Lunenburg Budget Project · Data architecture

The database, table by table

Every table, what it holds, and which years it covers.

tables

views

70,234

rows

983

columns

79%

columns described

School money and school subjects28 tables · 33,557 rows

The district's budget, its own funds, and what the town publishes about its schools.

Tier 1 · totalswhat was appropriated and what is left, by department

school_special_revenue_fy26_q3 school Tier 1 · totals fact 62 rowsone row per school fund

The schools' own funds at one moment (FY26 period 9): per fund, revenue received, salaries paid, other expenditure, encumbered, and the balance left. Dollars.

column	type	what it is
`fund`	TEXT	The MUNIS fund number. School money spans 13xx, 15xx, 22xx and 26xx-29xx. There is no single range that means school.
`name`	TEXT	The name as the source prints it. MUNIS truncates account names to ten characters.
`revenue`	TEXT	Money received into the fund. dollars Not the same quantity as spending, and the two must never be added.
`salaries`	TEXT	The salary portion of the fund's spending. dollars
`expenditure`	TEXT	The non-salary portion of the fund's spending. dollars
`encumbered`	TEXT	Committed by purchase order but not yet spent. dollars
`balance`	TEXT	not described yet

First 3 rows, in rowid order:

fund	name	revenue	balance
2748	FY24 117 EVIDENCE	-0.0	-0.0
2653	FY24 FAMILY & COMMUNITY #237	-0.0	433.7
2582	EECBG ENERGY EFFICIENCY GRANT	-0.0	316.48

total_expenses_history school Tier 1 · totals fact 28 rows 2014–2027one row per year per stage

One dollar total per year per stage — the district's own stated total expenses, as each document states it.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`stage`	TEXT	What the figure IS: proposed, settled, or actual. Rule 1. A rate measured from one stage to another is partly growth and partly the step between them.
`total`	TEXT	The row's total, as printed. dollars
`documents_disagree`	TEXT	Two documents state different figures for the same thing. Kept rather than resolved: which is right is often not established.

First 3 rows, in rowid order:

fy	stage	total	documents_disagree
2018	settled	6034260	1
2022	proposed	6008705	0
2026	settled	9169982	0

total_salaries_history school Tier 1 · totals fact 28 rows 2014–2027one row per year per stage

One dollar total per year per stage — the district's own stated total salaries.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`stage`	TEXT	What the figure IS: proposed, settled, or actual. Rule 1. A rate measured from one stage to another is partly growth and partly the step between them.
`total`	TEXT	The row's total, as printed. dollars
`documents_disagree`	TEXT	Two documents state different figures for the same thing. Kept rather than resolved: which is right is often not established.

First 3 rows, in rowid order:

fy	stage	total
2018	proposed	13374150
2020	proposed	14407125
2024	proposed	16226166

Tier 2 · categorieswhether a category — guidance, special education, transport — is over

budget_figure school the spine Tier 2 · categories fact 9,503 rows 2014–2027one row per line per year per stage per variant

Dollar amounts per budget LINE from the district's budget documents — no counts, no people. Each figure carries the stage it represents (proposed / settled / actual) and, for FY27, which of four whole budget scenarios it belongs to. FY2014–FY2027, 9,503 figures, plus a flag where two documents state different amounts for the same line-year.

Split on stage AND variant. FY27 carries four whole budgets.

column	type	what it is
`line_key`	TEXT	Our normalised join key for a budget line. Normalisation differs between budget_figure and workbook_figure — joining them on line_key matches only a fraction. Join on the normalised label instead.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`stage`	TEXT	What the figure IS: proposed, settled, or actual. Rule 1. A rate measured from one stage to another is partly growth and partly the step between them.
`variant`	TEXT	Which whole-budget scenario the figure belongs to. FY27 has four — Balanced, Core Budget, Level Service, Restoration. Without this you sum all four budgets together.
`value`	REAL	The figure, as printed. dollars
`documents_disagree`	INTEGER	Two documents state different figures for the same thing. Kept rather than resolved: which is right is often not established.
`doc_id`	TEXT	The archive path of the document this row's figures came from. Every fact row carries one. A figure that cannot name its document is not loaded.

First 3 rows, in rowid order:

line_key	label	fy	stage	variant	value	doc_id
hs librarian	H.S. Librarian	2027	proposed	Core Budget	55175.0	sources/district-budget/text/final-budget-d…
hs library books	H.S. Library Books	2017	proposed	empty	0.0	sources/district-budget/text/fy18-non-perso…
hs repair p e equipment	H.S. Repair P.E. Equipment	2025	proposed	empty	0.0	sources/district-budget/text/fy26-school-de…

workbook_figure school the spine Tier 2 · categories fact 3,261 rows 2023, 2024, 2025, 2026, 2027, 2029one row per worksheet row per year per column

The FY27 budget workbook unpivoted to one row per worksheet row × year × column, so every figure is quotable by its spreadsheet coordinate. column_kind says which printed column — final_budget, actual_to_date, encumbered_to_date. Dollars only.

every cell quotable by its row.

column	type	what it is
`row`	INTEGER	The worksheet row number in the source spreadsheet. Rule 13: quotes a coordinate, so a figure can be traced to its cell.
`line_key`	TEXT	Our normalised join key for a budget line. Normalisation differs between budget_figure and workbook_figure — joining them on line_key matches only a fraction. Join on the normalised label instead.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`column_kind`	TEXT	not described yet
`value`	REAL	The figure, as printed. dollars
`row_kind`	TEXT	not described yet
`doc_id`	TEXT	The archive path of the document this row's figures came from. Every fact row carries one. A figure that cannot name its document is not loaded.

First 3 rows, in rowid order:

row	line_key	fy	column_kind	value	row_kind	doc_id
139	m.s. general school supplies	2024	actual	7475.99	line	sources/budget-workbooks/fy27-proposals.xlsx
188	telephone maintenance/system	2026	actual_to_date	0.0	line	sources/budget-workbooks/fy27-proposals.xlsx
237	executive secretaries/supt.	2025	actual	184375.15	line	sources/budget-workbooks/fy27-proposals.xlsx

athletics_history school Tier 2 · categories fact 175 rows 2014–2026one row per athletics figure per year

Athletics money from BOTH sides, and these are three different quantities that must never be added: what the town appropriated from the general fund, what the revolving fund took IN, and what the revolving fund SPENT. Dollars per year with the basis and source document for each.

Appropriation, fund revenue and fund spending are three quantities. Never add two of them.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`side`	TEXT	not described yet
`item`	TEXT	The thing being priced or counted.
`amount`	TEXT	The figure, as printed. dollars
`basis`	TEXT	Whether a figure is published, contractual, statutory, our measurement or our assumption.
`source`	TEXT	Where the figure came from, as the extractor recorded it. A loose reference — a filename — not a resolved address.

First 3 rows, in rowid order:

fy	side	item	amount	basis	source
2016	revolving	Athletic Secretary	7796	actual	district-budget/docs/fy19-proposed-athletic…
2018	revolving	Track and Field Payment	40000	budget	district-budget/docs/fy19-proposed-athletic…
2020	general	Athletic Coaches	91601.0	actual	fy24-approved-budget.txt

budget_line school Tier 2 · categories dimension 688 rowsone row per budget line

The district's LINE LIST — no money at all. One row per budget line with its section, its function group and what kind of line it is. This is the dimension budget_figure and workbook_figure hang their dollars on.

with its section and function group.

column	type	what it is
`line_key`	TEXT	Our normalised join key for a budget line. Normalisation differs between budget_figure and workbook_figure — joining them on line_key matches only a fraction. Join on the normalised label instead.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`section`	TEXT	The budget book's own section heading.
`function_group`	TEXT	Our grouping of DESE function codes into readable categories.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.

First 3 rows, in rowid order:

line_key	label	section	function_group	kind
special ed transp system	Special Education Transp - System	NULL	NULL	NULL
curriculum adoption/system	Curriculum Adoption/System	EXPENSES	2110 - System Curriculum Adop	line
h.s. guidance secretary	H.S. Guidance Secretary	SALARIES	2710 - Guidance Salaries	line

line_history school Tier 2 · categories fact 9,503 rows 2014–2027one row per line per year per stage

Every budget line × year × stage as EXTRACTED, before disagreements between documents are resolved. Dollars. This is the raw reading; budget_figure is the settled one.

every year

column	type	what it is
`key`	TEXT	The row's join key.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`stage`	TEXT	What the figure IS: proposed, settled, or actual. Rule 1. A rate measured from one stage to another is partly growth and partly the step between them.
`variant`	TEXT	Which whole-budget scenario the figure belongs to. FY27 has four — Balanced, Core Budget, Level Service, Restoration. Without this you sum all four budgets together.
`value`	TEXT	The figure, as printed. dollars
`documents_disagree`	TEXT	Two documents state different figures for the same thing. Kept rather than resolved: which is right is often not established.
`source`	TEXT	Where the figure came from, as the extractor recorded it. A loose reference — a filename — not a resolved address.

First 3 rows, in rowid order:

key	label	fy	stage	variant	value	source
es principal asst prin	E.S. Principal/Asst. Prin.	2026	settled	empty	241000	final-budget-document.txt
hs nurses	H.S. Nurses	2017	proposed	empty	68100	fy18-personnel-supt-recommended-budget-revi…
plus total revolving fund	plus TOTAL Revolving Fund	2017	actual	empty	113819	fy19-proposed-athletics-budget.txt

lps_budget_lines school Tier 2 · categories fact 356 rowsone row per budget book line

The district's budget book in ITS OWN shape: one row per line with every year's column side by side (FY23 actual through the four FY27 scenarios) rather than unpivoted. Includes the workbook's own comments column, which is where the district explains a line to itself.

in its own shape and column order.

column	type	what it is
`row`	TEXT	The worksheet row number in the source spreadsheet. Rule 13: quotes a coordinate, so a figure can be traced to its cell.
`section`	TEXT	The budget book's own section heading.
`function_group`	TEXT	Our grouping of DESE function codes into readable categories.
`line_item`	TEXT	not described yet
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`fy23_actual`	TEXT	not described yet
`fy24_actual`	TEXT	not described yet
`fy25_actual`	TEXT	not described yet
`fy25_budget`	TEXT	not described yet
`fy26_final`	TEXT	not described yet
`fy26_actual_td`	TEXT	not described yet
`fy26_encumb_td`	TEXT	not described yet
`fy27_restoration`	TEXT	not described yet
`fy27_core`	TEXT	not described yet
`fy27_level_service`	TEXT	not described yet
`fy27_balanced`	TEXT	not described yet
`forecast_outyear`	TEXT	not described yet
`restoration_2_24_26`	TEXT	not described yet
`comments`	TEXT	not described yet

First 3 rows, in rowid order:

row	section	function_group	line_item	kind	fy23_actual	fy24_actual	fy25_actual	fy25_budget	fy26_final	fy26_actual_td	fy26_encumb_td	fy27_restoration	fy27_core	fy27_level_service	fy27_balanced	forecast_outyear	restoration_2_24_26	comments
50	EXPENSES	2210 - H.S. Principals Office	H.S. Graduation Expenses	line	7934.71	8250.65	8038.54	10000.0	10378.0	33.35	1616.65	10378.0	10378.0	10378.0	10378.0	10378.0	10378.0	empty
116	EXPENSES	2420 - P.S. Instr. Equipment	P.S. Science Supplies (includes STEM)	line	3329.04	6357.76	3332.24	5290.0	5222.0	0.0	556.21	3500.0	3500.0	3500.0	3500.0	3858.75	3500.0	empty
198	EXPENSES	4230 - M.S. Repairs	4230 - M.S. Repairs	line	empty	empty	empty	empty	empty	empty	empty	empty	empty	empty	empty	1800.0	empty	empty

ood_tuition_history school Tier 2 · categories fact 30 rows 2014–2027one row per year per stage

Out-of-district special education tuition in DOLLARS, split private vs collaborative, per year per stage. Not a count of children — placement_counts is that, and the two do not divide into each other.

private and collaborative.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`stage`	TEXT	What the figure IS: proposed, settled, or actual. Rule 1. A rate measured from one stage to another is partly growth and partly the step between them.
`private`	TEXT	not described yet
`collaborative`	TEXT	Children placed with an educational collaborative. count Children, not dollars.
`total`	TEXT	The row's total, as printed. dollars
`documents_disagree`	TEXT	Two documents state different figures for the same thing. Kept rather than resolved: which is right is often not established.

First 3 rows, in rowid order:

fy	stage	private	collaborative	total
2017	proposed	581345	74189	655534
2019	actual	706879	251616	958495
2021	proposed	589156	130113	719269

sped_para_history school Tier 2 · categories fact 26 rows 2014–2027one row per year per stage

Budgeted DOLLARS for special education paraprofessionals, split by school, per year per stage. A budget line, never a headcount.

A budget line, not a headcount.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`stage`	TEXT	What the figure IS: proposed, settled, or actual. Rule 1. A rate measured from one stage to another is partly growth and partly the step between them.
`ps`	TEXT	not described yet
`es`	TEXT	not described yet
`ms`	TEXT	not described yet
`hs`	TEXT	not described yet
`ace`	TEXT	not described yet
`total`	TEXT	The row's total, as printed. dollars
`documents_disagree`	TEXT	Two documents state different figures for the same thing. Kept rather than resolved: which is right is often not established.

First 3 rows, in rowid order:

fy	stage	ps	es	ms	hs	ace	total
2018	actual	169424	122049	214030	136132	0	641635
2021	actual	342946	129681	286383	136485	24850	920345
2026	settled	424416	225152	373590	275473	43742	1342373

sped_teacher_history school Tier 2 · categories fact 21 rows 2017–2027one row per year per stage

Budgeted DOLLARS for special education teachers, split by school, per year per stage. A budget line, never a headcount.

A budget line, not a headcount.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`stage`	TEXT	What the figure IS: proposed, settled, or actual. Rule 1. A rate measured from one stage to another is partly growth and partly the step between them.
`ps`	TEXT	not described yet
`es`	TEXT	not described yet
`ms`	TEXT	not described yet
`hs`	TEXT	not described yet
`ace`	TEXT	not described yet
`total`	TEXT	The row's total, as printed. dollars
`documents_disagree`	TEXT	Two documents state different figures for the same thing. Kept rather than resolved: which is right is often not established.

First 3 rows, in rowid order:

fy	stage	ps	es	ms	hs	ace	total
2021	actual	516646	362179	407895	378784	57649	1723153
2021	settled	516646	362179	407895	378784	57649	1723153
2022	actual	541721	342512	407145	454812	31049	1777239

sped_transport_history school Tier 2 · categories fact 25 rows 2015–2027one row per year per stage

Budgeted DOLLARS for special education transportation, per year per stage.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`stage`	TEXT	What the figure IS: proposed, settled, or actual. Rule 1. A rate measured from one stage to another is partly growth and partly the step between them.
`system`	TEXT	not described yet
`total`	TEXT	The row's total, as printed. dollars
`documents_disagree`	TEXT	Two documents state different figures for the same thing. Kept rather than resolved: which is right is often not established.

First 3 rows, in rowid order:

fy	stage	system	total	documents_disagree
2020	actual	223350	223350	1
2022	proposed	297843	297843	0
2023	settled	329360	329360	0

Tier 3 · accountswhat one account spent

fund_1301_cash_journal school Tier 3 · accounts fact 277 rows 2024–2026one row per transaction

TRANSACTIONS — the athletics revolving fund line by line: effective and posting dates, journal, references, vendor, amount, running balance, check and warrant numbers, and the comments the clerk typed. The only transaction-level data in the archive, obtained by records request.

from a records request.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`source_row`	TEXT	not described yet
`eff_date`	TEXT	not described yet
`post_date`	TEXT	not described yet
`src`	TEXT	not described yet
`src_meaning`	TEXT	not described yet
`period`	TEXT	MUNIS accounting period: a point in time INSIDE the fiscal year. A period is not a stage. Period 9 and period 12 are two different reports of the same year; adding them adds one report to the other.
`journal`	TEXT	not described yet
`ref1`	TEXT	not described yet
`po_ref2`	TEXT	not described yet
`ref3`	TEXT	not described yet
`reference`	TEXT	not described yet
`amount`	TEXT	The figure, as printed. dollars
`running_balance_derived`	TEXT	not described yet
`check_no`	TEXT	not described yet
`warrant`	TEXT	not described yet
`voucher`	TEXT	not described yet
`vendor`	TEXT	not described yet
`comments`	TEXT	not described yet

First 3 rows, in rowid order:

fy	source_row	eff_date	post_date	src	src_meaning	period	journal	ref1	po_ref2	ref3	reference	amount	running_balance_derived	warrant	voucher	vendor	comments
2024	20	2024-05-23	2024-05-23	APP	accounts payable — warrant disbursement	11	476	58 24	empty	empty	58 24	-5504.91	49527.63	empty	empty	empty	empty
2025	84	2024-08-07	2024-08-14	GRV	general journal reversal / correction	2	106	12WP24	empty	empty	2412001378	-586.50	17029.10	empty	empty	empty	TO CORRECT FY
2026	13	2026-04-02	2026-04-02	APP	accounts payable — warrant disbursement	10	40	44 26	empty	empty	44 26	-9900.00	160044.80	empty	empty	empty	empty

Not about moneyCounts, names, dates and documents — everything this archive holds that is not a dollar figure

athletic_fee_schedule school fact 31 rows 2024–2027one row per fee per year

The published fee a family pays to play: dollars per fiscal year × school level × item, with the date it was set and the document that set it. Includes family caps and sibling discounts.

by level and year.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`school_year`	TEXT	not described yet
`level`	TEXT	School level, where the row is about athletics rather than accounts. Means something different in `account` — see that table's entry.
`item`	TEXT	The thing being priced or counted.
`amount`	TEXT	The figure, as printed. dollars
`unit`	TEXT	What the figure is counted in.
`set_on`	TEXT	The date the rate or fee was set.
`source`	TEXT	Where the figure came from, as the extractor recorded it. A loose reference — a filename — not a resolved address.
`source_file`	TEXT	The file the figure was read from.
`source_ref`	TEXT	Where in the source the figure sits — a cell, a line, a page. Rule 13: a coordinate, so a figure can be checked rather than taken.
`verified`	TEXT	not described yet

First 3 rows, in rowid order:

fy	school_year	level	item	amount	unit	set_on	source	source_file	source_ref	verified
2025	2024-25	HS	third_child	85.00	per student per sport per season	empty	District athletics workbook, by sport (Town…	munis-ledgers/account-details/athletics-by-…	Spring!L3	Spring!L3
2025	2024-25	MS	reduced_fee	26.00	per student per sport per season	empty	District athletics workbook, by sport (Town…	munis-ledgers/account-details/athletics-by-…	Spring!U3	Spring!U3
2026	2025-26	HS	full_pay	325.00	per student per sport per season	2025-02-26	School Committee minutes, 26 February 2025 …	minutes/text/school-committee/2025-02-26-mi…	up to $325	quoted in source

athletics_by_sport school fact 960 rows 2024–2026one row per sport per season per metric

Per sport × season × level: participation COUNTS and DOLLAR figures mixed in one table, told apart by metric. raw keeps the value exactly as printed and is_numeric says whether it parsed, because some cells are text.

column	type	what it is
`season`	TEXT	Which athletic season.
`level`	TEXT	School level, where the row is about athletics rather than accounts. Means something different in `account` — see that table's entry.
`sport`	TEXT	The sport.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`metric`	TEXT	Which quantity this row measures.
`value`	TEXT	The figure, as printed. dollars
`raw`	TEXT	The value exactly as printed, before parsing. Rule 13: what the source says, beside what we made of it.
`is_numeric`	TEXT	Whether the raw value parsed as a number.
`cell`	TEXT	The spreadsheet cell the figure came from. A coordinate, so the figure can be checked.

First 3 rows, in rowid order:

season	level	sport	fy	metric	value	raw	is_numeric	cell
Fall	HS	Boys CC	2024	Costs for all 3 Seasons	empty	MIAA 1/3 4550. -1516.67	0	AX6
Spring	HS	Baseball	2025	Expected Fees Received	4592.5	4592.5	1	X6
Spring	HS	Softball	2025	Full Pay	21.0	21	1	F5

athletics_by_sport_reconciliation school provenance 342 rows 2024–2026one row per printed total

Not data — checks. For each total the source prints, what our rows sum to, the difference, and whether it ties.

column	type	what it is
`season`	TEXT	Which athletic season.
`scope`	TEXT	Whose money or whose subject the table is about: school, town, both, state or archive. `both` is not a hedge — the ledger and the chart of accounts genuinely carry every department, and splitting them would be inventing a division the source does not make.
`metric`	TEXT	Which quantity this row measures.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`printed`	TEXT	The total the source itself prints. dollars
`printed_cell`	TEXT	Where that printed total sits.
`printed_raw`	TEXT	The printed total exactly as it appears.
`summed_from_rows`	TEXT	What our rows add to. dollars
`difference`	TEXT	Ours less the source's. dollars
`ties`	TEXT	Whether the two agree.

First 3 rows, in rowid order:

season	scope	metric	fy	printed	printed_cell	printed_raw	summed_from_rows	ties
Winter	ALL	Equipment	2024	2567.2	BH23	2567.2	2567.2	1
Winter	HS	Coaches	2025	36133.0	AQ16	36133	36133.0	1
Winter	HS	Full Pay	2024	118.0	E16	118	118.0	1

grant_award school fact 79 rows 2020, 2021, 2022, 2023, 2024, FY21-24one row per grant per year

Grant AWARDS: name, amount in dollars, who holds it, and the fiscal year — except one ESSER award that genuinely spans FY21–24, which is why fy is TEXT here.

by name and amount.

column	type	what it is
`fy`	TEXT	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`name`	TEXT	The name as the source prints it. MUNIS truncates account names to ten characters.
`amount`	REAL	The figure, as printed. dollars
`owner`	TEXT	Who holds the grant.
`documents_disagree`	INTEGER	Two documents state different figures for the same thing. Kept rather than resolved: which is right is often not established.
`doc_id`	TEXT	The archive path of the document this row's figures came from. Every fact row carries one. A figure that cannot name its document is not loaded.
`url`	TEXT	The publisher's own address for the document.
`sha256`	TEXT	Checksum of our copy. Here because a file can be replaced at the same URL without the URL changing.

First 3 rows, in rowid order:

fy	kind	name	amount	owner	documents_disagree	doc_id	url	sha256
2022	federal	Special Education, 240 Grant	399152.0	Director of Special Services	1	fy25-superintendent-39-s-budget-update.txt	https://drive.google.com/file/d/1yJNhIyBLVT…	9169e2700def0c1a2b6bebbc55d4e7f737ea5c3a735…
2022	federal	Special Education, AARP	8229.0	Director of Special Services	0	fy25-superintendent-39-s-budget-update.txt	https://drive.google.com/file/d/1yJNhIyBLVT…	9169e2700def0c1a2b6bebbc55d4e7f737ea5c3a735…
2023	state	Grant	25000.0	empty	0	fy25-superintendent-39-s-budget-update.txt	https://drive.google.com/file/d/1yJNhIyBLVT…	9169e2700def0c1a2b6bebbc55d4e7f737ea5c3a735…

grants_history school fact 79 rows 2020, 2021, 2022, 2023, 2024, FY21-24one row per grant per year

The same awards with their source addresses and checksums, plus whether documents disagree about an amount and whether it is ESSER money.

fy is TEXT — see grant_award.

column	type	what it is
`fy`	TEXT	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`name`	TEXT	The name as the source prints it. MUNIS truncates account names to ten characters.
`amount`	TEXT	The figure, as printed. dollars
`owner`	TEXT	Who holds the grant.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`doc`	TEXT	not described yet
`esser`	TEXT	not described yet
`disagreement`	TEXT	not described yet
`source_url`	TEXT	not described yet
`our_copy`	TEXT	not described yet
`sha256`	TEXT	Checksum of our copy. Here because a file can be replaced at the same URL without the URL changing.

First 3 rows, in rowid order:

fy	kind	name	amount	owner	page	doc	disagreement	source_url	our_copy	sha256
2021	federal	Title IV, 309 Grant	18466.0	Director of Teaching & Learning	35	fy25-superintendent-39-s-budget-update.txt	empty	https://drive.google.com/file/d/1yJNhIyBLVT…	https://lunenburgbudgetproject.org/docs/dis…	9169e2700def0c1a2b6bebbc55d4e7f737ea5c3a735…
2023	federal	Title IV, Grant, 309 Grant	12904.0	empty	16	fy24-superintendent-39-s-recommended-budget…	empty	https://drive.google.com/file/d/1yJNhIyBLVT…	https://lunenburgbudgetproject.org/docs/dis…	9169e2700def0c1a2b6bebbc55d4e7f737ea5c3a735…
2024	state	Arts & Cultural Vitality, 718 Grant	5200.0	Director of Teaching & Learning	32	fy25-superintendent-39-s-budget-update.txt	empty	https://drive.google.com/file/d/1yJNhIyBLVT…	https://lunenburgbudgetproject.org/docs/dis…	9169e2700def0c1a2b6bebbc55d4e7f737ea5c3a735…

line_history_coverage school provenance 86 rowsone row per source document

Not figures — a report on extraction. Per source document: how many data rows, how many carried a usable layout, how many fell short of the expected column count, and why a document was or was not usable.

column	type	what it is
`document`	TEXT	The source file the figure was read from.
`doc_year`	TEXT	not described yet
`data_rows`	TEXT	not described yet
`rows_with_layout`	TEXT	not described yet
`rows_short_of_columns`	TEXT	not described yet
`figures`	TEXT	not described yet
`header_line`	TEXT	not described yet
`header_years`	TEXT	not described yet
`header_kinds`	TEXT	not described yet
`doc_date`	TEXT	not described yet
`covers`	TEXT	not described yet
`reason`	TEXT	not described yet

First 3 rows, in rowid order:

document	doc_year	data_rows	header_line	header_years	header_kinds	doc_date	covers	reason
fy21-recommended-budget-presentation-4-22-2…	2021	30	empty	FY20 Budget FY21 Budget	empty	2020-04-22	FY2020 FY2021	header wrapped: line 101 states 2 fiscal ye…
fy23-quarterly-budget-update.txt	2023	1	empty	FY23 Quarterly Budget Update November 11, 2…	empty	2022-11-11	FY2023	single-column table: line 3 states one fisc…
school-committee-approved-fy26-budget-perso…	2026	2	empty	School Committee Approved FY26 Budget	empty	empty	FY2026	single-column table: line 2 states one fisc…

line_history_disagreements school provenance 1,497 rows 2014–2027one row per disagreeing reading

One row per disagreeing READING: where two documents state different dollar amounts for the same line-year, both values are kept with their source and the spread between them. Nothing here is resolved.

Kept rather than resolved. Which document is right is often not established.

column	type	what it is
`key`	TEXT	The row's join key.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`stage`	TEXT	What the figure IS: proposed, settled, or actual. Rule 1. A rate measured from one stage to another is partly growth and partly the step between them.
`variant`	TEXT	Which whole-budget scenario the figure belongs to. FY27 has four — Balanced, Core Budget, Level Service, Restoration. Without this you sum all four budgets together.
`source`	TEXT	Where the figure came from, as the extractor recorded it. A loose reference — a filename — not a resolved address.
`value`	TEXT	The figure, as printed. dollars
`is_kept`	TEXT	not described yet
`spread`	TEXT	not described yet
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.

First 3 rows, in rowid order:

key	label	fy	stage	variant	source	value	is_kept	spread	kind
hs general school supplies	H.S. General School Supplies	2025	proposed	empty	fy26-school-dept-approved-budget-3-12-25.txt	16000	1	5500	documents
maintenance system salaries	Maintenance/System Salaries	2019	proposed	empty	fy19-supt-proposed-personnel-budget.txt	292184	1	12094	documents
ps guidance supplies	P.S. Guidance Supplies	2025	proposed	empty	3-12-25-town-manager-39-s-budget.txt	250	0	50	documents

placement_counts school fact 15 rows 2011–2025one row per year

CHILDREN, not dollars. Students placed outside the district, split collaborative / day / residential, measured on 1 March each year from SIMS Report 7. FY2013–FY2025. Carries its own checks: whether the parts sum to the total, and whether each year agrees with the next year's restatement of it.

split collaborative/day/residential.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`as_of`	TEXT	The date the count was taken. Placement counts are measured on 1 March, not at year end.
`total`	TEXT	The row's total, as printed. dollars
`collaborative`	TEXT	Children placed with an educational collaborative. count Children, not dollars.
`day`	TEXT	Children in a day placement. count Children, not dollars.
`residential`	TEXT	Children in a residential placement. count Children, not dollars.
`collaborative_within_day`	TEXT	Whether the collaborative count is inside the day count.
`collaborative_basis`	TEXT	How the collaborative figure was arrived at.
`report_says_prior_year`	TEXT	The prior-year figure this report restates. Each year states last year's number, which checks the chain.
`parts_tie`	TEXT	Whether the parts sum to the printed total.
`chain_agrees`	TEXT	Whether this year's report agrees with last year's about last year.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`document`	TEXT	The source file the figure was read from.
`checks`	TEXT	Which checks this row passed.

First 3 rows, in rowid order:

fy	as_of	total	collaborative	day	residential	collaborative_within_day	collaborative_basis	report_says_prior_year	parts_tie	chain_agrees	page	document	checks
2011	2011-03-01	15	empty	13	2	1	subset of day placements	empty	yes	n/a	35	FY2011	parts 13+2=15 ties to stated 15
2017	2017-03-01	14	2	8	4	empty	parallel category	19	yes	yes	102	FY2017	parts 2+8+4=14 ties to stated 14; report sa…
2023	2023-03-01	7	3	3	1	empty	parallel category	8	yes	yes	120	FY2023	parts 3+3+1=7 ties to stated 7; report says…

report_enrollment_mcas school extract 337 rows 2011–2025one row per printed line

Enrollment counts and MCAS results as the annual reports print them. Mixed measures in ordinal columns; read column_meaning before using any of them.

v1..v8 are ORDINALS. Split on status.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
enrollment_mcas	2012	FY2012	33	1	Health Promotion	PK	empty	row	empty	empty	empty	empty	Confirmed 25 \| Communicable Disease Number …	43	41	empty	empty	empty	empty	empty	empty	2	empty	empty	empty	no check	no total printed
enrollment_mcas	2014	FY2014	120	empty	empty	empty	empty	page_number	the page number, not a line of the table	empty	empty	empty	LUNENBURG PRIMARY SCHOOL \| 2014-2015 \| The …	120	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	empty	no check	no total printed
enrollment_mcas	2024	FY2024	109	1	empty	Grade 6 ELA	empty	row	empty	empty	empty	empty	trailing figures	49	40	empty	empty	empty	empty	empty	empty	2	empty	empty	empty	no check	no total printed

report_monty_tech school extract 70 rows 2017–2017one row per printed line

Figures for Montachusett Regional Vocational Technical School as the annual reports print them. A DIFFERENT DISTRICT: Lunenburg pays it an assessment, and its figures must never be added to Lunenburg's own schools.

Monty Tech is a DIFFERENT DISTRICT. Never add it to Lunenburg's schools.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
monty_tech	2017	FY2017	143	empty	Debt Service …	Retired Employee Benefits 859,100	empty	row	empty	empty	empty	empty	General Fund- Income and Expense Summary \| …	5250.00	821348.00	969651.00	1004045.00	1068842.00	64797.00	empty	empty	6	empty	empty	empty	no check	no total printed
monty_tech	2017	FY2017	143	empty	Debt Service …	FUNCTION 8000 DEBT RETIREMENT 1,213,204	empty	subtotal	subtotal: equals the 2 rows above it, in 4 …	empty	empty	empty	General Fund- Income and Expense Summary \| …	empty	1210099.00	1208628.00	1204669.00	1201263.00	-3406.00	empty	empty	5	empty	empty	empty	no check	no total printed
monty_tech	2017	FY2017	143	empty	Debt Service …	Total $24,785,240 $	empty	subtotal	empty	empty	empty	empty	General Fund- Income and Expense Summary \| …	empty	25548074.00	26228813.00	26364958.00	26895680.00	530722.00	empty	empty	5	yes	empty	empty	no check	no total printed

role_classification school dimension 1,030 rowsone row per printed roster title

Our classification of each printed roster title into a role category and grade, with how it was classified. Judgement, recorded so it can be disputed.

column	type	what it is
`role_raw`	TEXT	The job title exactly as printed.
`grade_or_dept`	TEXT	not described yet
`rows`	TEXT	How many rows the extract produced. count
`role_category`	TEXT	not described yet
`role_grade`	TEXT	not described yet
`classified_by`	TEXT	not described yet

First 3 rows, in rowid order:

role_raw	grade_or_dept	rows	role_category	role_grade	classified_by
empty	Technology IT:	9	technology	empty	heading-technology
Admin. Assistant	empty	3	secretary	empty	clerical
Occup. Therapist	Specialists	1	specialist	empty	heading-specialists

staff_position_map school dimension 534 rowsone row per printed roster title

A lookup, not data: each printed roster title, how many times it was seen, and the position it maps to.

column	type	what it is
`role_as_printed`	TEXT	not described yet
`position`	TEXT	The job, classified by us from the printed title. A roster carries no FTE, so a 0.4 music teacher and a full-timer are one row each.
`times_seen`	TEXT	not described yet

First 3 rows, in rowid order:

role_as_printed	position	times_seen
(TLC) Theraputic Learning Center	Special Education Teacher	1
A	empty	2
Music MS	Specialist Teacher	1

staff_roster_counts school fact 699 rows 2011–2025one row per school per position per year

Counts of the names above, per school × position × year. A count of names the town printed. It is not a staffing level and cannot be turned into one.

A count of NAMES, with no FTE and no funding source. It is not a staffing level.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`school`	TEXT	Which school, as the roster prints it.
`position`	TEXT	The job, classified by us from the printed title. A roster carries no FTE, so a 0.4 music teacher and a full-timer are one row each.
`count`	TEXT	How many. count

First 3 rows, in rowid order:

fy	school	position	count
2011	primary	Classroom Teacher	9
2012	high	(unmapped)	28
2019	primary	(unmapped)	13

staff_roster_entries school fact 3,815 rows 2011–2025one row per person per year

PEOPLE — one row per name the town printed, per year, with the school, the page, the printed job title and our classification of it. FY2011–FY2025, 3,815 entries. No FTE, no salary, no funding source, and undated within the year.

name by name

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`school`	TEXT	Which school, as the roster prints it.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`name`	TEXT	The name as the source prints it. MUNIS truncates account names to ten characters.
`role_raw`	TEXT	The job title exactly as printed.
`position`	TEXT	The job, classified by us from the printed title. A roster carries no FTE, so a 0.4 music teacher and a full-timer are one row each.
`position_from`	TEXT	not described yet
`grade_or_dept`	TEXT	not described yet
`line`	TEXT	not described yet
`document`	TEXT	The source file the figure was read from.

First 3 rows, in rowid order:

fy	school	page	name	role_raw	position	position_from	grade_or_dept	line	document
2014	high	110	Joseph Adamowicz	empty	Paraprofessional	section heading	Paraprofessionals:	7	4120-fy-2014-annual-town-report.pdf
2016	primary	110	Sheila Bilotta	Teacher	Classroom Teacher	printed role	K3	28	4123-fy-2016-annual-town-report.pdf
2019	middle	131	Kathleen Hyatt	empty	Paraprofessional	section heading	Paraprofessionals	25	4126-fy-2019-annual-town-report.pdf

Town money and town subjects18 tables · 17,762 rows

The other 67 departments, the receipts that fund them, and what the annual reports print about the town.

Tier 1 · totalswhat was appropriated and what is left, by department

annual_report_receipts town Tier 1 · totals extract 1,137 rows 2011–2023one row per revenue line per year

Town REVENUE — every receipt line the annual reports print, by name and amount, per year, FY2011–FY2023. Only five years are checked against the report's own printed total.

as printed.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`source`	TEXT	Where the figure came from, as the extractor recorded it. A loose reference — a filename — not a resolved address.
`amount`	TEXT	The figure, as printed. dollars
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`document`	TEXT	The source file the figure was read from.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.

First 3 rows, in rowid order:

fy	source	amount	status	document	page
2015	CEMETERIES BURIALS	4000.00	checked	4121-fy-2015-annual-town-report.pdf	22
2017	MEDICAL RECORDS	135.00	checked	4124-fy-2017-annual-town-report.pdf	22
2022	DOG LICENSES	10670.00	checked	4129-fy-2022-annual-town-report.pdf	26

report_appropriations town Tier 1 · totals extract 4,665 rows 2011 … 2025 (14 years)one row per printed line

APPROPRIATIONS as the annual reports print them — the largest extract here, 4,665 rows across sixteen years. Values sit in ORDINAL columns (v1..v8) whose meaning changes page to page; column_meaning names them where the table states an identity that fixes them, and says not established where it does not.

v1..v8 are ORDINALS. Read column_meaning and split on status.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
appropriations	2014	FY2014	141	1	Health & Sanitation	62 Town Highway Garage	empty	row	empty	empty	omnibus-budget	empty	GENERAL FUND APPROPRIATIONS \| SUMMARY & CLA…	1000.00	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	v1=voted (from the shape of the column, not…	check failed	voted: 28,979,584.00 vs 29,079,384.00 (-99,…
appropriations	2015	FY2015	28	empty	UNCLASSIFIED	Choice Tuitions	empty	row	empty	empty	accountant-schedule	empty	TOWN OF LUNENBURG \| GENERAL FUND APPROPRIAT…	531000.00	531000.00	524349.00	6651.00	empty	empty	empty	empty	4	empty	empty	v1=appropriated \| v2=available \| v3=expende…	check failed	appropriated: 31,922,571.97 vs 32,216,220.5…
appropriations	2016	FY2016	144	1	Unclassified	Group Health Insurance	empty	row	empty	empty	omnibus-budget	empty	TOWN OF LUNENBURG \| GENERAL FUND APPROPRIAT…	1907185.00	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	v1=voted (from the shape of the column, not…	check failed	voted: 33,802,373.00 vs 33,802,328.00 (+45.…

revenue_history town Tier 1 · totals fact 504 rows 2014, 2015, 2017, 2018, 2022one row per revenue source per year

The checked subset of the above joined to our classification: the printed name, our normalised key, what KIND of revenue it is (levy, state aid, local receipt, transfer, one-time), the amount, and the document and page it came from. Five years.

with our classification of what KIND it is.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`printed_name`	TEXT	not described yet
`source_key`	TEXT	not described yet
`class`	TEXT	Our classification of what KIND of revenue this is. Says where the money came FROM. It does NOT say what the money went on to pay for — general fund money is fungible and no document apportions it.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`amount`	REAL	The figure, as printed. dollars
`document`	TEXT	The source file the figure was read from.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.

First 3 rows, in rowid order:

fy	printed_name	source_key	class	label	amount	document	page
2014	DISHONORED CHECK FEE	DISHONOREDCHECKFEE	local	DISHONORED CHECK FEE	475.0	4120-fy-2014-annual-town-report.pdf	22
2017	POLICE PERMITS	POLICEPERMITS	local	POLICE PERMITS	5012.5	4124-fy-2017-annual-town-report.pdf	22
2017	SUBDIVISION/FORM A FILING FEES	SUBDIVISIONFORMAFILINGFEES	local	SUBDIVISION/FORM A FILING FEES	2100.0	4124-fy-2017-annual-town-report.pdf	22

special_revenue_funds town Tier 1 · totals extract 2,387 rows 2011–2025one row per fund per year

Every fund OUTSIDE the general appropriation, FY2011–FY2025: balance brought forward, receipts, disbursements, balance carried forward. Grants, revolving funds and gifts that pay for real staff and appear nowhere in the appropriation. Dollars.

fifteen years

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`fund`	TEXT	The MUNIS fund number. School money spans 13xx, 15xx, 22xx and 26xx-29xx. There is no single range that means school.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`is_subtotal`	TEXT	This row is a total of other rows, not a fund or a line of its own. Must be excluded from column sums or the rows it summarises are counted twice.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`columns_tying`	TEXT	not described yet
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

fy	edition	group	fund	page	is_subtotal	columns_as_printed	v1	v2	v3	v4	v5	v6	n_values	ruler_spanned	row_check	derived_cell	columns_tying	status	reconciliation
2012	FY2012	LIBRARY	Water Enterprise Fund	70	empty	empty	12380.65	17308.06	15950.86	13737.85	empty	empty	4	empty	fund balances: forward + receipts - disburs…	empty	0	check failed	v1: 2,897,297.02 vs printed 2,766,954.67 (+…
2016	FY2016	SCHOOL DEPARTMENT	Extended Day Revolving Fund	33	empty	FISCAL YEAR ENDING 6/30/16 ; FORWARD \| TOTA…	112884.10	300590.66	246943.94	166530.82	empty	empty	4	empty	fund balances: forward + receipts - disburs…	empty	0	check failed	v1: 2,711,371.31 vs printed 2,704,480.72 (+…
2018	FY2018	GENERAL GOVERNMENT	Sewer Gift Fund	35	empty	FISCAL YEAR ENDING 6/30/18 ; FORWARD \| TOTA…	387.09	empty	empty	387.09	empty	empty	2	empty	empty	empty	2	check failed	v1: 3,560,171.36 vs printed 3,564,136.36 (-…

town_ledger_fy26_q3 town Tier 1 · totals fact 67 rowsone row per department

The town's 67 departments at FY26 period 9: appropriation, transfers, revised, expended, encumbered, available and percent used. EXPENSE side only, department grain only.

by department.

column	type	what it is
`dept`	TEXT	The MUNIS department number. A department is not a purpose: school retiree health sits in the insurance department.
`name`	TEXT	The name as the source prints it. MUNIS truncates account names to ten characters.
`original`	TEXT	The appropriation as originally voted. dollars
`transfers`	TEXT	Money moved into or out of the appropriation after it was voted. dollars This is why `original` and `revised` differ.
`revised`	TEXT	The appropriation after transfers. dollars
`expended`	TEXT	Spent. dollars
`encumbered`	TEXT	Committed by purchase order but not yet spent. dollars
`available`	TEXT	Revised less expended less encumbered. dollars
`pct_used`	TEXT	Share of the revised budget used. percent

First 3 rows, in rowid order:

dept	name	original	revised	expended	encumbered	available	pct_used
525	PHYSICALS	3500.0	3500.0	416.0	234.0	2850.0	18.6
754	LOAN ADMINISTRATIVE FEES	1230.0	1230.0	1230.47	0.0	0.0	100.0
825	STATE ASSESSMENTS	1028529.0	1028529.0	765004.0	0.0	263525.0	74.4

Not about moneyCounts, names, dates and documents — everything this archive holds that is not a dollar figure

ballot_questions town fact 7 rowsone row per ballot question

VOTES, not dollars — though each carries the amount at stake. Per ballot question: type, purpose, yes / no / blank counts, registered voters, turnout and the result.

column	type	what it is
`date`	TEXT	not described yet
`election`	TEXT	not described yet
`question`	TEXT	not described yet
`type`	TEXT	not described yet
`amount`	TEXT	The figure, as printed. dollars
`purpose`	TEXT	not described yet
`yes`	TEXT	not described yet
`no`	TEXT	not described yet
`blanks`	TEXT	not described yet
`total`	TEXT	The row's total, as printed. dollars
`registered`	TEXT	not described yet
`turnout_pct`	TEXT	not described yet
`result`	TEXT	not described yet
`fy_report`	TEXT	not described yet
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`read_by`	TEXT	Which instrument produced the figures: `vision` for a page read directly. A reading, whichever instrument made it. It is checked by the printed total, not trusted for its source.
`checks`	TEXT	Which checks this row passed.

First 3 rows, in rowid order:

date	election	question	type	amount	purpose	yes	no	blanks	total	registered	turnout_pct	result	fy_report	page	read_by	checks
2014-01-11	Special Town Election	QUESTION 1. DEBT EXCLUSION	Proposition 2½ debt exclusion	72975321	Bonds to construct a new Lunenburg Middle/H…	1414	543	0	1957	7059	27.7	PASSED	2014	152	plain	YES 413+234+323+444=1414; NO 163+111+140+12…
2016-05	Annual Town Election	QUESTION 1 – Pavement Management Plan	Proposition 2½ debt exclusion	4000000	Pavement Management Plan — town roads	181	127	55	363	empty	empty	PASSED	2016	193	plain	YES 58+27+34+62=181; NO 28+36+32+31=127; Bl…
2024-05-18	Annual Town Election	QUESTION 1. OVERRIDE	Proposition 2½ override	empty	Override approved by voters in May 2024; Ar…	753	664	empty	empty	empty	empty	PASSED	2024	174	plain	YES 189+149+157+258=753; NO 195+118+175+176…

capital_funding_history town fact 10 rows 2017–2026one row per year

Per year, how capital spending was PAID FOR: free cash, taxation, unexpended prior-year capital, other. Dollars.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`total`	TEXT	The row's total, as printed. dollars
`free_cash`	TEXT	not described yet
`taxation`	TEXT	not described yet
`unexpended_prior_year_capital`	TEXT	not described yet
`other`	TEXT	not described yet

First 3 rows, in rowid order:

fy	total	free_cash	taxation	unexpended_prior_year_capital
2017	619475.00	250000.00	349023.05	20451.95
2023	2162849.30	983034.39	967725.00	212089.91
2024	1039010.00	467268.98	556789.09	14951.93

capital_plan_fy27 town fact 22 rowsone row per project

The FY27 capital plan as proposed: project, department, cost, running cumulative, whether it was funded and from where.

ranked.

column	type	what it is
`rank`	TEXT	not described yet
`dept`	TEXT	The MUNIS department number. A department is not a purpose: school retiree health sits in the insurance department.
`project`	TEXT	not described yet
`cost`	TEXT	not described yet
`cumulative`	TEXT	not described yet
`funded`	TEXT	not described yet
`funding`	TEXT	not described yet

First 3 rows, in rowid order:

rank	dept	project	cost	cumulative	funded	funding
2	Police	Patrol Cruiser (replacement)	68627.00	418627.00	yes	free_cash_or_taxation
12	IT	Adult Activity Center Network Upgrade	20405.00	1830203.00	yes	free_cash_or_taxation
19	Schools	Heat for THES Maintenance Garage	60000.00	2718608.00	no	unfunded

free_cash_proof town fact 630 rowsone row per line of the state's proof

The state's own line-by-line proof of certified free cash, for nine towns — every component with its role in the calculation. Dollars.

for nine towns.

column	type	what it is
`town`	TEXT	not described yet
`year`	TEXT	not described yet
`line`	TEXT	not described yet
`amount`	TEXT	The figure, as printed. dollars
`role`	TEXT	What kind of table it is: fact, dimension, extract, classification, provenance or derived. Facts carry the measurements. Everything else describes, extracts, classifies or checks them.
`source_file`	TEXT	The file the figure was read from.
`source_ref`	TEXT	Where in the source the figure sits — a cell, a line, a page. Rule 13: a coordinate, so a figure can be checked rather than taken.
`label_ref`	TEXT	not described yet

First 3 rows, in rowid order:

town	year	line	amount	role	source_file	source_ref	label_ref
Lunenburg	2022	Free Cash Certified Prior Year	2666962.00	prior_year_certified	state-dls/free-cash-proof-lunenburg.xlsx	Sheet1!C4	Sheet1!A4
Townsend	2024	Add Overlay Surplus Closed (CL#13)	0.00	component	state-dls/free-cash-proof-townsend.xlsx	Sheet1!E13	Sheet1!A13
Uxbridge	2022	Revenue Deficits	0.00	component	state-dls/free-cash-proof-uxbridge.xlsx	Sheet1!C6	Sheet1!A6

report_capital_projects town extract 616 rows 2011–2025one row per printed line

Capital project funds as the annual reports print them: appropriated, expended, encumbered and balance, per article.

v1..v8 are ORDINALS. Split on status.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
capital_projects	2016	FY2016	36	empty	empty	Technology, School Department	empty	row	empty	empty	empty	empty	CAPITAL PROJECT FUNDS FY 2016 \| APPROPRIATE…	30000.00	28228.93	1771.07	empty	empty	empty	empty	empty	3	empty	empty	empty	check failed	v1: 1,207,111.79 vs 11,120,094.93 (-9,912,9…
capital_projects	2023	FY2023	44	empty	empty	Public Safety Building Engineering Design f…	empty	row	empty	empty	empty	empty	CAPITAL PROJECT FUNDS FY 2023 \| EXPENDED/EN…	0.00	25000.00	empty	empty	empty	empty	empty	empty	2	empty	empty	empty	check failed	v1: 542,843.30 vs 843,181.55 (-300,338.25) …
capital_projects	2025	FY2025	30	empty	empty	3114 11/23 TC Passios Roof	empty	row	empty	empty	empty	empty	CAPITAL PROJECT FUND BALANCE DETAIL \| TOWN …	300000.00	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	empty	no check	no total printed

report_debt town extract 354 rows 2011, 2017, 2019, 2020, 2021, 2023, 2024, 2025one row per printed line

Outstanding DEBT and repayment schedules as the annual reports print them. Dollars, town-wide — school building debt is inside these figures and is not separated out.

v1..v8 are ORDINALS. Split on status.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
debt	2020	FY2020	45	empty	empty	PRINCIPAL	empty	row	empty	empty	empty	empty	TOWN OF LUNENBURG DEBT REPAYMENT SCHEDULE A…	965000.00	1070000.00	18510000.00	empty	empty	empty	empty	empty	3	empty	empty	empty	no check	no total printed
debt	2020	FY2020	45	empty	empty	PRINCIPAL	empty	row	empty	empty	empty	empty	TOWN OF LUNENBURG DEBT REPAYMENT SCHEDULE A…	empty	empty	1045000.00	empty	empty	empty	empty	empty	1	empty	empty	empty	no check	no total printed
debt	2025	FY2025	39	empty	empty	TOTAL MASS WATER POL TRUST-SEWER	empty	subtotal	empty	empty	empty	empty	TRUST & STABILIZATION FUNDS HELD BY OTHER B…	73779.00	73792.00	73806.00	73820.00	73835.00	73881.00	73896.00	73913.00	9	empty	empty	empty	checked	v3: 2,981,199.55 vs 2,981,199.55 [v-numbers…

report_dept_activity town extract 515 rows 2011–2025one row per printed line

Departmental activity tables — permits issued, inspections done, cases handled. Mostly COUNTS rather than dollars, and what is counted differs by department and by year.

v1..v8 are ORDINALS. Split on status.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
dept_activity	2016	FY2016	53	empty	Cemetery Commissioners, Ch	* Arthur Weidenhoff	empty	row	empty	empty	empty	empty	CEMETERY DEPARTMENT \| Cemetery Commissioner…	81	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	empty	no check	no total printed
dept_activity	2016	FY2016	53	empty	Cemetery Commissioners, Ch	Murray S. Smith Jr.	empty	row	empty	empty	empty	empty	CEMETERY DEPARTMENT \| Cemetery Commissioner…	81	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	empty	no check	no total printed
dept_activity	2017	FY2017	88	empty	CEMETERY NAME	empty	empty	row	empty	empty	empty	empty	CEMETERY DEPARTMENT - 2017 \| NAME AGE CEMET…	50	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	empty	no check	no total printed

report_elections town extract 2,012 rows 2011–2025one row per printed line

Election results as the annual reports print them: candidates, offices and vote counts in ordinal columns.

v1..v8 are ORDINALS. Split on status.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
elections	2014	FY2014	159	1	empty	TOTAL	empty	subtotal	parts sum to the printed total	empty	empty	empty	trailing figures	1187	955	1053	1052	4247	empty	empty	empty	5	empty	empty	empty	check failed	181 full rows tie, 1 do not; 28 partial not…
elections	2017	FY2017	186	1	empty	BLANKS	empty	row	parts sum to the printed total	empty	empty	empty	trailing figures	171	96	124	173	564	empty	empty	empty	5	empty	empty	empty	check failed	42 full rows tie, 5 do not; 12 partial not …
elections	2024	FY2024	173	1	empty	BLANKS	empty	row	parts sum to the printed total	empty	empty	empty	trailing figures	99	77	84	114	374	empty	empty	empty	5	empty	empty	empty	check failed	96 full rows tie, 14 do not; 124 partial no…

report_gross_wages town extract 3,545 rows 2011 … 2025 (13 years)one row per printed line

GROSS WAGES paid per named employee per year, as the annual reports print them. 3,545 rows. Actual pay rather than budget, including overtime and details.

v1..v8 are ORDINALS. Split on status.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
gross_wages	2016	FY2016	202	empty	FIRE	VASADY-KOVACS	empty	row	empty	empty	empty	empty	FIRE SCHOOL \| NEIL SCHOOL \| SCHOOL \| SCHOOL	5683.64	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	empty	no check	no total printed
gross_wages	2021	FY2021	166	empty	empty	RIOS JENNIFER	empty	row	empty	empty	empty	empty	OUDAIN DUVV CIV NUTAND	67253.64	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	empty	no check	no total printed
gross_wages	2025	FY2025	182	empty	empty	WENTWORTH	empty	row	empty	empty	empty	empty	2025 GROSS WAGES	51673.54	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	empty	no check	no total printed

report_officials town extract 237 rows 2011 … 2024 (13 years)one row per printed line

Names and offices of town officials as each annual report lists them. No money.

v1..v8 are ORDINALS. Split on status.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
officials	2011	FY2011	13	empty	empty	*Shirley A. Levesque	empty	row	empty	empty	empty	empty	Marion Benson, c/o Planning Board \| D.P.W. …	80	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	empty	no check	no total printed
officials	2023	FY2023	166	empty	Health & Sanitation	Highway Labor	empty	row	empty	empty	empty	empty	APPOINTED OFFICIALS \| INSPECTOR OF PLUMBING…	59	empty	empty	empty	empty	empty	empty	empty	1	empty	587,160.00	empty	no check	no total printed
officials	2023	FY2023	167	empty	Highway	Vehicle Mtc - Highway	empty	row	empty	empty	empty	empty	APPOINTED OFFICIALS \| INSPECTOR OF PLUMBING…	64	empty	empty	empty	empty	empty	empty	empty	1	empty	62,750.00	empty	no check	no total printed

report_trust_funds town extract 642 rows 2011 … 2025 (13 years)one row per printed line

Trust fund balances and activity as the annual reports print them.

v1..v8 are ORDINALS. Split on status.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
trust_funds	2011	FY2011	73	empty	LIBRARY FU	EMMA DIMOND LIBRARY BOOKS	empty	subtotal	subtotal: equals the 2 rows below it, in 2 …	empty	empty	empty	TRUST FUNDS \| BEGINNING BEGINNING DISBURSE …	300.00	125.46	0.00	3.09	0.00	empty	empty	empty	5	empty	empty	empty	check failed	v1: 354,751.46 vs 1,905,698.11 (-1,550,946.…
trust_funds	2019	FY2019	46	empty	LIBRARY FUNDS	PHILIP CROUSE	empty	row	empty	empty	empty	empty	TOWN OF LUNENBURG \| TRUST FUNDS STABILIZATI…	66.69	4.48	empty	30.00	0.00	72.00	1.63	empty	6	empty	empty	empty	check failed	v1: 2,856,386.21 vs 2,528,358.08 (+328,028.…
trust_funds	2020	FY2020	41	empty	LIBRARY FUNDS	empty	empty	footnote	printed below the grand total, which ends t…	empty	empty	empty	TOWN OF LUNENBURG Page 1 of 2 \| TRUST FUNDS…	8129.00	empty	empty	28434.40	empty	empty	empty	empty	2	empty	empty	empty	check failed	v2: 29,909.86 vs 7.00 (+29,902.86) [v-numbe…

report_valuation town extract 316 rows 2011–2019one row per printed line

Assessed property VALUATION by class, as the annual reports print it. Dollars — and the base the tax levy is computed on.

v1..v8 are ORDINALS. Split on status.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
valuation	2016	FY2016	37	empty	empty	Industrial	empty	row	empty	empty	empty	empty	BOARD OF ASSESSORS \| Fiscal 2016 Assessment…	empty	empty	empty	19994000.00	19.61	392082.34	empty	empty	3	empty	1.6667%	empty	no check	no total printed
valuation	2019	FY2019	43	empty	empty	empty	empty	row	empty	empty	empty	empty	Valuation and Tax History \| Fiscal Year Tax…	1442248677.00	4.23	empty	empty	empty	empty	empty	empty	2	empty	empty	empty	no check	no total printed
valuation	2019	FY2019	43	empty	Assessor's Account for	Charges to 6/30/2019	empty	row	empty	empty	empty	empty	Valuation and Tax History \| Fiscal Year Tax…	118953.20	133830.01	empty	empty	empty	empty	empty	empty	2	empty	empty	empty	no check	no total printed

report_vital_records town extract 96 rows 2011 … 2025 (14 years)one row per printed line

Counts of births, marriages and deaths as the annual reports print them.

v1..v8 are ORDINALS. Split on status.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`panel`	TEXT	Which block of the page this row came from, where a page prints several tables. Two panels on one page have different column rulers. Summing across panels mixes them.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`label_missing`	TEXT	The row's label did not survive extraction. The figures may still be good. A missing label is not a missing row.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`row_check`	TEXT	Whether the row satisfies an identity the table itself states.
`derived_cell`	TEXT	A cell we computed from the row's own other cells, recorded here and NEVER written into a v column. An inference kept out of the readings, on purpose.
`table_family`	TEXT	Which printed table shape this row came from.
`line_no`	TEXT	Position of the row on the page, counting from the top. Positional, so it identifies a row within a page and nothing beyond it.
`columns_as_printed`	TEXT	The column headings of this table, verbatim, as that page prints them. This is what says which printed column v1 actually is. It is the only thing that does.
`v1`	TEXT	ORDINAL, NOT A COLUMN NAME: the first column of THIS PAGE that held figures. dollars The ruler is built per page, so v1 on one page and v1 on the next can be different printed columns. Read column_meaning. Summing v1 down a run adds one page's APPROPRIATED to another's TOTAL EXPENDED.
`v2`	TEXT	ORDINAL: the second column of this page that held figures. dollars See v1.
`v3`	TEXT	ORDINAL: the third column of this page that held figures. dollars See v1.
`v4`	TEXT	ORDINAL: the fourth column of this page that held figures. dollars See v1.
`v5`	TEXT	ORDINAL: the fifth column of this page that held figures. dollars See v1.
`v6`	TEXT	ORDINAL: the sixth column of this page that held figures. dollars See v1.
`v7`	TEXT	ORDINAL: the seventh column of this page that held figures. dollars See v1.
`v8`	TEXT	ORDINAL: the eighth column of this page that held figures. dollars See v1.
`n_values`	TEXT	How many money cells this row actually carried. count Fewer than the table's column count means blank cells, and which ones were blank is positional.
`ruler_spanned`	TEXT	The column ruler could not cut this row cleanly. Flagged rather than dropped: a dropped row shifts nothing and is invisible.
`repaired_cells`	TEXT	Cells we corrected, and what the correction was. A repair is a change we made to a reading. It is recorded so it can be disputed.
`unparsed_cells`	TEXT	Cells we could not read. Kept rather than dropped: a dropped cell is invisible.
`column_meaning`	TEXT	What each ordinal column is, where the table states an identity that fixes it. Says `not established` where the table does not fix it, and that is not the same as unknown-and-guessable.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`reconciliation`	TEXT	The per-column result of that check: what we summed against what the report printed.

First 3 rows, in rowid order:

dataset	fy	edition	page	panel	group	label	label_missing	kind	row_check	derived_cell	table_family	line_no	columns_as_printed	v1	v2	v3	v4	v5	v6	v7	v8	n_values	ruler_spanned	unparsed_cells	column_meaning	status	reconciliation
vital_records	2011	FY2011	13	empty	empty	*Betty Ann Sikora	empty	row	empty	empty	empty	empty	JANUARY JULY \| Nancy L Woodruff 79 N *Richa…	72	empty	empty	empty	empty	empty	empty	empty	1	empty	empty	empty	no check	no total printed
vital_records	2012	FY2012	105	1	empty	TOTAL	empty	subtotal	empty	empty	empty	empty	trailing figures	2760	2154	2520	2688	10122	empty	empty	empty	5	empty	empty	empty	no check	no total printed
vital_records	2013	FY2013	92	1	empty	TOTAL	empty	subtotal	empty	empty	empty	empty	trailing figures	564	436	482	535	2017	empty	empty	empty	5	empty	empty	empty	no check	no total printed

Both at once — the ledger and the model24 tables · 7,000 rows

Tables that carry every department together. The town's books do not separate school from town, and neither do these: splitting them would invent a division the source does not make.

Tier 1 · totalswhat was appropriated and what is left, by department

balance_sheet school + town Tier 1 · totals fact 774 rows 2011–2022one row per line per fund column per year

What the town HOLDS, read from the COMBINED BALANCE SHEET — assets, liabilities and fund equity across six fund columns (general, special revenue, enterprise, capital project, trust and agency, long-term debt). Everything else in this archive measures money MOVING; this is the only table that measures what is sitting there.

Read from the page, like special_revenue_read. It is trusted because three checks hold: each column foots to its own printed total, the sheet's identity (liabilities + fund equity = assets) holds per column, and special revenue + enterprise fund equity ties to the special revenue schedule's carried-forward total — an INDEPENDENT document. One printed sheet does not balance itself (FY2019 trust and agency, $657,297.35 apart); that is the town's print, recorded not corrected.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page_printed`	TEXT	not described yet
`page_pdf`	TEXT	not described yet
`section`	TEXT	The budget book's own section heading.
`line`	TEXT	not described yet
`fund`	TEXT	The MUNIS fund number. School money spans 13xx, 15xx, 22xx and 26xx-29xx. There is no single range that means school.
`amount`	TEXT	The figure, as printed. dollars
`document`	TEXT	The source file the figure was read from.
`read_by`	TEXT	Which instrument produced the figures: `vision` for a page read directly. A reading, whichever instrument made it. It is checked by the printed total, not trusted for its source.

First 3 rows, in rowid order:

fy	edition	page_printed	page_pdf	section	line	fund	amount	document	read_by
2011	FY2011	53	56	assets	FY 11 REAL ESTATE TAXES	general	445664.60	sources/town-annual-reports/docs/4117-fy-20…	vision
2012	FY2012	53	56	assets	FY 12 REAL ESTATE TAXES	general	356747.49	sources/town-annual-reports/docs/4118-fy-20…	vision
2022	FY2022	24	24	assets	FY19 REAL ESTATE TAXES	general	-104.00	sources/town-annual-reports/docs/4129-fy-20…	vision

enterprise_balance_sheet school + town Tier 1 · totals fact 114 rows 2024–2025one row per line per fund column per printing per year

What the four ENTERPRISE funds hold at 30 June — Sewer, Water, Solid Waste and PEG Access, side by side, with a memorandum-only total column. Read from the COMBINING BALANCE SHEET — ENTERPRISE FUNDS, which is the only balance sheet FY2024 and FY2025 print; neither report contains the town-wide combined balance sheet balance_sheet holds for FY2011–FY2023.

RATEPAYER MONEY. Sewer and water user charges, betterments, trash fees and cable franchise money — none of it is available to the town's general operations and none of it is a tax dollar. This is NOT an extension of balance_sheet: that table's enterprise value is one of six fund-type COLUMNS on a different sheet, and the two must never be unioned. FY2024 prints the same sheet TWICE (printed pages 21 and 28), so printing is 1 or 2 and must be collapsed on before anything is summed — summing across it doubles the money. FY2024's memorandum column is $102,000.00 out on two offsetting lines; that is the town's print, recorded not corrected. Only cells that print a figure are recorded; the sheet's 0.00 and '-' template cells are not rows here.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page_printed`	TEXT	not described yet
`page_pdf`	TEXT	not described yet
`printing`	TEXT	not described yet
`section`	TEXT	The budget book's own section heading.
`line`	TEXT	not described yet
`fund`	TEXT	The MUNIS fund number. School money spans 13xx, 15xx, 22xx and 26xx-29xx. There is no single range that means school.
`amount`	TEXT	The figure, as printed. dollars
`document`	TEXT	The source file the figure was read from.
`read_by`	TEXT	Which instrument produced the figures: `vision` for a page read directly. A reading, whichever instrument made it. It is checked by the printed total, not trusted for its source.

First 3 rows, in rowid order:

fy	edition	page_printed	page_pdf	printing	section	line	fund	amount	document	read_by
2024	FY2024	21	21	1	liabilities	Deferred revenue: Special assessments	total_memorandum	3139121.60	sources/town-annual-reports/docs/4132-fy-20…	vision
2024	FY2024	28	28	2	liabilities	Warrants payable	peg_access	283.50	sources/town-annual-reports/docs/4132-fy-20…	vision
2025	FY2025	21	25	1	assets	Due to/Due From	solid_waste	26973.93	sources/town-annual-reports/docs/4130-fy-20…	vision

fund_activity school + town Tier 1 · totals fact 61 rows 2026–2026one row per fund per year per period

Per fund per period: opening balance, revenue in, salaries, other expenditure, encumbered and closing balance. FY2026 period 9 only — and opening_balance is NULL for every one of the 61 funds.

spent and holds.

column	type	what it is
`fund`	TEXT	The MUNIS fund number. School money spans 13xx, 15xx, 22xx and 26xx-29xx. There is no single range that means school.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`period`	INTEGER	MUNIS accounting period: a point in time INSIDE the fiscal year. A period is not a stage. Period 9 and period 12 are two different reports of the same year; adding them adds one report to the other.
`opening_balance`	REAL	What the fund held at the start of the year. dollars
`revenue`	REAL	Money received into the fund. dollars Not the same quantity as spending, and the two must never be added.
`salaries`	REAL	The salary portion of the fund's spending. dollars
`expenditure`	REAL	The non-salary portion of the fund's spending. dollars
`encumbered`	REAL	Committed by purchase order but not yet spent. dollars
`closing_balance`	REAL	What the fund held at the end of the period. dollars
`doc_id`	TEXT	The archive path of the document this row's figures came from. Every fact row carries one. A figure that cannot name its document is not loaded.

First 3 rows, in rowid order:

fund	fy	period	opening_balance	revenue	expenditure	closing_balance	doc_id
2765	2026	9	NULL	0.0	0.0	-502.0	sources/town-ledgers/fund-balances/special-…
2769	2026	9	NULL	0.0	0.0	4.55	sources/town-ledgers/fund-balances/special-…
1311	2026	9	NULL	22485.91	2910.68	109397.82	sources/town-ledgers/fund-balances/special-…

special_revenue_read school + town Tier 1 · totals fact 1,882 rows 2011–2023one row per fund per year

The special revenue schedule READ from the rendered page rather than OCR'd: balance brought forward, total receipts, total disbursements, balance carried forward, per fund. The only version of this table that ties to its own printed grand total — FY2022, 167 funds, all four columns to the penny and every row balancing.

A figure read off an image is a READING, not a published number — same status as the OCR's. It is trustworthy because the report's own printed total and the table's own identity both check it, not because a model produced it. Years that do not tie are not in here.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`fund`	TEXT	The MUNIS fund number. School money spans 13xx, 15xx, 22xx and 26xx-29xx. There is no single range that means school.
`forward`	TEXT	Balance the fund brought forward from the previous year, as the schedule prints it. dollars
`receipts`	TEXT	Total money the fund took in during the year. dollars Not the same quantity as disbursements and must never be added to it.
`disbursements`	TEXT	Total money the fund paid out during the year. dollars
`carried`	TEXT	Balance the fund carried forward into the next year. dollars Ties to the NEXT year's `forward`, which is a check across years the schedule makes possible.
`row_ties`	TEXT	Whether this row satisfies forward + receipts − disbursements = carried. The identity the table itself states. A row that does not tie is a transcription error or a printing error, and which one is not decided here.
`document`	TEXT	The source file the figure was read from.
`read_by`	TEXT	Which instrument produced the figures: `vision` for a page read directly. A reading, whichever instrument made it. It is checked by the printed total, not trusted for its source.

First 3 rows, in rowid order:

fy	edition	page	group	fund	forward	receipts	disbursements	carried	row_ties	document	read_by
2020	FY2020	32	PUBLIC SAFETY	Community Policing FY 96	6.55	empty	empty	6.55	yes	sources/town-annual-reports/docs/4127-fy-20…	vision
2017	FY2017	32	LIBRARY	FY07 Public Libraries Fund	571.17	empty	empty	571.17	yes	sources/town-annual-reports/docs/4124-fy-20…	vision
2013	FY2013	68	SCHOOL DEPARTMENT	Full Day Kindergarten Grant	1470.00	48382.00	49852.00	0.00	yes	sources/town-annual-reports/docs/4119-fy-20…	vision

Tier 2 · categorieswhether a category — guidance, special education, transport — is over

peg_access school + town Tier 2 · categories fact 110 rows 2015–2025one row per printed expense line per year

What the town's PEG Access / Public Access Cable operation spent, line by line, FY2015-FY2025, read from the expense table each annual town report prints. From FY2020 the report also prints a % of Expense column and it is recorded here beside the amount.

CABLE FRANCHISE MONEY, not tax money. Comcast pays it under the Town's cable licence; from FY2020 it sits in a PEG Access enterprise fund established under MGL c.44 s53F1/2 and is appropriated by its own Town Meeting article. None of it is available to general operations. The line LIST changes: Insurance disappears after FY2020 and Indirect Costs appears, so a line-by-line series across that boundary is comparing different things. FY2017's ten lines add to 450.01 MORE than that report's own printed TOTAL; the total is what the rest of the record uses. Join to peg_access_printed_totals for the total each year actually prints -- never sum these rows and call it the year's expenses without saying so.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page_printed`	TEXT	not described yet
`page_pdf`	TEXT	not described yet
`ordinal`	TEXT	not described yet
`line`	TEXT	not described yet
`amount`	TEXT	The figure, as printed. dollars
`pct`	TEXT	not described yet
`quote`	TEXT	What the source says, verbatim.
`document`	TEXT	The source file the figure was read from.
`read_by`	TEXT	Which instrument produced the figures: `vision` for a page read directly. A reading, whichever instrument made it. It is checked by the printed total, not trusted for its source.

First 3 rows, in rowid order:

fy	edition	page_printed	page_pdf	ordinal	line	amount	pct	quote	document	read_by
2022	FY2022	72	72	5	Office Supplies	0.00	0.00	$0.00	sources/town-annual-reports/docs/4129-fy-20…	vision
2022	FY2022	72	72	9	Videographers	25516.28	20.06	$25,516.28	sources/town-annual-reports/docs/4129-fy-20…	vision
2023	FY2023	80	80	8	New Equip	15205.50	9.32	$15,205.50	sources/town-annual-reports/docs/4131-fy-20…	vision

variance_by_group school + town Tier 2 · categories derived 59 rowsone row per function group

Computed, not read: per function group, how far spending ran from budget across years — budgeted, spent, net, gross, churn, and the best and worst years.

across years.

column	type	what it is
`function_group`	TEXT	Our grouping of DESE function codes into readable categories.
`years`	TEXT	not described yet
`budgeted`	TEXT	not described yet
`spent`	TEXT	Money paid out of the fund. dollars Not the same quantity as revenue.
`net`	TEXT	not described yet
`gross`	TEXT	not described yet
`churn`	TEXT	not described yet
`worst_year`	TEXT	not described yet
`best_year`	TEXT	not described yet

First 3 rows, in rowid order:

function_group	years	budgeted	spent	net	gross	churn	worst_year	best_year
2320 - Therapeutic Services	5	3093610	3014841	-78769	78769	1.00	-0.0532	-0.0100
(unmapped)	4	2722759	2748038	25279	62963	2.49	-0.0371	0.0590
2325 - Special Education Substitutes	5	322364	274236	-48128	153472	3.19	-0.6591	0.9754

Tier 3 · accountswhat one account spent

ledger_snapshot school + town the spine Tier 3 · accounts fact 983 rows 2026–2026one row per account per year per period per document

BOTH revenue and expense accounts from the town's MUNIS books, told apart by account_type — revenue is stored NEGATIVE. Six money measures per row: original appropriation, transfers, revised, expended, encumbered, available, plus percent used. Two grains in the same table (level): 67 department roll-ups and 916 detail accounts. FY2026 only, at two periods — 9 carries the department rows, 12 the account rows.

at a moment in the year.

column	type	what it is
`account_id`	TEXT	The full MUNIS account, fund-org-object. The stable identity. Never identify an account by its name.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`period`	INTEGER	MUNIS accounting period: a point in time INSIDE the fiscal year. A period is not a stage. Period 9 and period 12 are two different reports of the same year; adding them adds one report to the other.
`original`	REAL	The appropriation as originally voted. dollars
`transfers`	REAL	Money moved into or out of the appropriation after it was voted. dollars This is why `original` and `revised` differ.
`revised`	REAL	The appropriation after transfers. dollars
`expended`	REAL	Spent. dollars
`encumbered`	REAL	Committed by purchase order but not yet spent. dollars
`available`	REAL	Revised less expended less encumbered. dollars
`pct_used`	REAL	Share of the revised budget used. percent
`rounded_columns`	TEXT	Which columns the source printed rounded. Sets the tolerance a reconciliation may use — one dollar per rounded row, not a blanket.
`doc_id`	TEXT	The archive path of the document this row's figures came from. Every fact row carries one. A figure that cannot name its document is not loaded.

First 3 rows, in rowid order:

account_id	fy	period	original	transfers	revised	expended	available	rounded_columns	doc_id
0100-11462-573100	2026	12	750.0	0.0	750.0	250.0	500.0	empty	sources/town-ledgers/expenses/glytdbud-expe…
0100-11751-511000	2026	12	0.0	0.0	0.0	12796.8	-12796.8	empty	sources/town-ledgers/expenses/glytdbud-expe…
0100-14112-521016	2026	12	750.0	-750.0	0.0	0.0	0.0	empty	sources/town-ledgers/expenses/glytdbud-expe…

account school + town Tier 3 · accounts dimension 981 rowsone row per MUNIS account

The chart of accounts — no money at all. One row per MUNIS account number split into fund, department, org and object, with its DESE function code, its name, whether it is revenue or expense, whether it is a roll-up or a detail account, and the years it appears.

department and function code.

column	type	what it is
`account_id`	TEXT	The full MUNIS account, fund-org-object. The stable identity. Never identify an account by its name.
`fund`	TEXT	The MUNIS fund number. School money spans 13xx, 15xx, 22xx and 26xx-29xx. There is no single range that means school.
`fund_name`	TEXT	The fund's name as MUNIS prints it. CHAPTER 658 REVOLVING FUND is athletics. Identify a fund by number.
`dept`	TEXT	The MUNIS department number. A department is not a purpose: school retiree health sits in the insurance department.
`org`	TEXT	The MUNIS org segment of the account.
`object`	TEXT	The MUNIS object code — what kind of thing the money buys.
`account_string`	TEXT	not described yet
`function`	TEXT	The DESE function code. Joins a CATEGORY, never a line.
`name`	TEXT	The name as the source prints it. MUNIS truncates account names to ten characters.
`account_type`	TEXT	Whether the account is expense or revenue. Revenue is stored NEGATIVE. Omitting this once made the town's whole budget compute as minus $997,871.
`level`	TEXT	School level, where the row is about athletics rather than accounts. Means something different in `account` — see that table's entry.
`first_seen_fy`	INTEGER	not described yet
`last_seen_fy`	INTEGER	not described yet

First 3 rows, in rowid order:

account_id	fund	fund_name	dept	org	object	account_string	function	name	account_type	level	first_seen_fy	last_seen_fy
0100-11462-534400	0100	GENERAL FUND	146	11462	534400	0100-1-146-0000-00-0-00-2-534400	NULL	POSTAGE	expense	account	2026	2026
0100-19142-570002	0100	GENERAL FUND	914	19142	570002	0100-9-914-0000-00-0-00-2-570002	NULL	LIFE INS	expense	account	2026	2026
0100-01001-432800	0100	GENERAL FUND	NULL	01001	432800	NULL	NULL	ZN BD FILE	revenue	account	2026	2026

account_names school + town Tier 3 · accounts dimension 124 rowsone row per abbreviated account name

A decoder, not data: what each ten-character MUNIS abbreviation expands to, and the basis for each expansion.

The expansions are ours except where basis says otherwise.

column	type	what it is
`dept`	TEXT	The MUNIS department number. A department is not a purpose: school retiree health sits in the insurance department.
`object`	TEXT	The MUNIS object code — what kind of thing the money buys.
`org`	TEXT	The MUNIS org segment of the account.
`code`	TEXT	not described yet
`readable`	TEXT	not described yet
`basis`	TEXT	Whether a figure is published, contractual, statutory, our measurement or our assumption.
`note`	TEXT	What we recorded about this row.

First 3 rows, in rowid order:

dept	object	org	code	readable	basis	note
300	511002	empty	HS GUID SE	High school guidance secretary	department context	Object 511002 is the secretarial salary obj…
300	512103	empty	KIND SUBS	Kindergarten substitutes	district document	The district budget prints “Kind Regular Su…
300	519104	empty	UNIFSPORTS	Sports uniforms	district document	The district budget prints “Replacement of …

crosswalk school + town Tier 3 · accounts dimension 0 rowsone row per budget line mapped to an account

EMPTY, deliberately. It would map a budget line to a ledger account, and it cannot: MUNIS truncates names to ten characters so MS GUIDANC and HS GUIDANC collide. Only the CATEGORY joins, through account.function.

Deliberately empty. MUNIS truncates to ten characters so MS GUIDANC and HS GUIDANC collide; only the CATEGORY can be joined, through account.function. Populating this would record an inference as a mapping.

column	type	what it is
`line_key`	TEXT	Our normalised join key for a budget line. Normalisation differs between budget_figure and workbook_figure — joining them on line_key matches only a fraction. Join on the normalised label instead.
`account_id`	TEXT	The full MUNIS account, fund-org-object. The stable identity. Never identify an account by its name.
`method`	TEXT	not described yet
`confidence`	TEXT	not described yet
`evidence`	TEXT	What establishes the claim in this row.
`doc_id`	TEXT	The archive path of the document this row's figures came from. Every fact row carries one. A figure that cannot name its document is not loaded.

No rows. That is the honest state of this table, not a load that failed — see its entry in SCHEMA.md.

munis_ledger school + town Tier 3 · accounts fact 983 rows 2026–2026one row per account per report

The MUNIS chart of accounts exactly as DELIVERED, before our normalisation — including the report parameters (totals_only, suppress_zero) that decide what a report shows and therefore what is missing from it.

column	type	what it is
`doc_id`	TEXT	The archive path of the document this row's figures came from. Every fact row carries one. A figure that cannot name its document is not loaded.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`period`	TEXT	MUNIS accounting period: a point in time INSIDE the fiscal year. A period is not a stage. Period 9 and period 12 are two different reports of the same year; adding them adds one report to the other.
`fund`	TEXT	The MUNIS fund number. School money spans 13xx, 15xx, 22xx and 26xx-29xx. There is no single range that means school.
`fund_name`	TEXT	The fund's name as MUNIS prints it. CHAPTER 658 REVOLVING FUND is athletics. Identify a fund by number.
`funds_covered`	TEXT	not described yet
`account_type`	TEXT	Whether the account is expense or revenue. Revenue is stored NEGATIVE. Omitting this once made the town's whole budget compute as minus $997,871.
`level`	TEXT	School level, where the row is about athletics rather than accounts. Means something different in `account` — see that table's entry.
`dept`	TEXT	The MUNIS department number. A department is not a purpose: school retiree health sits in the insurance department.
`org`	TEXT	The MUNIS org segment of the account.
`object`	TEXT	The MUNIS object code — what kind of thing the money buys.
`account`	TEXT	not described yet
`function`	TEXT	The DESE function code. Joins a CATEGORY, never a line.
`name`	TEXT	The name as the source prints it. MUNIS truncates account names to ten characters.
`original`	TEXT	The appropriation as originally voted. dollars
`transfers`	TEXT	Money moved into or out of the appropriation after it was voted. dollars This is why `original` and `revised` differ.
`revised`	TEXT	The appropriation after transfers. dollars
`expended`	TEXT	Spent. dollars
`encumbered`	TEXT	Committed by purchase order but not yet spent. dollars
`available`	TEXT	Revised less expended less encumbered. dollars
`pct_used`	TEXT	Share of the revised budget used. percent
`totals_only`	TEXT	not described yet
`suppress_zero`	TEXT	not described yet
`rounded_columns`	TEXT	Which columns the source printed rounded. Sets the tolerance a reconciliation may use — one dollar per rounded row, not a blanket.

First 3 rows, in rowid order:

doc_id	fy	period	fund	fund_name	funds_covered	account_type	level	dept	org	object	account	function	name	original	transfers	revised	expended	encumbered	available	suppress_zero	rounded_columns
sources/town-ledgers/expenses/glytdbud-expe…	2026	12	0100	GENERAL FUND	empty	expense	account	300	S1011022	535003	0100-3-300-1210-01-1-02-2-535003	1210	DUES/MTGS	8000.0	-3584.8	4415.2	3521.0	0.0	894.2	1	empty
sources/town-ledgers/expenses/glytdbud-expe…	2026	12	0100	GENERAL FUND	empty	expense	account	411	14112	540000	0100-4-411-0000-00-0-00-2-540000	0000	OFF SUPPLI	1100.0	0.0	1100.0	392.19	523.31	184.5	1	empty
sources/town-ledgers/revenue/glytdbud-reven…	2026	9	0100	GENERAL FUND	0100 GENERAL FUND	revenue	account	empty	01001	417900	empty	empty	INT FARM A	0.0	0.0	0.0	0.0	0.0	0.0	0	original,transfers,revised

Not about moneyCounts, names, dates and documents — everything this archive holds that is not a dollar figure

balance_sheet_printed_totals school + town provenance 288 rows 2011–2022one row per printed total per fund column per year

The totals each balance sheet PRINTS — total assets, total liabilities, total fund equity — quoted with the page they appear on.

The independent check on balance_sheet, so it must be read off the page and never computed from our own rows.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page_printed`	TEXT	not described yet
`page_pdf`	TEXT	not described yet
`total_row`	TEXT	not described yet
`fund`	TEXT	The MUNIS fund number. School money spans 13xx, 15xx, 22xx and 26xx-29xx. There is no single range that means school.
`amount`	TEXT	The figure, as printed. dollars
`quote`	TEXT	What the source says, verbatim.

First 3 rows, in rowid order:

fy	edition	page_printed	page_pdf	total_row	fund	amount	quote
2012	FY2012	53	56	TOTAL ASSETS	general	6185491.72	TOTAL ASSETS row, printed page 53
2018	FY2018	26	26	TOTAL LIABILITIES/FUND EQUITY	capital_project	1671644.20	TOTAL LIABILITIES/FUND EQUITY row, printed …
2020	FY2020	22	22	TOTAL LIABILITIES/FUND EQUITY	special_revenue	2650524.60	TOTAL LIABILITIES/FUND EQUITY row, printed …

enterprise_balance_sheet_printed_totals school + town provenance 75 rows 2024–2025one row per printed total per fund column per printing per year

The totals the enterprise balance sheet PRINTS — Total Assets, Total Liabilities, Total Fund Equity, Total Liabilities and Fund Equity, and the sheet's own PROOF row — quoted with the raw cell text and the page.

The independent check on enterprise_balance_sheet, so it must be read off the page and never computed from our own rows. The PROOF row is the sheet's own self-check and is what a reading is reconciled to. FY2025 prints its Proof as '(0.00)' and '-' in different columns; the amount recorded is 0.00 and the raw glyph is in quote.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page_printed`	TEXT	not described yet
`page_pdf`	TEXT	not described yet
`printing`	TEXT	not described yet
`total_row`	TEXT	not described yet
`fund`	TEXT	The MUNIS fund number. School money spans 13xx, 15xx, 22xx and 26xx-29xx. There is no single range that means school.
`amount`	TEXT	The figure, as printed. dollars
`quote`	TEXT	What the source says, verbatim.

First 3 rows, in rowid order:

fy	edition	page_printed	page_pdf	printing	total_row	fund	amount	quote
2024	FY2024	21	21	1	TOTAL LIABILITIES AND FUND EQUITY	sewer	5573383.92	printed 'Total Liabilities and Fund Equity'…
2024	FY2024	28	28	2	TOTAL LIABILITIES AND FUND EQUITY	water	65079.88	printed 'Total Liabilities and Fund Equity'…
2025	FY2025	21	25	1	TOTAL LIABILITIES	sewer	3172048.99	printed 'Total Liablities' row, FY2025 prin…

fiscal_period school + town dimension 13 rowsone row per MUNIS period

What each MUNIS accounting period means and how many months of the year had elapsed. Period 13 is the year-end close after the lapse period, not a thirteenth month.

Period 13 is the year-end close after the lapse period, not a thirteenth month.

column	type	what it is
`period`	INTEGER	MUNIS accounting period: a point in time INSIDE the fiscal year. A period is not a stage. Period 9 and period 12 are two different reports of the same year; adding them adds one report to the other.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`months_elapsed`	REAL	not described yet
`is_final`	INTEGER	not described yet

First 3 rows, in rowid order:

period	label	months_elapsed
10	April	10.0
11	May	11.0
12	June	12.0

fund school + town dimension 68 rowsone row per MUNIS fund

One row per MUNIS fund: its number, its name, what kind of fund it is and what restricts it. No money.

with its kind and restriction.

column	type	what it is
`fund`	TEXT	The MUNIS fund number. School money spans 13xx, 15xx, 22xx and 26xx-29xx. There is no single range that means school.
`name`	TEXT	The name as the source prints it. MUNIS truncates account names to ten characters.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`restriction`	TEXT	not described yet

First 3 rows, in rowid order:

fund	name	kind	restriction
1311	SCHOOL GIFT FUND	gift	NULL
2709	FY12 REG DISSEMINATION GR#321	NULL	NULL
2653	FY24 FAMILY & COMMUNITY #237	NULL	NULL

money_assumptions school + town our model 6 rowsone row per assumption

Every assumption the money model rests on, what it rests on, and what single document would settle it.

with what would settle it.

column	type	what it is
`assumption`	TEXT	not described yet
`evidence`	TEXT	What establishes the claim in this row.
`rests_on`	TEXT	not described yet
`settled_by`	TEXT	not described yet

First 3 rows, in rowid order:

assumption	evidence	rests_on	settled_by
`SCHRESSTIP` is a school resource officer s…	Read from the abbreviation, in the police d…	A $6,800 line. Immaterial, and still an inf…	The account’s full name from MUNIS, which t…
The district workbook total ties to the dep…	They land 0.4% apart for FY2026.	Used to argue the $26m is the town’s bill r…	Whether the 0.4% is netting or simply that …
The pension assessment includes school non-…	Standard Massachusetts practice: teachers i…	The reason the pension is treated as a scho…	The WRRS annual actuarial valuation, which …

money_classification school + town our model 252 rowsone row per revenue account or fund or department

Our model, not measurement. One row per revenue account, fund or department: what we classified it as, HOW that was established, and WHY. 252 rows covering every money node in the school and town diagrams.

how and why carry the basis so a classification can be argued with.

column	type	what it is
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`key`	TEXT	The row's join key.
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`label`	TEXT	The row's name, as printed. Not a key. MUNIS truncates to ten characters, so MS GUIDANC and HS GUIDANC collide.
`classification`	TEXT	Our classification of the node.
`how`	TEXT	How that classification was established.
`why`	TEXT	Why the classification was made, in words that can be argued with.

First 3 rows, in rowid order:

kind	key	group	label	classification	how	why
revenue_account	SALE STAND	onetime	One-time money	general_fund_revenue	named	Free cash and proceeds. Spendable once
report_receipt	SENIORCENTERRENTALFEES	local	SENIOR CENTER RENTAL FEES	general_fund_revenue	residual	Local receipts — a residual: fees, permits,…
report_receipt	STUDENTPARKINGFEES	local	STUDENT PARKING FEES	general_fund_revenue	residual	Local receipts — a residual: fees, permits,…

money_edges school + town our model 11 rowsone row per traced edge

Where a restricted fund's money is PERMITTED to go, with the statute or grant award that permits it. An edge is a presumption about purpose, never an observation of spending.

An edge is the statute or grant award — a PRESUMPTION about purpose, never an observation of spending.

column	type	what it is
`source`	TEXT	Where the figure came from, as the extractor recorded it. A loose reference — a filename — not a resolved address.
`use`	TEXT	not described yet
`basis`	TEXT	Whether a figure is published, contractual, statutory, our measurement or our assumption.
`why`	TEXT	Why the classification was made, in words that can be argued with.

First 3 rows, in rowid order:

source	use	basis	why
Fund 1301 (athletics)	Athletics spending	restricted	The fund is athletics — established from it…
Grant funds 26xx–29xx	The purpose of each grant	restricted	Restricted by the grant award rather than b…
State teachers’ pension system	District teaching staff	unknown	Not appropriated by Lunenburg, not in our a…

money_gaps school + town our model 39 rowsone row per gap

The honest list of what CANNOT be answered — per side (money in, money out), what is missing and why the published documents do not close it.

This table is the honest list of what cannot be answered.

column	type	what it is
`side`	TEXT	not described yet
`what`	TEXT	not described yet
`why`	TEXT	Why the classification was made, in words that can be argued with.

First 3 rows, in rowid order:

side	what	why
document_wanted	DESE End of Year Financial Report	Published by the state. — closes: Separates…
people	Which fund pays which post	This is the load-bearing one. The rosters n…
money_out	School share of Medicare, active health ins…	Five of the seven FY2026 insurance accounts…

peg_access_identities school + town provenance 23 rows 2015–2025one row per arithmetic statement each PEG statement makes about itself

The arithmetic the printed statement states about itself, written over the row ORDINALS in peg_access_printed_totals -- 1 + 2 = 3 and so on. It is transcribed from the page, because the arithmetic changes every few years.

This is what verify_peg_access.py recomputes, and it is also a completeness check: every printed row must be reached by some identity. FY2020's row 5 is not, and that is recorded as a printed defect rather than removed.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`statement`	TEXT	not described yet
`identity`	TEXT	not described yet
`note`	TEXT	What we recorded about this row.

First 3 rows, in rowid order:

fy	edition	statement	identity	note
2017	FY2017	revenue_vs_expenses	3 - 4 = 5	empty
2022	FY2022	revenue_vs_expenses	1 + 2 = 3	empty
2024	FY2024	revenue_vs_expenses	1 + 2 + 3 = 4	empty

peg_access_printed_totals school + town provenance 77 rows 2015–2025one row per printed figure per statement per year

The REVENUE vs EXPENSES statement each report prints, row by row in printed order, plus the line-item TOTAL and what the prose says the revenue was -- each with the raw cell text, the printed page and the PDF index.

total_row is the label the page prints and it is NOT unique: FY2019's statement carries two rows labelled Subtotal. Use ordinal. The rows do not mean the same thing across years -- FY2015-FY2019 Starting Balance is the prior year's Ending Balance, FY2021 and FY2025 it is the BUDGET, and FY2022 has no balance at all, so TOTAL there is one year's surplus and not a fund balance. precision says whether the statement is printed in cents or whole dollars; FY2024 and FY2025 are in dollars while the expense table beside them is in cents.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page_printed`	TEXT	not described yet
`page_pdf`	TEXT	not described yet
`statement`	TEXT	not described yet
`ordinal`	TEXT	not described yet
`total_row`	TEXT	not described yet
`amount`	TEXT	The figure, as printed. dollars
`pct`	TEXT	not described yet
`precision`	TEXT	not described yet
`quote`	TEXT	What the source says, verbatim.
`document`	TEXT	The source file the figure was read from.

First 3 rows, in rowid order:

fy	edition	page_printed	page_pdf	statement	ordinal	total_row	amount	pct	precision	quote	document
2017	FY2017	68	68	prose	1	REVENUE STATED IN PROSE	107849.80	empty	cents	The overall payments received for FY17 were…	sources/town-annual-reports/docs/4124-fy-20…
2019	FY2019	81	83	revenue_vs_expenses	6	Comcast Tech Fund (Undesignated Fund)	124796.01	empty	cents	$124,796.01	sources/town-annual-reports/docs/4126-fy-20…
2024	FY2024	63	63	revenue_vs_expenses	2	Enterprise Fund Interest	24498.00	empty	dollars	$ 24,498	sources/town-annual-reports/docs/4132-fy-20…

rate_register school + town fact 62 rows 2024–2029one row per rate

Every rate and fee the model uses: the value, what kind of value it is, when it was set, when it expires, its source document and cell reference, and whether the model actually uses it.

with where it was set.

column	type	what it is
`fy`	TEXT	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`category`	TEXT	not described yet
`unit`	TEXT	What the figure is counted in.
`item`	TEXT	The thing being priced or counted.
`value`	TEXT	The figure, as printed. dollars
`value_type`	TEXT	not described yet
`set_on`	TEXT	The date the rate or fee was set.
`expires`	TEXT	not described yet
`source`	TEXT	Where the figure came from, as the extractor recorded it. A loose reference — a filename — not a resolved address.
`source_file`	TEXT	The file the figure was read from.
`source_ref`	TEXT	Where in the source the figure sits — a cell, a line, a page. Rule 13: a coordinate, so a figure can be checked rather than taken.
`status`	TEXT	Whether this row's table was CHECKED against a total the report itself prints. `checked`, `check failed`, `no check`. NOTHING may be aggregated without splitting on this — check failed is not slightly-worse data, it is data whose columns do not sum to the printed total.
`in_model`	TEXT	not described yet

First 3 rows, in rowid order:

fy	category	unit	item	value	value_type	set_on	expires	source	source_file	source_ref	status	in_model
2024	athletic_fee	MS	reduced_fee	26.00	dollars	empty	empty	District athletics workbook, by sport (Town…	munis-ledgers/account-details/athletics-by-…	Spring!T3	verified	reference
2025	athletic_fee	ANY	unified_track	100.00	dollars	empty	empty	LHS Athletics FAQ (rschoolteams.com)	district-budget/text/lhs-athletics-faq.txt	Unified Track =$100.00	verified	reference
2027	contract_cola	Paraprofessionals	cost-of-living adjustment to the salary sca…	2.0	percent	empty	June 30, 2028	Lunenburg school employee contracts — Parap…	contracts/CONTRACTS.md	Paraprofessionals \| FY26–FY28 \| 3.0% · 2.0%…	recorded	deliberately not used — the line has never …

special_revenue_printed_totals school + town provenance 13 rows 2011–2023one row per edition

The GRAND TOTAL each annual report PRINTS for the special revenue schedule, quoted with its page. Four columns.

This is the independent check on special_revenue_read, so it must be read off the page and never computed from our own rows — the OCR misread this very figure by $1,010 and the reconciliation was measured against it for months.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`forward`	TEXT	Balance the fund brought forward from the previous year, as the schedule prints it. dollars
`receipts`	TEXT	Total money the fund took in during the year. dollars Not the same quantity as disbursements and must never be added to it.
`disbursements`	TEXT	Total money the fund paid out during the year. dollars
`carried`	TEXT	Balance the fund carried forward into the next year. dollars Ties to the NEXT year's `forward`, which is a check across years the schedule makes possible.
`quote`	TEXT	What the source says, verbatim.

First 3 rows, in rowid order:

fy	edition	page	forward	receipts	disbursements	carried	quote
2015	FY2015	33	2599893.63	5285016.77	5180429.68	2704480.72	GRAND TOTAL row, page 33
2021	FY2021	35	5596048.89	7088091.52	6269397.74	6414742.67	GRAND TOTAL row, page 35
2023	FY2023	43	8663349.91	9102807.73	7735057.86	10031099.78	GRAND TOTAL row, page 43

stated_figure school + town fact 2 rows 2025–2025one row per figure the town stated

Figures the town or district ASSERTED in their own words, with the quote, who said it, when, and what it supersedes. Claims, kept as claims.

with the quote.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`metric`	TEXT	Which quantity this row measures.
`amount`	REAL	The figure, as printed. dollars
`stated_on`	TEXT	not described yet
`stated_by`	TEXT	not described yet
`basis`	TEXT	Whether a figure is published, contractual, statutory, our measurement or our assumption.
`doc_id`	TEXT	The archive path of the document this row's figures came from. Every fact row carries one. A figure that cannot name its document is not loaded.
`source_ref`	TEXT	Where in the source the figure sits — a cell, a line, a page. Rule 13: a coordinate, so a figure can be checked rather than taken.
`quote`	TEXT	What the source says, verbatim.
`supersedes`	REAL	not described yet
`note`	TEXT	What we recorded about this row.

First 2 rows, in rowid order:

fy	metric	amount	stated_on	stated_by	basis	doc_id	source_ref	quote	supersedes	note
2025	school_surplus	582115.44	2025-09-03	Superintendent (memo read into the record b…	close	sources/meetings/text/school-committee/2025…	line 36	recently discovered in the amount of $582,1…	NULL	The Chair read a prepared statement about a…
2025	school_surplus	603885.97	2025-09-17	Business Administrator (Mr. McNamara) to th…	close	sources/meetings/text/school-committee/2025…	line 123	the surplus number has gone up to $603,885.…	582115.44	The $21,770.53 difference is encumbrances r…

The state's figures2 tables · 5,964 rows

DESE's own published measures, for Lunenburg and every other district. All-funds, so NOT comparable with a general fund appropriation.

Not about moneyCounts, names, dates and documents — everything this archive holds that is not a dollar figure

dese_measure state fact 2,982 rows 2009–2025one row per district per year per measure

The state's published measures for EVERY Massachusetts district, per year: enrollment, per-pupil spending, selected populations. Each value carries whether it reconciles to DESE's own printed total.

column	type	what it is
`lea`	TEXT	The state's district code.
`district`	TEXT	The district's name as DESE prints it.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`measure`	TEXT	Which DESE measure this row reports.
`value`	REAL	The figure, as printed. dollars
`reconciles`	TEXT	Whether this figure ties to DESE's own printed total.
`doc_id`	TEXT	The archive path of the document this row's figures came from. Every fact row carries one. A figure that cannot name its document is not loaded.

First 3 rows, in rowid order:

lea	district	fy	group	measure	value	reconciles	doc_id
01620000	Lunenburg	2025	MCAS Performance	ELA Grade 10 % Meets or Exceeds	0.57	yes	sources/state-dese/radar-district-compariso…
06730000	Groton-Dunstable	2024	Expenditures Per Pupil	Pupil Services	2073.0	yes	sources/state-dese/radar-district-compariso…
06730000	Groton-Dunstable	2012	Expenditures Per Pupil	Teachers	4483.0	yes	sources/state-dese/radar-district-compariso…

dese_radar state fact 2,982 rows 2009–2025one row per district per year per measure

DESE's ALL-FUNDS per-pupil expenditure figures for every district. All funds means grants and revolving funds are included, so these are NOT comparable with a general fund appropriation.

All-funds, so NOT comparable with the town's general fund appropriation.

column	type	what it is
`lea`	TEXT	The state's district code.
`district`	TEXT	The district's name as DESE prints it.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`group`	TEXT	The department heading the row sits under, as printed. A heading governs every row below it until the next one.
`measure`	TEXT	Which DESE measure this row reports.
`value`	TEXT	The figure, as printed. dollars
`reconciles`	TEXT	Whether this figure ties to DESE's own printed total.
`doc_id`	TEXT	The archive path of the document this row's figures came from. Every fact row carries one. A figure that cannot name its document is not loaded.

First 3 rows, in rowid order:

lea	district	fy	group	measure	value	reconciles	doc_id
06100000	Ashburnham-Westminster	2019	MCAS Performance	Math Grades 3-8 % Meets or Exceeds	0.518	yes	sources/state-dese/radar-district-compariso…
06160000	Ayer Shirley School District	2017	MCAS Performance	Math Grades 3-8 % Meets or Exceeds	0.433	yes	sources/state-dese/radar-district-compariso…
07350000	North Middlesex	2019	Expenditures Per Pupil	Operations and Maintenance	1513	yes	sources/state-dese/radar-district-compariso…

About the archive itself9 tables · 5,951 rows

Which document a figure came from, what has been read, and what these tables mean.

Not about moneyCounts, names, dates and documents — everything this archive holds that is not a dollar figure

annual_report_catalogue archive provenance 866 rows 2011–2025one row per table per annual report

A catalogue of what EXISTS, not of what has been extracted. Every table found by reading all sixteen annual reports end to end: its printed heading verbatim, its pages, its grain, its printed total, and whether it is extractable.

A catalogue of what EXISTS, not of what has been extracted.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`document`	TEXT	The source file the figure was read from.
`name`	TEXT	The name as the source prints it. MUNIS truncates account names to ten characters.
`printed_heading`	TEXT	not described yet
`pages`	TEXT	Which pages of the source this covers.
`what_it_is`	TEXT	not described yet
`grain`	TEXT	What one row of this table represents. The field that decides whether two tables can be joined or summed.
`columns`	TEXT	not described yet
`layout`	TEXT	not described yet
`printed_total`	TEXT	The total the source itself prints, quoted. This is the independent check. An extract that does not tie to it is not trusted.
`approx_rows`	TEXT	not described yet
`extractable`	TEXT	not described yet
`notes`	TEXT	What we recorded about this row while reading it.
`cluster`	TEXT	not described yet

First 3 rows, in rowid order:

fy	edition	document	name	printed_heading	pages	what_it_is	grain	columns	layout	printed_total	approx_rows	extractable	notes	cluster
2013	FY2013	4119-fy-2013-annual-town-report.pdf	Nashoba Environmental Health Department act…	Environmental Health Department / Environme…	16	Counts of regional health district activity…	one activity type	(activity) \| (count)	label with dot leaders, count right	empty	12	clean	THE MOST COMPLETE OF THE THREE YEARS: Lunen…	activity counts department environmental
2017	FY2017	4124-fy-2017-annual-town-report.pdf	Monty Tech assessment ratio percentages by …	ASSESSMENT RATIO PERCENTAGES	147	Each member town's operating and capital ap…	one member town	COMMUNITIES \| TRANSPORTATION & OTHER OPERAT…	Seven-column table	TOTALS 100.00% 100.00% 0.00% 100.00% 100.00…	18	clean	Footnote: '* Capital Percentages are used f…	assessment monty percentages ratio
2018	FY2018	4125-fy-2018-annual-town-report.pdf	Fire department calls for service, 2018	empty	96	Calendar 2018 fire department run volume sp…	one call category	category \| count	Two side-by-side label/number pairs embedde…	stated in the prose above it: 'your Fire De…	4	clean	Fires & Hazardous Conditions 452, Emergency…	calls department fire service

annual_report_contents archive provenance 199 rows 2011–2025one row per extracted table

What was actually READ: which tables, from which pages, how many figures, and whether the extract is checkable against a printed total.

from which pages

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`table`	TEXT	Which table this row describes.
`pages`	TEXT	Which pages of the source this covers.
`page_count`	TEXT	not described yet
`pages_with_figures`	TEXT	not described yet
`figure_rows`	TEXT	not described yet
`figures`	TEXT	not described yet
`columns`	TEXT	not described yet
`read_by`	TEXT	Which instrument produced the figures: `vision` for a page read directly. A reading, whichever instrument made it. It is checked by the printed total, not trusted for its source.
`printed_total`	TEXT	The total the source itself prints, quoted. This is the independent check. An extract that does not tie to it is not trusted.
`checkable`	TEXT	not described yet
`csv_fields`	TEXT	not described yet
`grain`	TEXT	What one row of this table represents. The field that decides whether two tables can be joined or summed.

First 3 rows, in rowid order:

fy	table	pages	page_count	pages_with_figures	figure_rows	figures	columns	read_by	printed_total	checkable	csv_fields	grain
2018	payroll	5,248-254	8	7	304	586	7	layout+plain	empty	no	fy, department, position_count, gross	AGGREGATE ONLY -- never one row per named p…
2019	capital_project	36	1	1	9	189	2	ocr	empty	no	fy, project_number, project_name, balance	one row per capital project per year
2025	enterprise	25,35,53,64,67,140,142-143,152	9	8	96	155	6	layout+plain	empty	no	fy, fund, line, amount	one row per enterprise fund line per year

annual_report_survey archive provenance 2,751 rows 2011–2025one row per page surveyed

A measurement of PAGES, not of figures: per page, the extraction mode, how many gutters and money tokens it holds, how many characters and rows. This is how the right instrument is chosen per page.

A survey, not data: it measures pages, it does not read figures.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`document`	TEXT	The source file the figure was read from.
`page`	TEXT	The page number printed on the page, not the PDF's index. The two differ by a cover offset that is not constant across years.
`mode`	TEXT	not described yet
`gutters`	TEXT	not described yet
`money`	TEXT	not described yet
`chars`	TEXT	not described yet
`rows`	TEXT	How many rows the extract produced. count
`sections`	TEXT	not described yet
`totals`	TEXT	not described yet
`note`	TEXT	What we recorded about this row.

First 3 rows, in rowid order:

fy	document	page	mode	gutters	money	chars	rows	sections	totals	note
2016	4123-fy-2016-annual-town-report.pdf	174	plain	2	39	2291	39	empty	empty	empty
2022	4129-fy-2022-annual-town-report.pdf	26	plain	1	110	3450	54	receipts	GRAND TOTAL=46,695,138.62	empty
2022	4129-fy-2022-annual-town-report.pdf	154	layout	0	0	3297	0	town_meeting	empty	text, no figures

column_glossary archive provenance 139 rowsone row per DISTINCT column name across the whole database

One row per DISTINCT column name across the whole database: what it means, its unit, whether it is published or ours, and the caution attached to it.

Keyed on the NAME, not on table plus name, because v1 and status mean the same dangerous thing everywhere they appear. Where a name genuinely differs by table, the table entry says so.

column	type	what it is
`column`	TEXT	not described yet
`meaning`	TEXT	What the column holds.
`unit`	TEXT	What the figure is counted in.
`basis`	TEXT	Whether a figure is published, contractual, statutory, our measurement or our assumption.
`caution`	TEXT	What goes wrong if this is used without knowing the above. Every entry here was written after getting it wrong at least once.

First 3 rows, in rowid order:

column	meaning	unit	basis	caution
v4	ORDINAL: the fourth column of this page tha…	dollars	derived	See v1.
variant	Which whole-budget scenario the figure belo…	empty	ours	FY27 has four — Balanced, Core Budget, Leve…
path	Where our copy lives in the archive.	empty	ours	empty

dataset_document archive provenance 225 rowsone row per dataset per document

The join between an extracted dataset and the document it came from: which pages, how many rows, whether it reconciled, and the publisher's own label, size and checksum.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`rows`	TEXT	How many rows the extract produced. count
`pages`	TEXT	Which pages of the source this covers.
`page_count`	TEXT	not described yet
`reconciled`	TEXT	not described yet
`partial`	TEXT	not described yet
`document`	TEXT	The source file the figure was read from.
`publisher_label`	TEXT	not described yet
`upstream`	TEXT	The publisher's address, as deeply as it goes.
`sha256`	TEXT	Checksum of our copy. Here because a file can be replaced at the same URL without the URL changing.
`bytes`	TEXT	Size of our copy. bytes

First 3 rows, in rowid order:

dataset	edition	rows	pages	page_count	document	publisher_label	upstream	sha256	bytes
special-revenue-funds	FY2015	129	29,30,31,32,33	5	4121-fy-2015-annual-town-report.pdf	FY 2015 Annual Town Report	https://www.lunenburgma.gov/DocumentCenter/…	712afbb458b2a452a1c4b3ce618478c8f59a31d7f04…	7590404
report-officials	FY2021	21	16,90,96	3	4128-fy-2021-annual-town-report.pdf	FY 2021 Annual Town Report	https://www.lunenburgma.gov/DocumentCenter/…	7ec4740034a2b9bd3aa7450f63c91d10e6ead472450…	82620223
report-enrollment-mcas	FY2018	13	112,125,137,139,143	5	4125-fy-2018-annual-town-report.pdf	FY 2018 Annual Town Report	https://www.lunenburgma.gov/DocumentCenter/…	364a1138df754668ee59a499cf3d76b6ac006d47a4a…	8052398

document archive provenance 616 rowsone row per source document

One row per source document: its archive path, how it reached us, its publisher URL, whether that link still opens (link_state), whether our copy still matches the publisher's (copy_state), and both checksums. Also which columns a workbook HIDES from a reader.

link_state says whether the publisher's copy still opens; copy_state whether it still matches ours.

column	type	what it is
`doc_id`	TEXT	The archive path of the document this row's figures came from. Every fact row carries one. A figure that cannot name its document is not loaded.
`path`	TEXT	Where our copy lives in the archive.
`source_type`	TEXT	not described yet
`basis`	TEXT	Whether a figure is published, contractual, statutory, our measurement or our assumption.
`ledger_at`	TEXT	not described yet
`hidden_columns`	TEXT	not described yet
`url`	TEXT	The publisher's own address for the document.
`link_state`	TEXT	not described yet
`copy_state`	TEXT	not described yet
`remote_sha256`	TEXT	not described yet
`local_sha256`	TEXT	not described yet

First 3 rows, in rowid order:

doc_id	path	source_type	basis	ledger_at	hidden_columns	url	link_state	copy_state	remote_sha256	local_sha256
sources/district-budget/text/fy23-quarterly…	sources/district-budget/text/fy23-quarterly…	ledger	budget	2	empty	NULL	NULL	NULL	NULL	5d3c93143333601a78715745056dc3d794dc2c55ed4…
sources/town-annual-reports/text/4128-fy-20…	sources/town-annual-reports/text/4128-fy-20…	forward	budget	empty	empty	NULL	NULL	NULL	NULL	751b434df07d4e54eb17aa35601fb49eda03ce43cb1…
sources/town-budget/text/a106-fy24-finance-…	sources/town-budget/text/a106-fy24-finance-…	forward	budget	empty	empty	NULL	NULL	NULL	NULL	e5b4f1ab3fe6bf3e0a1424ba2e318db988221897dc4…

extraction_plan archive provenance 553 rows 2011–2025one row per planned extract

What is planned to be read out of the annual reports and what state each item is in — including items deliberately not attempted, and why.

and what state each is in.

column	type	what it is
`dataset`	TEXT	Which extract family this row belongs to.
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`pages`	TEXT	Which pages of the source this covers.
`printed_heading`	TEXT	not described yet
`table`	TEXT	Which table this row describes.
`grain`	TEXT	What one row of this table represents. The field that decides whether two tables can be joined or summed.
`printed_total`	TEXT	The total the source itself prints, quoted. This is the independent check. An extract that does not tie to it is not trusted.
`checkable`	TEXT	not described yet
`extractable`	TEXT	not described yet
`rows`	TEXT	How many rows the extract produced. count
`notes`	TEXT	What we recorded about this row while reading it.

First 3 rows, in rowid order:

dataset	fy	edition	pages	printed_heading	table	grain	printed_total	checkable	extractable	rows	notes
officials	2016	FY2016	7	ELECTED OFFICIALS	Elected officials -- state and federal	one officeholder	empty	no	messy	9	Substantively identical to FY2015 p8.
officials	2016	FY2016	10-15	APPOINTED OFFICIALS	Appointed town officials	one appointed position/member	empty	no	messy	180	Header: 'TERMS ARE FOR ONE YEAR UNLESS OTHE…
monty_tech	2024	FY2024	118	OUR SCHOOL COMMUNITY	Monty Tech member towns	one member town	empty	no	clean	18	'Barre' is spelled correctly this year (FY2…

report_anomalies archive provenance 521 rows 2011–2025one row per anomaly

Things found while reading the annual reports that need a person: a total that does not tie, a heading that moved, a column that vanished.

column	type	what it is
`fy`	INTEGER	Fiscal year, ending 30 June. FY2026 runs 1 July 2025 to 30 June 2026. year INTEGER everywhere except grant_award/grants_history (a multi-year ESSER award) and rate_register (one rate with no year).
`edition`	TEXT	Which annual town report the row was read from, as that report names itself. FY2016 has TWO editions — the report and an addendum. Summing on fy rather than edition counts that year twice.
`document`	TEXT	The source file the figure was read from.
`pages`	TEXT	Which pages of the source this covers.
`table`	TEXT	Which table this row describes.
`kind`	TEXT	What sort of row or record this is. Meaning differs by table — see the per-table entry.
`source`	TEXT	Where the figure came from, as the extractor recorded it. A loose reference — a filename — not a resolved address.
`detail`	TEXT	not described yet

First 3 rows, in rowid order:

fy	edition	document	pages	table	kind	source	detail
2016	FY2016	4123-fy-2016-annual-town-report.pdf	empty	empty	note	surprise	The fire department publishes a genuine wor…
2016	FY2016-addendum	4122-fy-2016-annual-town-report-addendum.pdf	9-12	Collection of taxes -- receivables summary …	unreadable / OCR damage	table note	The main report's pp47-49 print the header …
2020	FY2020	4127-fy-2020-annual-town-report.pdf	40-41	Trust funds fiscal year 2020 summary	unreadable / OCR damage	table note	Here the table is transposed into column ru…

table_semantics archive provenance 81 rowsone row per table in this database

This page's own source. One row per table: its role, whose money it is, how finely it resolves money, its grain, what it holds, and what goes wrong if it is used carelessly.

Self-describing, and checked both ways: a table with no row here fails the build, and a row naming a table that no longer exists fails it too.

column	type	what it is
`table_name`	TEXT	Which table this row describes. Named `table_name` rather than `table` because `table` is an SQL reserved word and this table exists to be queried.
`role`	TEXT	What kind of table it is: fact, dimension, extract, classification, provenance or derived. Facts carry the measurements. Everything else describes, extracts, classifies or checks them.
`scope`	TEXT	Whose money or whose subject the table is about: school, town, both, state or archive. `both` is not a hedge — the ledger and the chart of accounts genuinely carry every department, and splitting them would be inventing a division the source does not make.
`tier`	TEXT	How finely the money is resolved: 1 totals, 2 categories, 3 accounts. Blank where the table is not about money. From the join map. Tier 1 answers what was appropriated and what is left; tier 2 which category is over; tier 3 what one account spent. The tier below 3 — which revenue paid which expense — no report reaches.
`grain`	TEXT	What one row of this table represents. The field that decides whether two tables can be joined or summed.
`what_it_answers`	TEXT	The question one row of this table answers.
`caution`	TEXT	What goes wrong if this is used without knowing the above. Every entry here was written after getting it wrong at least once.
`default_query`	TEXT	The query the schema page's modal opens this table with. Set only where a bare SELECT is unhelpful — a fact table keyed on a code shows numbers with no words until its dimension is joined back.

First 3 rows, in rowid order:

table_name	role	scope	tier	grain	what_it_answers	caution	default_query
money_gaps	classification	both	empty	one row per gap	The honest list of what CANNOT be answered …	This table is the honest list of what canno…	empty
report_debt	extract	town	empty	one row per printed line	Outstanding DEBT and repayment schedules as…	v1..v8 are ORDINALS. Split on `status`.	empty
staff_roster_entries	fact	school	empty	one row per person per year	PEOPLE — one row per name the town printed,…	name by name	empty

How this is arranged

Grouped by whose money it is, because that is the question people arrive with. Two other facets ride along as badges on every table.

badge	what it tells you
Tier 1 · totals	What was appropriated and what is left, by department. Not which category, not which school, not which fund paid.
Tier 2 · categories	Whether guidance, special education or transport is over — and whether the town's books and the district's budget agree on a code. Not which school inside a category: names truncate at ten characters, so `MS GUIDANC` and `HS GUIDANC` are both 2710.
Tier 3 · accounts	What one account spent, and how much came in per fund.
fact dimension	How the table is built. Facts carry the figures; dimensions say what a key means and are joined to, never summed.
the spine	The three grains everything else describes, extracts or checks.

There is a tier below 3, and no report reaches it. Which revenue paid which expense. That is the grey edge in the join map, and it is why this archive measures appropriations and says so.

The columns you cannot leave out

None of these raises an error when omitted. Each returns a number that is the sum of two different things, which is why every one of them has been got wrong here at least once.

column	in	what happens if you omit it
`status`	annual report extracts	`checked`, `check failed`, `no check`. Summing across them mixes verified figures with unverified ones.
`level`	`account`, `ledger_snapshot`	`department` roll-ups sit in the same table as their own detail. Summing both adds the budget to itself.
`account_type`	`account`	Revenue is stored NEGATIVE. Omitting it once made the town’s whole budget compute as minus $997,871.
`period`	`ledger_snapshot`	Two periods are two different MUNIS reports of the same year, not two halves of it.
`stage`	`budget_figure`	`proposed`, `settled`, `actual`. Rule 1: a rate measured across stages is partly growth and partly the step between them.
`variant`	`budget_figure`	FY27 has four whole budgets — Balanced, Core, Level Service, Restoration. Without this you sum all four.
`v1 … v8`	annual report extracts	ORDINALS, not column names: the first column of THAT PAGE that held figures. Summing `v1` down a run adds one page’s APPROPRIATED to another’s TOTAL EXPENDED.

And one that is now fixed. fy was stored as TEXT in 44 tables and INTEGER in 18, so WHERE fy = 2023 answered against a third of the database and returned zero rows, with no error, against the rest. It is INTEGER everywhere now except three documented exceptions, and check_fy_types in build_db.py fails the build if a new table arrives without it.

Two databases, and they are not always the same one. Everything printed on this page — row counts, samples, coverage — is read from the LOCAL sources/data/lunenburg.db when the page is built. Open full table queries {api_base}, a copy pushed to Cloudflare D1 that can lag behind. Point it elsewhere with ?api=<base-url> on this page's own URL, or set LUNENBURG_API before building. If a table opens empty, or the button reports that it does not exist, that copy has not been pushed yet: python3 scripts/sync_d1.py. This is stated rather than hidden because two sources that usually agree are the ones that mislead when they stop.

Views — the joins already written for you

A view is the safe way to ask a question that needs two tables. Each one carries the splitting columns above in its own WHERE, so it cannot be got wrong the way a hand-written join can.

view	columns
`v_appropriation_vs_spend`	fy, fund, dept, name, level, original, transfers, revised, expended, encumbered, surplus, surplus_pct, doc_id
`v_budget_disagreement`	fy, stage, variant, line_key, label, source, value, is_kept, spread, kind
`v_budget_scenario`	line_key, label, fy, stage, variant, value, doc_id
`v_burn`	dept, name, fy, period, year_elapsed, spent_share, committed_share, pace_gap, available_now, doc_id
`v_function_budget_vs_ledger`	function_code, fy, period, ledger_revised, ledger_expended, ledger_encumbered, accounts, budget_lines, doc_id
`v_fund_year`	fund, name, kind, restriction, fy, period, revenue, spent, closing_balance, doc_id
`v_interfund`	fund, fund_name, org, object, name, fy, period, budgeted, revised, received, still_to_come, pct_received, doc_id
`v_line_budget_vs_actual`	line_key, label, fy, settled, proposed, actual, documents_disagree
`v_provenance`	fact, doc_id, path, basis, copy_state, rows
`v_revenue`	fund, fund_name, org, object, name, fy, period, budgeted, revised, received, still_to_come, pct_received, doc_id
`v_revenue_classified`	fy, period, fund, account, class, class_label, budgeted, received
`v_spending_classified`	fy, period, dept, department, control, control_how, voted, revised, expended
`v_staff_roster`	fy, school, page, name, role_raw, grade_or_dept, position_legacy, role_category, role_grade, classified_by, document
`v_state_aid`	fund, fund_name, org, object, name, fy, period, budgeted, revised, received, still_to_come, pct_received, doc_id
`v_transfer_history`	dept, name, fy, period, original, transfers, revised, moved_since_last, doc_id
`v_workbook_budget_vs_actual`	line_key, row, fy, budget, actual, under_budget

What this database cannot tell you

Read from money_gaps, so this list is data rather than a thing somebody remembered to write down. Each is a question the published documents do not close.

side	what is missing	why it cannot be answered
document_wanted	DESE End of Year Financial Report	Published by the state. — closes: Separates district spending by FUND, which is the one thing the town’s budget documents never show.
document_wanted	The Cherry Sheet itself — what the state told Lunenburg to expect, receipt by receipt	This archive holds no Cherry Sheet for any year. The Cherry Sheet is the state’s own statement of estimated aid and assessments to a Massachusetts town, and it is what a tax rate is set on. Without it the only measurement of aid estimate error this project can make is one net line on the Division of Local Services free cash proof — `Excess/Shortfall Cherry Sheet Receipts (CL#8)` — which lumps Chapter 70, Unrestricted General Government Aid, the abatement reimbursements and every other receipt into a single figure. So a year that came in $390,814 short can be reported as short and cannot be reported as short in anything. — closes: DLS Cherry Sheet CS 1-ER (estimated receipts) and CS 1-EB (estimated charges) for Lunenburg, FY2021 to FY2027.
document_wanted	The Town Manager’s revenue apportionment worksheet	Seen once; we do not hold it. — closes: Documents the convention by which general-fund revenue is presented as split across departments. It cannot make the edge traceable — nothing can — but it makes the convention citable.
document_wanted	The period 13 ledger	The year-end close. — closes: Reconciles FY2026 properly; period 12 is used for now.
document_wanted	The town’s Chapter 32B enrolment schedule	Not published anywhere in this archive; the town must hold one to administer the plan and set premium shares. — closes: Splits every insurance account between the town and the schools, in every year, instead of in the two years the split happens to be printed.
document_wanted	WRRS annual actuarial valuation, by member unit	Published by the retirement system. — closes: Sizes the school share of the $2.39M pension assessment.
document_wanted	Where bus fee receipts are booked	A revenue account, a fund, or a statement that they are netted before booking. — closes: The fee schedule is published and verified; `STUDENTBUS` shows zero. A charged fee with no observable destination is the clearest single gap in this model.
document_wanted	`glytdbud-expense` for the special revenue funds	The identical report the town already runs for the general fund and for each of the four enterprise funds, pointed at funds 13xx/22xx/26xx–29xx instead. — closes: Turns every restricted edge above from presumption into traced fact — athletics, lunch, circuit breaker, extended day, and every grant.
held	What the $102,000.00 is that the FY2024 enterprise balance sheet classifies two ways at once	The Combining Balance Sheet - Enterprise Funds in the FY2024 annual town report — printed twice, on printed pages 21 and 28, identically — prints 'Reserved for expenditures' as 400,000.00 in the Sewer column and 502,000.00 in the Totals (Memorandum Only) column, and prints 'Unreserved retained earnings' as 2,973,842.15 in the memorandum column where the four fund columns (1,813,817.53 Sewer, 51,333.93 Water, 181,853.82 Solid Waste, 1,028,836.87 PEG Access) add to 3,075,842.15. The memorandum column is $102,000.00 high on the first line and $102,000.00 low on the second, so the two cancel and Total Fund Equity (3,483,633.65), Total Assets, the identity and the sheet's own printed PROOF row all still tie in every column — compensating errors, invisible to any check that reads only totals. WHAT IS NOT ESTABLISHED: which presentation is right, or what the $102,000.00 is. A reservation sitting in a fund column that prints blank, a memorandum column keyed to a different worksheet row, and a reclassification made only in the total column all fit the same signature, and nothing on the page distinguishes them. These are ratepayer funds; nothing here is general-fund money. — closes: the Town Accountant's FY2024 trial balance for the enterprise funds, which would state the reserved and unreserved balance of each fund separately.
held	What the $87,293.86 is that the FY2023 annual report states twice and differently	Two tables printed in the same FY2023 annual town report state the money held in special revenue and enterprise funds, and they disagree. The combined balance sheet on printed page 24 gives TOTAL FUND EQUITY of $6,131,303.68 for SPECIAL REVENUE and $3,812,502.24 for ENTERPRISE, $9,943,805.92 together. The special revenue schedule in the same report prints a GRAND TOTAL carried forward of $10,031,099.78. The difference is $87,293.86, where FY2011, FY2013, FY2019 and FY2022 agree to the cent. It is not a misreading: the FY2023 page was transcribed and every column foots to its own printed TOTAL ASSETS, TOTAL LIABILITIES and TOTAL FUND EQUITY, the accounting identity holds in all six columns, and the long-term debt mirror agrees. So FY2023 is transcribed and deliberately NOT written into the dataset, which is why the combined balance sheet here stops at FY2022 while the special revenue schedule runs to FY2023. One explanation has been tested and RULED OUT: the fund 'Sale of Cemetery Lots' carried $87,308.09 in FY2023, within $14.23 of the gap, which suggested a fund counted in one statement and classified elsewhere in the other. It is not that — the same fund appears in the schedule in all thirteen years including the four where the two documents agree to the cent, so it is counted consistently and the near-match is a coincidence. What the difference IS remains unestablished; nothing here says which document is right, nor whether it is connected to the $17,861.24 of grant funds the town restated between the FY2022 and FY2023 reports. Those are two different amounts and nothing joins them. — closes: the Town's FY2023 trial balance or general ledger fund balances as of 30 June 2023, which would say which of its own two published tables the accounting system agrees with.
held	What the PEG Access enterprise fund actually held at 30 June 2025	Three figures for the same fund on the same date appear in the same FY2025 annual town report and no document in this archive prints a line between any two of them. The PAC report's FY25 REVENUE vs EXPENSES table (printed page 60, pdf 64) gives an Ending Balance of $30,632, computed on its own face as Beginning Balance $208,772 less Expenses $178,140 -- and the prose beside it says the $208,772 was "the budget starting balance", so the ending balance is unspent spending authority rather than money. The Treasurer's Cash schedule in the same report prints `Bartholomew- PEG Access Enterprise Fund 782,591.31$`. The Combining Balance Sheet - Enterprise Funds (printed page 21, pdf 25) prints PEG Access Total Fund Equity of $1,072,661.74, against $1,028,836.87 the year before -- a rise of $43,824.87, where the statement's own revenue less expenses is $262,565 - $178,140 = $84,425. The two do not reconcile and nothing on either page relates them. WHAT IS NOT ESTABLISHED: which quantity any of the three is; a budget balance, a bank balance and a fund equity are three different things and the report labels none of them as such. -- closes: the Town Accountant's FY2025 year-end statement of revenues, expenditures and changes in fund balance for the PEG Access and Cable Related Enterprise Fund, which the annual report does not contain.
held	What the town held town-wide at 30 June 2024 and 30 June 2025	The FY2024 and FY2025 annual town reports print NO combined balance sheet across all fund types and account groups — the schedule the reports carried every year from FY2011 to FY2023. In both editions the table of contents lists 'Balance Sheet for FY Ending June 30, 2024' at Page 21 and 'Balance Sheet for FY Ending June 30, 2025' at Page 21 respectively, and printed page 21 of each holds the Combining Balance Sheet - Enterprise Funds instead; FY2024 prints that enterprise sheet a second time on page 28, which is where its contents page says it should be. So for two consecutive years the only balance sheet the town published covers four ratepayer enterprise funds and says nothing about the general fund, special revenue, capital projects, trust and agency, or long-term debt. The consequence is that the general fund's cash, receivables, reserves and undesignated fund balance — the figures free cash is built out of — cannot be read for either year, and the balance_sheet dataset stops at FY2023 for that reason rather than by choice. WHAT IS NOT ESTABLISHED: whether the town-wide sheet was produced and dropped in layout or never produced; nothing in either report says. — closes: the Town Accountant's year-end balance sheet submitted to the Division of Local Services for FY2024 and FY2025, or the town's audited financial statements for those years.
money_in	Bus fees	charged by published policy · no destination found in any ledger
money_in	Grants received in earlier years	spent now, booked then · nine funds, no FY26 revenue
money_in	How much of free cash is the schools’	Free cash is certified for the town as a whole. Neither the DLS proof nor the Town’s own undesignated fund balance roll-forward attributes any component to any department, so nothing published says how much of it the school department’s unspent appropriation produced — nor how much of any appropriation of it reaches the schools. The projection on this site is built from budget columns and free cash is derived from actuals; they are subtracted once, side by side, and never joined. — closes: a departmental turnback schedule for the year-end close, which would split the unexpended appropriations line by department.
money_in	How much of the FY2027 enacted state-aid increase is receipts and how much is offsetting assessments	model/finance.py grows a FY2027 base of $11,404,917 + $471,121. The first is the Governor’s GROSS cherry sheet Total Receipts, printed by the FY2027 Town Meeting booklet. The second is recorded in fy27-and-the-override.md as NET — "+$471,121 net (Ch.70, charter/choice receiving tuition, Smart Growth, UGGA, vets/elderly exemptions; offset by higher charter assessments)". A gross figure and a net figure are being added, so the gross FY2027 receipt estimate the projection actually grows is not established, and neither is how much of the increase the Town keeps. This archive holds no Cherry Sheet for any year, which is the only document that prints receipts and assessments side by side. — closes: the FY2027 Cherry Sheet for Lunenburg on the enacted budget (DLS form CS 1-ER), which states each receipt and each assessment separately.
money_in	Student activity accounts	held by the school under its own authority · not in the town’s books at all
money_in	Trust fund income	scholarships, cemetery, stabilisation · extract is check-failed
money_in	What quantity the Town’s own published state-aid growth rates of 5.3% and 1.4% measure	The FY2027 Town Meeting booklet states that "From FY13-FY23 state aid increased at an annual rate of 5.3%, but in the last four years, it has slowed to approximately 1.4%" (3765-town-meeting-booklet-including-warrant, line 338), and does not say which figure it grew. Rebuilding the Town’s own `Subtotal State Aid` line from its own revenue/expenditure worksheets — 23 consecutive budget years, FY2005 to FY2027, notes/findings/state-aid-budget-series.csv — gives 4.5660% for FY2013→FY2023 and 1.9144% for FY2023→FY2027. Neither published rate is reproducible from the Town’s own published series, so a rate quoted to Town Meeting cannot be checked against anything. — closes: the Cherry Sheets themselves (DLS form CS 1-ER for Lunenburg, one per year), or the Town Manager’s working file behind the FY2027 booklet’s revenue narrative.
money_in	Where the PEG Access balance went between FY2019 and FY2021	The fund's carried balance is continuous in the report from FY2015 to FY2019 -- each edition's Starting Balance is the previous edition's Ending Balance to the cent -- and then stops being a series. FY2019 prints a Total of $573,837.37 (starting balance, plus revenue, less expenses, plus the Comcast Tech Fund of $124,796.01); FY2020 opens at $447,813.36, $126,024.01 lower; FY2021 opens at $272,000.00, which the same page's prose states is the budget rather than a balance. One document explains part of it: Article 16 of the FY2019 Annual Town Meeting established a PEG Access Enterprise Fund under MGL c.44 s53F1/2 and transferred into it the balances of the PEG Access and Cable Related Receipts Reserved For Appropriation Fund and the Comcast Tech Capital Grant Fund as of 30 June 2019. So the fund changed legal form at exactly this point, which is why the statement changes shape. WHAT IS NOT ESTABLISHED: the amounts transferred, and therefore whether the $126,024.01 step is the transfer, a closing entry, a restatement, or something else. There is also a smaller unexplained step earlier, FY2017's Ending Balance $303,909.77 against FY2018's Starting Balance $303,551.54, $358.23 apart, with no fund change to account for it. -- closes: the Town Accountant's opening and closing trial balances for the PEG Access and Cable Related funds for FY2018, FY2019 and FY2020, showing the transfers in and out.
money_in	Whether Lunenburg’s free cash is unusual for a town its size	The DLS proof carries no denominator — no population, budget, revenue or levy for any of the nine towns, Lunenburg included. So Littleton’s $11.3M against Shirley’s $272K says nothing about which is closer to its own target, and free cash as a share of the operating budget cannot be computed for a peer at all. Composition and volatility are ratios and do compare; the level does not, and is not shown. — closes: the Tax Rate Recapitulation (Schedule A) for each town, which prints the operating budget the ratio needs.
money_in	Whether a shortfall in state aid falls on the schools or on the rest of the town	Chapter 70 is 35% of the school appropriation and the town has no vote on it, so which side of the budget absorbs a miss is the question a resident actually has. Nothing published answers it. The free cash proof states one net cherry sheet figure for the whole town; the general ledger’s state aid accounts share no organisation code with any expense account, 0 of 222, so no aid dollar can be followed to anything it paid for; and the appropriation is voted before the final aid number is known. A year where aid rose and the appropriation rose with it and a year where aid was flat and the town covered the difference are different facts about the world and identical facts on the page. — closes: the Town’s FY-end recap (Form A-1) beside the Cherry Sheet for the same year, which states estimated and actual receipts on one page.
money_in	Which departments’ unspent appropriations produced the free cash	Money appropriated and not spent is the largest single component of Lunenburg’s free cash — 47% of five years of it, and 66% of the record 2025 certification. It is published as one town-wide figure across 67 departments. Prudence, a post left vacant, a project that slipped into the next year and a line that was over-budgeted from the start are four different facts about the world and one number on the page. — closes: the period 13 general ledger, showing budget, expended and turnback for each department at the year-end close.
money_in	Why DLS certifies a different amount from the free cash its own proof identifies	The Division of Local Services’ free cash proof cross-foots exactly: in all 45 town-years of the nine-town set the eleven component lines sum, to the cent, to the workbook’s own `Identified Free Cash July 1,`. But that is not the figure DLS certifies. `Current Year Calculation` — the number the Town quotes, the number the Finance Director announced, and the number that carries into the next year — differs from it in all 45, by $361,912 for Lunenburg in 2025 and by 69% for Uxbridge in 2024. The workbook prints no reconciling line, and nothing in this archive states what the adjustment is. — closes: DLS’s free cash certification letter to the Town, which states the adjustments made between the balance sheet the Town submitted and the amount certified.
money_in	Why the Town’s undesignated fund balance and DLS’s free cash differ	Two publishers describe the same 30 June balance sheet and print three different figures. The Town’s annual report roll-forward gives an undesignated fund balance of $4,236,488.24 at 30 June 2025; DLS identifies $3,716,282 of free cash on the same date and certifies $3,354,370. Each worksheet cross-foots exactly to its own printed total, and the FY2024 report’s closing balance is the FY2025 report’s opening one to the cent — so neither is in error, and no document in this archive prints a line between any pair of the three. — closes: the Town’s free cash submission to DLS (the balance sheet and the reconciliation of receivables and deferred revenue that accompanies it).
money_out	Debt service by project	two accounts for ALL town borrowing · school buildings cannot be separated
money_out	School share of Medicare, active health insurance and the insurance reserves	Five of the seven FY2026 insurance accounts — active health insurance, Medicare, life insurance, insurance cost control and Public Employee Committee expenses — carry nothing in their names that says which side of the town they are for, so the school portion of them is not separable in any year. — closes: The town’s Chapter 32B enrolment schedule, which counts insured employees and retirees by department.
money_out	School share of the pension	inside WRRS · town and school staff together, no published split
money_out	School share of the town’s health insurance before FY2026	The town’s insurance department prints one undivided `Health Insurance CH 32B` line in fifteen of the sixteen annual reports · a school share is named only in the FY2023 report and in the FY2026 ledger, so no year between them can be split. — closes: The town’s Chapter 32B enrolment schedule, which counts insured employees and retirees by department and would split every year rather than the two that happened to be printed.
money_out	What any special revenue fund bought	no expense report exists for these funds · purpose is presumed, never observed
money_out	What the PEG Access fund spent in FY2020 and in FY2024	In both years the annual town report prints two totals for the same year's PEG expenses and they differ. FY2020, both tables on printed page 75: `FY20 Line Item Expenses by Percentage` foots to Total $116,933.22 and `FY20 REVENUE vs EXPENSES` prints Expenses $116,933.54 -- 32 cents apart. FY2024: the statement on printed page 63 prints Expenses $174,150 and the line-item table on printed page 64 foots to Total $174,151.16, and since that statement is printed in whole dollars throughout, rounding 174,151.16 gives 174,151 rather than 174,150. Each table foots to its own printed total exactly, so neither is internally wrong; they disagree with each other. FY2017 is a third case of the same shape: its ten printed expense lines add to $74,665.74 against a printed TOTAL of $74,215.73, a difference of $450.01, of which $450.00 is the `Vendor Expenses` line that the facing pie chart also omits. WHAT IS NOT ESTABLISHED: which figure is the year's expenditure in any of the three years. -- closes: the MUNIS year-to-date expense report (glytdbud) for the PEG Access and Cable Related Enterprise Fund for FY2017, FY2020 and FY2024, the same report the town already runs for the general fund.
money_out	What the capital transfers bought	two transfer lines · no project detail
money_out	What the four `ADJ EXP` journal entries moved, and from where	$254,121.18 — 65% of everything that reached the athletics revolving fund in FY2025 — arrived on four general-journal entries whose entire description is an expense adjustment made per a memo. An entry raising cash in a fund and labelled as an expense adjustment fits at least three readings: expenses charged here and later moved elsewhere, a transfer in from another fund, or a correction of charges posted to the wrong account. Those are different facts about the world and identical facts on the page, and the same shape of entry could be adjusting any fund in the town without either budget document showing a trace. — closes: the five memos the ledger names, dated 08/12/24, 01/30/25, 05/02/25, 07/02/25 and 08/20/2025.
money_out	Whether a general fund athletics line is net of the revolving fund	In FY2024 the town’s comparable athletics lines came to $124,301 against $351,643 the district’s own sport-by-sport workbook totals for the same categories — so the appropriation covered 35% of the cost. That is measurable for FY2024 and FY2025 only, because those are the only years anybody published both halves. For every other year, and every other programme, a line that rose because the cost rose looks exactly like a line that rose because a fund stopped paying part of it. — closes: the MUNIS `Account_Detail` export for the general fund athletics orgs S3066672 and S3066671, all object codes.
people	How many children Chapter 70 is actually paid for	The aid is calculated on a `Foundation enrollment` of 1,599 in DESE’s FY27 Chapter 70 summary. DESE publishes three other counts for the same district — Student Headcount, In-District FTE Pupils and Total FTE Pupils — and in the latest year published they are 1,563, 1,568.9 and 1,665.9. Four numbers, one agency, none of them equal, measured on different dates and counting different things. So every per-pupil figure on this project, including aid per pupil, is a division by a count whose definition has to be stated with it, and no published document reconciles the foundation count to the children in the buildings. — closes: DESE’s FY27 foundation enrollment worksheet for Lunenburg, which states the October 1 count by grade and category the formula was run on.
people	How many children play sports	Two town documents give 593 and 649 for the same year and count different things: one counts fee categories, the other counts participations by sport, and a participation is not a child — one student playing three seasons is three of them. Some workbook rows are negative, subtracting out-of-district athletes back out, so a season total is a net figure rather than a headcount of bodies. Nothing published is an unduplicated count. — closes: an unduplicated athlete roster, or a participation report that states whether it counts students or participations.
people	How many people the schools actually employ	The town publishes per-school staff ROSTERS in every annual report, FY2011 to FY2025 — a list of names. A roster carries no FTE, so a 0.4 music teacher and a full-timer are one row each, and it is a point in time, undated within the year. A count of names the town printed is a real quantity and it is not a staffing level. — closes: a staffing report stating FTE by position and year.
people	Whether a budgeted position was filled	A budget line is money the town has to raise, not a person in a chair. A line funded all year for a post vacant half of it looks identical on the page to one that was filled. Bounded, not settled: the rosters show who was printed, which narrows the question without answering it. — closes: a filled-position list, or payroll by position rather than by name.
people	Which fund pays which post	This is the load-bearing one. The rosters name people and the budget names dollars, and nothing published joins them to a funding source. So a paraprofessional line rising cannot be distinguished from a grant that was paying for those posts ending — which is exactly what the special education escalator rests on. — closes: DESE's End of Year Financial Report, which separates spending by fund.

Examples — real questions, run against the live database

Every query below is run against the live database when this page is built, so a question that stops answering breaks the build rather than sitting here looking answerable. Two of them return nothing, and that is the honest state of the archive rather than a fault.

partlyEvery source of revenue in FY2022, by name and amount, with our classification

SELECT fy, printed_name, label, class, amount, document, page FROM revenue_history WHERE fy = 2022 ORDER BY amount DESC

100 rows

fy	printed_name	label	class	amount	document	page
2022	REAL ESTATE TAXES	REAL ESTATE TAXES	levy	28756474.97	4129-fy-2022-annual-town-report.pdf	26
2022	CH 70 SCHOOL AID	CH 70 SCHOOL AID	state	7823618.0	4129-fy-2022-annual-town-report.pdf	26
2022	TRANSFERS FROM OTHER FUNDS	TRANSFERS FROM OTHER FUNDS	transfer	2391150.83	4129-fy-2022-annual-town-report.pdf	26
2022	MOTOR VEHICLE EXCISE	MOTOR VEHICLE EXCISE	local	1839037.96	4129-fy-2022-annual-town-report.pdf	26
2022	UNRESTRICTED GEN'L GOVT AID	UNRESTRICTED GEN'L GOVT AID	state	1162230.0	4129-fy-2022-annual-town-report.pdf	26

…and 95 more.

class says what KIND of revenue it is — levy, state aid, local receipt, transfer — which is a property of where the money came FROM. It does not say what the money went on to pay for, and for anything landing in the general fund nothing can: that money is fungible by law and no document apportions it. Only restricted funds have a traceable destination, and those are in money_edges.

The classification is joined in at build time from money_classification, so the name the report printed and the group we assigned are in one row.

not yetThe same question for FY2023

SELECT fy, printed_name, amount FROM revenue_history WHERE fy = 2023

0 rows

Returns nothing, and that is the correct answer rather than a failure. revenue_history carries only years whose extract was CHECKED against the report’s own printed total. FY2023 has 10 unchecked rows in annual_report_receipts; loading them here would make five verified years and one guess look identical.

answersWhat was budgeted for the Superintendent, FY2022 to FY2025

SELECT fy, stage, label, value FROM budget_figure WHERE label LIKE '%Superintendent%' AND fy BETWEEN 2022 AND 2025 ORDER BY fy, stage

10 rows

fy	stage	label	value
2022	actual	Superintendent	165000.0
2022	proposed	Superintendent	157000.0
2022	settled	Superintendent	165000.0
2023	actual	Superintendent	171000.0
2023	proposed	Superintendent	171000.0

…and 5 more.

One row per stage, and they are different quantities: proposed is what was asked for, settled what was voted, actual what the later document reported. Rule 1 — never compute a growth rate from one stage to another.

budget_figure is one row per line × year × stage × variant. Filtering on a label works here because "Superintendent" is unique; most lines are not, which is what line_key is for.

partlyEvery line in the school budget for FY2024, by name and amount

SELECT fy, stage, label, value FROM budget_figure WHERE fy = 2024 AND stage = 'proposed' ORDER BY value DESC

319 rows

fy	stage	label	value
2024	proposed	Health Insurance	3076306.0
2024	proposed	H.S. Teachers/Regular	2899395.0
2024	proposed	M.S. Teachers/Regular	2027997.0
2024	proposed	E.S. Teachers/Regular	1527124.0
2024	proposed	P.S. Teachers/Regular	1033468.0

…and 314 more.

FY2024 has no settled stage in the archive — only proposed and actual. So this is what the district ASKED for that year, not what the town voted. The stage you get is not the stage you named, and the coverage table below is the only place that difference is visible before you query.

Which stages exist varies by year, so a query written for one year can silently return nothing for another.

answersWhere the money actually went in FY2026, by department, with who decides

SELECT department, control, voted, expended FROM v_spending_classified WHERE fy = 2026 AND period = 9 ORDER BY expended DESC LIMIT 12

12 rows

department	control	voted	expended
SCHOOL DEPARTMENT	delegated	26247474.0	15736640.86
WRRS ASSESSMENT	assessed	2392572.0	2392572.0
INSURANCE	bargained	3713520.0	2240086.02
POLICE DEPARTMENT	discretionary	2406172.0	1686886.24
FIRE DEPARTMENT	discretionary	1717588.0	1169012.23

…and 7 more.

control is our classification of who holds the decision, not a published field. voted and expended are the town’s own figures; the column beside them is ours and is labelled that way everywhere it appears.

Spending IS traceable to a department, which is why this answers where the revenue question above cannot.

answersThe full ledger for one account across the year

SELECT l.fy, l.period, a.name, a.account_id, l.original, l.revised, l.expended, l.available FROM ledger_snapshot l JOIN account a USING (account_id) WHERE a.account_id LIKE '0100-S%' AND l.fy = 2026 ORDER BY l.expended DESC LIMIT 10

10 rows

fy	period	name	account_id	original	revised	expended	available
2026	12	HEALTH INS	0100-S5991992-570001	3701195.0	3701195.0	3622881.65	78313.35
2026	12	HS TEACHER	0100-S2066101-511001	2719099.0	2719099.0	2791452.42	-72353.42
2026	12	MS TEACHER	0100-S2055101-511001	2038757.0	2038757.0	2022044.44	16712.56
2026	12	ES TEACHER	0100-S2044101-511001	1552471.0	1552471.0	1556544.17	-4073.17
2026	12	PS TEACH R	0100-S2072101-511001	1172728.46	1172728.46	1095292.39	77436.07

…and 5 more.

period is a point in time inside the year, not a stage. Two periods of the same year are two different MUNIS reports and adding them double-counts.

The S prefix is the structural way to find school accounts — it does not depend on any name, and MUNIS names are ten characters typed by a person.

Generated by scripts/build_schema_page.py from the live database on 2026-09-07. Do not edit — run the script. The CSVs in sources/data/ are the source of truth; this database is a derived read model, rebuilt from scratch on every run.