Month-End Close — Instrumented

Six closes, 13 tasks with real dependencies, sign-off states, and a flux review with materiality thresholds. The close is a process with a critical path, not a heroic sprint — and this page names the current bottleneck. Synthetic, seeded data.

6.9 days

December close — down from 8.9 in July

2.5d

Longest task on the critical path — Revenue cut-off review — this is what to attack next

Flux breach missing commentary (Nov) — close held until resolved

Task-instances with prepare → review → approve ordering proven

max($50K, 5%)

Flux materiality — every breach explained or the close doesn't certify

Days to close — before and after attacking the path

Bank reconciliation was automated in October (blue → green). July's critical path ran subledger → bank_rec → rev_cutoff → flux → reporting; December's runs rev_data → rev_rec → rev_cutoff → flux → reporting. The improvement did not come from anyone working faster on average — it came from removing 2.7 days from the one chain that gates everything. The next day of improvement lives in the path's longest task — currently revenue cut-off review — and the dashboard will keep saying so until it moves.

December close checklist

Task	Owner	Start (day)	Duration	Finish
Close AP / AR subledgers	AP accountant	0.0	1.4	1.4
Payroll & benefits accrual	Payroll lead	0.0	1.0	1.0
Usage & billing data cut	RevOps	0.0	0.9	0.9	● critical path
Stock comp expense	Senior accountant	0.0	0.8	0.8
Revenue recognition run	Revenue accountant	0.9	1.3	2.3	● critical path
Bank reconciliations	Staff accountant	1.4	0.7	2.1
Opex accruals & prepaids	Senior accountant	1.4	1.4	2.8
FX rates & remeasurement	Senior accountant	1.4	0.8	2.1
Intercompany reconciliation	Senior accountant	2.1	1.2	3.4
Revenue cut-off review	Controller	2.3	2.5	4.7	● critical path
Consolidation & eliminations	Controller	3.4	1.0	4.4
Balance-sheet flux review	Controller	4.7	1.3	6.0	● critical path
Management reporting package	Head of Finance	6.0	0.9	6.9	● critical path

Highlighted rows form the critical path — recomputed from the dependency graph, not asserted. Validation requires the path's length to equal days-to-close exactly; a "critical path" that doesn't is a diagram, not a measurement.

Balance-sheet flux — breaches of max($50K, 5%)

Close	Line	Move	Commentary
2025-08	Deferred revenue	+$834K	Three multi-year prepays billed annually in advance
2025-10	Cash	−$1,400K	Q3 bonus payout plus annual insurance premium
2025-11	Prepaid expenses	+$294K	COMMENTARY MISSING — close not certified
2025-11	Accrued liabilities	+$343K	Legal settlement accrual per outside counsel range
2025-12	Lease liabilities	+$620K	GPU capacity agreement commenced (embedded lease, ASC 842)

November's prepaid-expenses move breached materiality with no commentary attached. The engine refuses to certify that close until the line is explained — the control catching the gap is the demonstration. (December's lease line, incidentally, is the GPU embedded lease from the ASC 842 register commencing.)

Validation — re-run before every publish

PASS no task starts before its dependencies finish (6 closes) — max violation 0.000000 days
PASS sign-off ordering holds for all 78 task-instances — prepare -> review -> approve, no exceptions
PASS critical-path length equals days-to-close, every month — max gap 0.00000000 days
PASS flux control: the one seeded commentary gap is caught, only it — flagged [('Prepaid expenses', '2025-11')]
PASS materiality threshold max($50K, 5%) applied exactly — recomputed line by line, month by month
PASS December close at least 2 days faster than July — 8.9 -> 6.9 days
PASS automation moved the critical path off bank reconciliation — July path ends reporting; December path ends reporting

Synthetic, seeded data (random.seed(1231)). Generator, DAG, and checks: github.com/Lumimama/finance/month-end-close.