There is a difference between an accrual for something genuinely uncertain and an accrual for something entirely certain that simply has not been invoiced yet. The first is accounting. The second is a workaround for a timing problem, repeated monthly, and treated as if it were the first.
The test
Look at last month’s accruals and mark each one: did we know the amount, within a few percent, before the period ended? For most teams somewhere between half and two thirds get marked yes.
Those are not judgements. They are known facts that arrived after the door closed — usually because the supplier invoices in arrears, or because an internal approval had not finished, or because the information existed in a system the ledger cannot see.
Each one has a different fix
- Known amount, late invoice. Accrue from the purchase order or the contract, not from the invoice. The number is already in the commitment.
- Known amount, late approval. This is an approval design problem, not an accounting one — see approval rules people actually follow.
- Known amount, wrong system. The usage data exists; nothing connects it to the ledger. This is the one that stays broken longest because it belongs to nobody.
Why they persist
Because an accrual works. It produces defensible numbers, the auditors accept it, and the cost is spread thinly across every month rather than concentrated anywhere anyone would notice. The estimate becomes institutional, someone eventually stops checking it against the actual, and the variance starts drifting quietly.
What to do this quarter
Take the three largest recurring accruals and trace where the number could have come from earlier. In my experience two of the three have a source that predates the close by weeks, and connecting it is a smaller job than the monthly estimate everyone has stopped questioning.
