If a vendor delivered a service in March but doesn't invoice until April, accrual accounting still records that expense in March, the month the service was actually received, rather than waiting for the invoice to arrive. This gives a more accurate picture of a given period's real financial activity than simply tracking cash movements.

Calculating the right accrual requires judgment: estimating an unbilled amount based on historical patterns, contract terms, or partial delivery evidence, then reversing the accrual once the real invoice arrives. Getting this wrong in either direction distorts that period's reported results.

Frequently Asked Questions

How is accrual accounting different from cash accounting?

Cash accounting only records a transaction when money actually moves. Accrual accounting records it when the underlying economic event happens, the service delivered, the revenue earned, regardless of when the cash follows. Most companies of any real size are required to use accrual accounting for financial reporting.

Why are accrual calculations one of the harder parts of monthly close?

Because the actual invoice usually hasn't arrived yet, so the amount has to be estimated from historical vendor patterns, contract terms, or partial evidence, and then trued up once the real number is known, a judgment call rather than a simple lookup.

How does an AI agent learn to estimate accruals accurately?

By tracking patterns over multiple close cycles, noticing, for example, that a specific vendor always bills roughly ten days late, or that a certain expense category spikes predictably in a given quarter, and adjusting its estimates based on that accumulated history rather than a static rule.

What happens if an accrual estimate turns out to be wrong?

It gets reversed and corrected in the period the real invoice arrives, a normal and expected part of the process, though a pattern of consistently inaccurate estimates for a given vendor or category is worth investigating rather than repeatedly correcting after the fact.