Cash forecast accuracy: how to review misses without rebuilding the model

Save the forecast before replacing estimates with actuals. Record its preparation date, opening cash, collection assumptions, and major payments. Without that version, the team can mistake an updated forecast for evidence that the original estimate was accurate.

Select a completed week or month and reconcile actual cash to bank records. Match the same entities, currencies, and accounts. A forecast covering operating cash cannot be tested against a total that includes newly added restricted accounts without a bridge.

Classify differences as receipts or payments occurring earlier or later, amounts changing, or items missing entirely. A customer paying one week late needs a different correction from an invoice that was never valid.

For illustration, a $15,000 receipt expected on Friday but collected Monday is a timing miss. A $15,000 invoice reduced to $12,000 is an amount change. Combining them under revenue shortfall obscures the process that needs attention.

Review collection promises, payroll changes, tax payments, inventory commitments, and large one-time bills. Identify which owner knew about the change and when finance received it.

Recurring surprises can indicate a missing input process. A purchasing commitment absent from the forecast should trigger a source change, with future purchase orders or approved commitments supplied on a regular cadence.

Revise customer collection timing, payment schedules, or forecast coverage based on the observed pattern. Avoid adding complexity to small expenses while a few large receipts drive most of the error.

Use ranges or scenarios when the input is uncertain. Record the uncertainty instead of assigning a precise date that nobody can support.

Note the forecast miss, its cause, the correction, and the owner. Track whether the same issue recurs in later periods. Review the effect on actual management decisions, such as hiring or payment timing.

The forecast should become easier to maintain as inputs improve. Retaining the old version and an explained variance lets the team distinguish better planning from a model that merely looks accurate after being edited.