Review variance invoice by invoice against a locked copy of last week's forecast. First determine which forecast rows collected in the assigned week, earlier, later, or remain open. Then classify why the evidence changed. Build the next forecast from a new reconciled receivables snapshot without rewriting the prior expectation.
A useful variance review connects two different source states. The prior forecast contains an open balance, assigned week, evidence, confidence label, owner, and next review date. The current accounting source contains observed receipt applications and remaining open balances. Intuit's payment guidance shows that payments are recorded against selected outstanding invoices, which is why an invoice-level comparison is stronger than comparing only total cash.
Lock the comparison population
Before calculating or describing variance, save:
- The prior forecast version and its exact week date ranges
- The prior invoice open amount and expected collection week
- The evidence and confidence label used at that time
- The actual-receipt cutoff and current snapshot date
- The current invoice balance or applied-payment evidence
- Any credit, dispute, approval, or source correction observed later
Do not remove a forecast row because it became inconvenient. Do not add an after-cutoff receipt to the prior actuals. These controls keep the review reproducible.
Use outcome and reason codes
Assign each prior row one observed outcome:
- Collected in the assigned week
- Collected earlier than assigned
- Collected later than assigned
- Partially collected with a remaining balance
- Still open at the cutoff
- Unresolved because the source cannot yet be reconciled
Then add a separate reason code only when the evidence supports it: promise held, promise missed, approval completed, approval delayed, dispute opened, dispute resolved, payment applied after cutoff, credit changed the balance, or original source error corrected. Leave the reason unconfirmed when the records do not explain the outcome.
Separate timing, amount, and scope variance
Do not compress every miss into one variance number. Classify three independent dimensions:
| Dimension | Comparison | Example treatment | | --- | --- | --- | | Timing | Assigned week versus observed receipt week | Keep the original amount and explain the dated movement | | Amount | Forecasted receipt amount versus observed applied amount | Record a partial receipt, credit, deduction, or unresolved residual | | Scope | Prior invoice population versus current source population | Identify a new invoice, removed duplicate, report filter, or corrected source row |
One invoice can have more than one dimension. A partial receipt one week late is both an amount and timing variance. A filtered-out invoice is a scope issue, not a customer collection miss. This split keeps operating changes from being misdiagnosed as forecasting error.
Improve rules rather than invent precision
The purpose is to identify which evidence rules need attention. Repeated misses tied to stale promises may justify a shorter promise-review interval. Repeated approval delays may justify an earlier internal handoff. A single outcome does not establish a universal customer probability or benchmark.
Oracle's forecast-method guidance distinguishes forecast approaches and notes that actual-versus-forecast behavior can inform future forecasts. For a small B2B service company, keep the improvement loop understandable: preserve invoice identity, record the original assumption, observe the outcome, and change a rule only when the evidence is repeatable.
The four-week forecast guide explains invoice-level weekly assignment. The forecast template supplies traceable fields, and the collection-assumptions template separates source dates from evidence and confidence.
Where AR Pulse fits
AR Pulse analyzes an authorized customer-provided QuickBooks receivables export and returns a ranked cash-action list plus a four-week collections forecast. It does not request QuickBooks credentials, post payments, contact customers, or promise that an assigned week will collect. Finance teams that want a reviewable starting forecast can apply for AR Pulse, then use the observed variance to strengthen future evidence rules.