Platform
A liability you can drill into
If you cannot get from the balance-sheet number back to the transaction lines that created it, you do not have an accrual — you have an opinion.
The financial problem
An estimated accrual is wrong for three quarters and then corrects violently.
- True-up variance of 10–30% is normal where the accrual comes from an assumed rate on aggregate volume.
- The variance concentrates — it lands in one period, usually at year end, distorting a period that had nothing to do with it.
- The estimate is fed its own error: next year’s assumed rate derives from this year’s settlement, which included the catch-up.
- When the auditor asks why the number was what it was, the method is the only available answer.
The technical problem
The accrual is not linked to the transactions, so nothing flows through.
- Computed in a spreadsheet from period totals, it cannot answer which lines produced it.
- A restatement of the base data does not propagate — the unwind is manual.
- Without a locked period the source data keeps moving underneath a number that has supposedly closed.
- Reporting aggregates live over transaction tables, so a useful cut takes minutes to hours and nobody explores.
The spine
Accrue → review → claim → settle → reconcile
Accrual engine
Computed in-database from transaction lines against locked periods, posted to mapped GL accounts with exact decimal maths.
Accrual review
A staging object with calendars and approval gates. Reviewers see variances and outliers, not every line.
Claims & recovery
Raised in-window, tracked through the partner's process, settled — including split, overpay and write-off.
Payouts
What you owe out, held as a distinct object with its own approvals and segregation of duties.
Period close
Locked periods every engine validates against, so a closed month cannot be silently restated.
Materialised reads
Financial positions served from snapshots rather than live aggregation, so the answer is instant.
Benchmarks
What a defensible close looks like
| Metric | Typical today | Target |
|---|---|---|
| Accrual variance at settlement | 10 – 30% | <2% |
| Posted balances drillable to source rows | Partially | 100% |
| Days to gross-to-net close | 25 – 45 | <5 business days |
| Reconciliation effort per close | Days of manual work | 60 – 90% reduction |
| Engines validating against period locks | Some | All, without exception |
| Deduction value resolved without research | 40 – 70% | >85% |
Bring your last close.
The most revealing first conversation is usually a walk through your most recent month-end: what was estimated, what was reconciled by hand, and what the true-up was.