Clustering
Identify account-to-account posting patterns that occur so rarely they fit no normal business process.
- General ledger alone
- Compares with last year
How the result is read
What counts as a flow has to be stated: observed on a two-line journal, derived by pro-rata allocation on a multi-line one, where the ledger never recorded which credit answered which debit. Rarity alone is not risk — the year-end close is made of rare flows by design — so the ranking is rarity weighted by amount, and a rare flow recurring at the same point each year is periodic rather than exceptional.
The worksheet, on the demo engagement
Clustering, as the application draws it for Harrow Precision Engineering Ltd, the fictional demo engagement — one row per account-pairing label.
- Outliers (no cluster)
- 3
- Top rare flow
- 4000 to 2800 x3
Result grid3 rows
| Rare account flow | Frequency | Example journal | Note |
|---|---|---|---|
| 4000 Revenue to 2900 Suspense | 3 | JV-43880 | outlier — fits no cluster |
| 6500 Prof fees to 2900 Suspense | 2 | JV-44120 | expense to suspense |
| 4000 Revenue to 1200 AR | 1 | — | within normal cluster |
In the application
Clustering is filed under Journal entries, index letter B. "Run all analyses" runs it with every other analysis the ledger can answer. The worksheet states the population it tested on its face — or "Not applicable" with the reason when a column it depends on is empty — and the Excel workbook the engagement exports files it under B, beside an index of the source files with their SHA-256 fingerprints.
Read it with
Run your next audit file test without the Excel crash.
One free engagement to evaluate, valid 60 days. The ledger never leaves the machine.
- 20 minutes
- Runs on a fictional engagement
- No client data required