Editorial draft for professional review. No unapproved author or reviewer is named. Updated:
What you will check
- Describe the specific signal rather than alleging fraud.
- Check the source and data quality before the business explanation.
- Distinguish a confirmed finding, justified repetition and missing evidence.
Describe the signal precisely
Instead of labelling an entry suspicious, record what the rule identified: a repeated document number, unusual account relationship, amount outside an expected range or posting outside normal hours. Include the period, source and comparison group.
A threshold useful for one transaction type may be unhelpful for another. Separate ordinary invoices, payroll, closing entries and one-off events before comparing them.
Check data quality before explaining the business event
A repeated import can resemble a duplicate posting. A misread decimal separator can produce an extreme amount. Check the original record, identifier and mapping before investigating the business reason for a variance.
Record whether the review covered the whole population or a selected sample. A large row count does not compensate for an incomplete source or establish the quality of audit evidence by itself.
Synthetic example: two identical amounts
Two EUR 2,400 invoices from the same supplier could be duplicates or two months of rent. Compare invoice numbers and dates, service periods, the agreement, reversals and payment links. Do not automatically delete a row because the amounts match.
If both rows carry the same source identifier, check for a duplicated export or import. If their identifiers differ, review the business documents. The outcome may be a confirmed duplicate, a justified repetition or an unresolved finding.
Statistics point to questions, not verdicts
Benford analysis compares leading-digit distributions, but it is not appropriate for every dataset. Assigned identifiers, restricted ranges and predetermined prices can produce structures unlike those expected in such a comparison.
IAASB material on analytical procedures emphasises professional judgement about the suitability of a procedure and the evidence obtained. Do not turn a single rule, statistical signal or AI summary into an allegation of fraud.
Close the finding with evidence and an owner
For a material finding, retain the signal description, checked source, explanation, conclusion, reviewer and date. If evidence is missing, state what is still needed. Accounting corrections should follow an approved process that preserves the record.
In a FynTwin demonstration, inspect how findings and underlying data are presented. Use AI commentary to help frame questions and summaries, then substantiate conclusions with documents and professional review. Revisit the ledger-preparation guide when checking the source.
Work through an example
The worksheet contains two EUR 2,400 invoices from the same synthetic supplier. One relates to January and the other to February. Matching amounts do not establish duplication.
Synthetic learning example. This is not a FynTwin export or a real client report. Amounts are in EUR; VAT, opening balances and closing entries are omitted for simplicity. Account codes are illustrative. The CSV is not a universal import template.
| Check | Expected result |
|---|---|
| Amount and supplier | Match on both invoices. |
| Document numbers | RENT-01 and RENT-02 differ. |
| Service period | January and February - potentially a recurring monthly expense. |
| Open check | Obtain the invoices, agreement and payment links before concluding. |
The initial CSV status is needs_evidence. Do not close the finding as justified solely on the ledger description; document what you actually verified.
Downloadable resource
Download the finding review worksheet (CSV)Two synthetic records with fields for evidence, explanation, conclusion, reviewer and date. No real supplier data is included.