Ten rows. Four questions. One explainable close.

Write down the rules before opening the file

The period is 21–27 September 2026, including both dates. Every amount is AUD. The fictional calculation uses sales excluding the supplied tax, with each line rounded to the nearest cent. Mia Chen has a 40% service rate and 10% retail rate. Noah Reed has 45% and 12%. These are example inputs, not suggested pay arrangements.

The source contains a reported commission column, but that number is evidence to compare, not the rule used to calculate the answer. For this case, the reviewer has confirmed the supplied rate register. The voucher is outside the example’s commission scope. Keeping those assumptions separate from the file prevents a spreadsheet total from silently choosing the policy.

The ten rows, without omissions

The table follows the physical order of the data records in the download. CSV row numbers are one greater because the header occupies row 1. The second A102 is deliberately identical on the fields used for duplicate detection. Its extra example note differs so you can see that an investigative note does not make it a second sale.

Question one: where did the refund come from?

A105 returns half the value of A100. Its signed gross amount is −$55.00 and supplied tax is −$5.00, producing a −$50.00 basis. At Mia’s 40% service rate, the adjustment is −$20.00. The source reports zero commission, so the row also has a commission difference. One review item can contain more than one issue.

The approval note links A105 to A100, identifies the half refund and states the −$20.00 effect. This is the fictional reviewer’s confirmed treatment. An unexplained negative number would not establish whether the refund belongs to the staff member, item or arrangement being reviewed.

Question two: which A102 survives?

Records 3 and 7 represent the same colour sale in overlapping exports. The first remains included at $90.00. The second is excluded with a reason naming the retained copy. The engine initially holds the repeated row, so excluding it does not cause another $90.00 drop from the displayed provisional commission. That distinction explains why resolving a warning can change the review count without changing the total.

Do not remove both copies, and do not approve the second merely because the reference is valid. Both statements can be true: the underlying sale is genuine, and the later imported representation of that sale should contribute nothing.

Questions three and four: a rate and a category

A106 has an $80.00 basis. The source’s $32.00 corresponds to 40%, while Noah’s supplied service rule gives $36.00 at 45%. The case approval records the rate evidence and the $4.00 difference. A difference is not, on its own, proof that the source is wrong; the example supplies the missing policy decision.

A107 has a voucher category, which the importer does not quietly reinterpret as a service. It remains a hard review item until the reviewer excludes it under this case’s stated scope. This treatment is specific to the fictional file. It does not establish a general commission rule for every gift-card sale or redemption.

Rebuild the result independently

The raw file adds to $1,056.00 gross, including the repeated sale and voucher. Their combined $330.00 gross is excluded. The remaining $726.00 includes $66.00 supplied tax, giving a $660.00 basis. Mia and Noah each contribute a $330.00 net basis, but their category mix and rates produce different commission totals.

Mia’s service commissions are $40.00, $60.00 and $40.00, with $3.00 retail and a −$20.00 refund: $123.00. Noah’s $90.00 and $36.00 service commissions plus $6.00 retail total $132.00. Together they equal $255.00. This independent addition is more informative than simply trusting a green completion indicator.

Repeat the close and inspect what remains

Download the source and the decision register. Import the source into an empty workspace, choose the stated dates and basis, and enter both staff rules. You should see four rows needing review. Resolve the refund, repeated sale, rate difference and voucher exactly as documented. Ordinary rows do not need invented approval notes.

We executed this fixture against the calculation engine on 27 September 2026. It produced zero pending rows after those decisions, $660.00 net and $255.00 commission. Finalisation retained all ten reviewed rows in the snapshot, including both exclusions, and eight eligible fingerprints. The draft became empty because every source row was inside the chosen period.