MIDA / DATA QUALITY

A practical checklist before turning a CSV into a report

Catch simple file problems before they become persuasive charts.

Keep one record per row

Use a single header row and avoid subtotals, merged cells or notes mixed into the data. Choose a clear grain: one order per row or one order line per row. Combining both in the same file can duplicate amounts. Export only the data sheet and keep the original export for reference.

Make identifiers and dates explicit

Include an order identifier and use YYYY-MM-DD for dates. Decide whether the date represents creation, payment, completion or refund. A channel name is not a unique order key by itself. For cross-channel files, use the source or channel together with the order ID when checking duplicates.

Check amounts and currencies

Use numeric amounts without currency symbols or thousands separators. Make decimal conventions clear. Add a currency column for files containing more than one currency. Never add SGD and VND into a single headline total without an approved conversion method and date.

Decide how statuses affect the report

Document which statuses are sales, refunds and cancellations. In the Workbench, paid, completed and delivered count as sales. Refunded and returned records subtract their positive amount; cancelled records are excluded from monetary totals. This is a simple reporting definition, not a replacement for your finance policy.

Review exclusions before acting

Import the file, map the fields and inspect validation results. Invalid records and duplicate keys should be explained, not silently counted. Export the quality report for follow-up with the source owner. Reconcile the accepted records against the original report before using them for decisions.