Guide · Excel finance automation
Short answer
A close spreadsheet should prove, every time it runs, that it read all the data, read it correctly, counted nothing twice and agrees to the ledger. Seven simple checks do most of that work, and a failed check should stop the result from being presented as clean.
Spreadsheet errors are rarely clever. Research collected by the European Spreadsheet Risks Interest Group has documented errors in a large share of operational spreadsheets, and most come from ordinary causes: a missing row, a duplicated paste, a date read as text. Checks catch those causes regardless of who built the formulas.
Put the checks on one visible sheet, each with a status of Pass, Flagged or Fail, so the reviewer sees them before the results.
If a system export changes its layout, every formula downstream can silently point at the wrong column. Check that each expected header exists before anything else runs. A missing column should stop the run.
Count the rows where an amount is not a number or a date is not a date, for example with COUNTIFS on helper columns or with explicit data types in Power Query. One unreadable amount means the total is wrong.
Exports and copy and paste both create duplicates. Build a key from the fields that should be unique, such as reference, date and amount, and count how often each key appears. Report duplicates rather than quietly removing them, because sometimes they are real.
Record how many rows each export contained and compare it with how many rows the workbook used. This is the simplest completeness check and catches truncated pastes and filtered ranges.
The opening balance, period activity and closing balance in the workbook should each agree to the general ledger or control report. Show both figures and the difference, not only a tick.
A reconciliation is complete when the difference is fully explained by listed items, such as timing items or unapplied cash, and the unexplained amount is zero. If an unexplained amount remains, the status should read "needs review", never "reconciled".
Each unmatched item should carry the rule it failed, the source row and who needs to look at it. An exception list without reasons just moves the investigation to someone else.
Separate two kinds of failure. Missing columns or unreadable data mean no reliable result is possible, so the run should stop. A reconciliation difference is different: show it, mark the run as needing review and let a person investigate. In both cases, a failed or blocked run must never look like a clean one.
The Excel workflow demo runs all seven checks on synthetic data. You can introduce a duplicate receipt and watch the status change.
Published 18 October 2026
General educational information only. Not accounting, tax, audit or legal advice. MatchPass Labs provides workflow automation and is not an accounting firm.
Common questions
Describe it in a few sentences. No files or preparation needed. If it is not worth automating, we will say so.