A repeatable spreadsheet reconciliation checklist
A reliable reconciliation proves more than whether two totals match. It explains which records paired, which did not, and which rules were used to classify differences.
Start with file identity and scope
Record the file names, extract dates, row counts, worksheet names, and hashes. Confirm both extracts cover the same population and cutoff before comparing individual values. A synthetic control record might be INV-100 with 100.00 in both files and INV-101 with 100.01 versus 100.00; the result should show one matched record and one value exception when tolerance is zero.
- Confirm reporting period and population.
- Check row and column counts.
- Keep a SHA-256 hash for each source file.
Control the matching population
Measure blank keys, duplicate keys, added records, and removed records separately. A matching algorithm cannot safely decide which duplicate record belongs to which counterpart without an additional rule.
Make tolerances explicit
A tolerance should correspond to rounding, timing, or another understood process. Record its unit and whether it is absolute or percentage-based. Never apply one broad tolerance to columns with different meanings. The engine returns at most 10,000 displayed differences and uses the same XLSX and CSV/TSV limits as the Excel compare tool.