Auditable Excel & CSV Cleanup: Messy Export to Verified Delivery

The challenge Messy customer exports often combine inconsistent headers, spacing, phone formats, dates, currencies, invalid values, and duplicate candidates. A useful cleanup must correct what can be corrected without silently inventing data or losing source traceability. The solution I built SheetMedic, a configurable Python workflow for CSV, TSV, and Excel exports. It maps an agreed schema, applies explicit cleaning rules, flags invalid or ambiguous values, identifies duplicate candidates, and reconciles every input row with the delivery. Demonstrated result The synthetic 12-row demonstration produces 5 accepted rows and 7 review rows, reports 5 validation issues, and identifies 2 duplicate groups containing 3 non-canonical candidates. All 12 source rows remain traceable. The delivery includes a four-sheet Excel workbook, accepted and review CSVs, an exception report, an exact cell-level change log, reconciliation evidence, and file hashes. Independent verification The workflow passes 122 product tests and 69 retained adversarial regression checks. Independent public-data testing covers five datasets totaling 15,494 unique records, including held-out parsing checks and 1,344,000 exported business-value comparisons. Client workflow I first inspect a representative sample and agree on the column mapping, regional conventions, validation rules, and acceptance criteria. I preserve originals, review flagged records, verify the exports, and deliver a reconciled workbook with documented assumptions. Ambiguous identities or business decisions are never resolved silently. This is an independent synthetic portfolio demonstration, not paid client work.