CSV Exception Report — Synthetic Demonstration

SYNTHETIC DEMONSTRATION — not client work.

This AI-assisted Python workflow compares two CSV exports using one exact key. It preserves every source record, keeps ambiguous duplicate keys out of automatic matches, and separates field differences from invalid or missing values.

The invented inventory fixture contains 9 left records and 7 right records. The result is 4 unique matched pairs and 8 unpaired records, with 2 field differences and 2 invalid values. The source-record count reconciles: 4 pairs account for 8 records, plus 8 unpaired records equals 16. Findings may overlap.

Deliverables include an HTML report, review CSV, raw-value JSON evidence, column mapping, reproducible command and file hashes. This page shows the one-page summary. The sample package includes the working code and test evidence.

Validation: the two-workflow suite passed 36 tests, including a seeded randomized conservation test. A separate synthetic benchmark compared 10,000 rows per input and found exactly 100 deliberately introduced changes. These are test results, not measured customer savings or production capacity.

Limits: exact matching only; no live integrations, fuzzy identity guesses or bookkeeping. The tool flags records for a person to decide.

For the $150 pilot, send the two column-header lists, the matching key and what needs comparing. Scope, permitted sample, data handling and delivery date are agreed before acceptance.