Compare a master file with a reference file, identify missing or unused records, and update only the columns you approve. Receive a structured result with a separate change and quality report.

The reconciled master as an agreed CSV or JSON file.

A report of matched records, unmatched master rows, unused reference rows and the number of changed cells.

Independent checks of the complete key relation and every output cell, with SHA-256 hashes binding the checked files.

Master row order, non-target cells and unmatched rows preserved according to the agreed rules.

Exactly two supplied UTF-8 CSV files using one agreed dialect, or two flat JSON arrays of objects with string values. Each file may be up to 2 MiB; the combined input may contain up to 25,000 rows and each file up to 100 columns. Both SKU/key columns must contain unique, nonempty strings. Matching is exact and case sensitive: leading zeros and other identifier characters are preserved.

We agree which file is the master, which is the reference, the key columns, and the explicit source-to-target column mapping. Missing master matches are kept unchanged or cause rejection, as agreed. Unused reference rows are reported or cause rejection, as agreed. Duplicate or empty keys must be resolved before reconciliation. Formula-like content is not executed; it needs review and an explicit safe JSON path.

Send a small non-sensitive sample of both files, formats, row/column counts, the key columns and intended output. Confirm your right to process the data and that AI-assisted preparation and checking are allowed. We confirm scope, price, deadline and correction terms before starting. Contact for pricing. The displayed two-day duration is indicative after complete inputs and scope confirmation; the agreed deadline governs.

The linked CSV cleanup, duplicate-removal and JSON-mapping projects are my own synthetic demonstrations of file handling and verification, not paid client results or claims that those examples demonstrate this entire two-file SKU workflow. The SKU reconciliation operation has separate execution and independent-validation evidence.

No fuzzy matching, guessed identities, XLSX/formulas/macros, nested JSON, web scraping, enrichment, live database/CRM/API writes or recurring synchronization. This service prepares files; it does not modify a live shop. Do not send credentials, payment details or sensitive personal data. AI assistance is disclosed; independent QA is not a claim of human review.

Message me with the two-file sample and the result you need. I will confirm whether the task fits before quoting.