Record Reconciliation
A line comparison can exaggerate changes when an exported table is sorted differently. This workbench identifies the same record using your unique business keys, then compares fields by column name. Separate reports show added, deleted, modified and unchanged records, plus the before and after values of each changed field.
Key features
- Identify records with 1–4 compound keys regardless of row order
- Reject empty or duplicate keys with CSV record locations
- Record-level status and field-level change tables
- Distinguish added/removed columns, absent values and empty strings
- Absolute and relative tolerance for selected numeric fields
- Preserve source strings and download complete CSV details
How to use
- Paste the baseline and comparison CSV or select UTF-8 files.
- Choose delimiters and read both tables.
- Select 1–4 shared column names that uniquely identify each record.
- Optionally choose numeric fields and tolerances, then reconcile.
- Review record statuses and field changes, then export the needed detail table.
Use cases
- Detect additions and removals in reordered weekly product lists
- Track changed order records identified by compound keys
- Separate measurement differences that fall within a chosen tolerance
- Find fields added to or removed from an export schema
Frequently asked questions
How does this differ from text or JSON diff?
Records match by unique business keys instead of line or array position. Column order is ignored, field names are matched, and snapshot additions, removals and optional numeric tolerances have separate meaning.
Which row wins when keys are duplicated?
Neither. The tool stops and reports the duplicate CSV record. Use CSV Table Join for one-to-many relationships. Reconciliation requires one identifiable record per key, possibly made from several columns.
Are empty cells the same as absent fields?
An empty string means the column exists but contains no text. An absent value means its row or column is missing from that snapshot. The detail CSV retains before_present and after_present flags.
Can tolerances change how keys match?
No. Keys remain strings. Tolerances apply only to explicitly selected shared non-key fields. Relative tolerance 0.01 means 1%; cells that are not supported numbers fall back to exact string comparison.
Which numbers and CSV files are supported?
UTF-8 comma, semicolon and tab files are accepted up to 1 MiB, 5,000 rows and 32 columns each. Numeric comparison uses finite decimal/exponent values within the stated safe magnitude and JavaScript floating-point arithmetic. It is not an arbitrary-precision accounting calculator; specify an appropriate tolerance for rounding differences.
Does CSV export include only filtered preview rows?
No. It exports the complete selected record or field-change table. Filters and 50-row pages affect the preview only. Potential spreadsheet formula strings receive a leading apostrophe in downloaded CSV, with a protected-cell count.
Privacy
Pasted CSV, file contents, settings and results are processed in this page’s browser memory. This tool does not automatically send inputs to a server, URL, analytics event or browser storage. Files are created only when you request a download. Input changes, Clear all and leaving the page discard the result.
Comments & questions