refynr.
Clean spreadsheets you can actually trust.
refynr catches what's quietly wrong — duplicates, mixed dates, placeholder blanks, the damage Excel does to IDs and codes — and proposes a fix for each, reason spelled out. Review, export a clean copy with an audit report, and save it as a recipe for next time.
How it works
From messy paste to clean export in under a minute
Paste or upload
Drop in CSV, Excel, JSON or Parquet — or paste straight from Excel or Google Sheets. Semicolon and pipe files just work. It's analysed on your device — no upload, no waiting.
See what's wrong
Get a data-health score and an itemised list of findings, each explained in plain English. Anything refynr can't fix safely — an impossible date, a suspect ID — is flagged for you, never guessed.
Review, export, repeat
Accept fixes one by one, edit cells inline, reshape columns — every step undoable. Export a clean copy plus an audit report, and save your decisions as a recipe to replay on next month's file.
What refynr catches
Deterministic fixes you review, one by one
Non-destructive by design
Every change is a reviewable patch with a before, after and reason. The engine never edits your input — it only ever builds a clean copy.
Private by default
Cleansing runs 100% in your browser. Your rows never leave your device — there's no upload, and nothing is sent to any server.
UK formats, done right
Postcodes, phone numbers, VAT numbers (checksum included), sort codes and Companies House numbers — the details US-centric tools get wrong on British data.
The everyday mess
Stray whitespace, inconsistent casing, duplicate rows, placeholder blanks like “NA” and “NULL”, mojibake encoding, malformed emails and numbers-stored-as-text — normalised automatically.
Catches what Excel broke
Stripped leading zeros, IDs collapsed into scientific notation, codes silently turned into dates, rows with shifted columns — the classic export injuries, flagged before they reach your analysis.
Honest when it can't be sure
A column mixing UK and US date orders? Ambiguous cells are flagged, not guessed. Probable duplicates are shown for review, never auto-deleted. Confidence is earned, not asserted.
A score you can't fool
Deterministic across validity, consistency, completeness and uniqueness — and one cratered dimension can't hide behind three clean ones. Accepting fixes shows you the exact gain first.
Repeatable with recipes
Save your cleaning decisions as a recipe — pure configuration, none of your data — and replay it on next month's export in one click, or headlessly in CI with the refynr CLI.
Compare & prove it
Diff two versions of a dataset row by row to see exactly what changed since the last export, and ship the audit report (PDF, HTML or Markdown) alongside the clean file.
Why not just use AI in Excel?
The cell assistants rewrite your data. refynr shows its working.
In-cell AI (Copilot, GPT-in-Sheets)
- ✗ Sends your rows to a cloud model
- ✗ Edits cells in place — the original is gone
- ✗ You trust the change; you can't inspect the rule
- ✗ Every run can differ
refynr
- ✓ Runs locally — your rows never leave the tab
- ✓ Builds a clean copy; your original is untouched
- ✓ Every change is a patch with a reason you can read
- ✓ Deterministic — same input, same result, every time
Your data stays yours
Most cleaning tools upload your spreadsheet to a server. refynr doesn't. The entire analysis and repair happens locally, in this browser tab. There's no file to leak and nothing to delete afterwards — because your data never left in the first place.