DataTray, and the file that would not open

There is a specific moment DataTray was built for. The export from the CRM is 900 MB. The log dump has more rows than a spreadsheet can hold. The vendor sent .parquet and Windows has no idea what that is. The CSV opens as one column because it is semicolon-delimited and your machine expects commas.

Today the usual options are: upload the file to a website you have never heard of, install Python, buy a BI licence, or give up. DataTray is the fourth option.

What is in the first build

Five tools, organised by what each one does to your file.

• Data Explorer — opens anything and changes nothing. Rows, a full column profile, a data-quality score with its weights printed, a data dictionary you can hand to somebody, and a SQL tab if you want one.
• Data Converter — CSV, TSV, Excel, Parquet, JSON, NDJSON and XML in the directions you actually need. Twelve files into one, or one into twelve.
• Dedupe & Merge — exact, normalized or fuzzy matching, with a report naming every row that was removed and what it lost to.
• Clean & Anonymize — the twenty things wrong with every export, fixed in one previewable pass, and the personal data found and masked in the same pass.
• Compare Files — schema drift first, then added, removed and changed rows, with tolerances so float noise is not reported as a change.

Recipes, so the twelve fixes you applied to this month's export can be re-run on next month's.

Three things that are true of every one of them

Nothing uploads. Not one of the five tools makes a network request. The query engine runs with remote filesystems switched off, so even SQL you paste in cannot reach http:// or s3://.

Nothing downloads after install. No AI models, no database engine, no extensions fetched the first time you open a spreadsheet. DataTray works on a machine with no internet connection at all.

Your original is never modified. Every run writes a new file with a dated, self-describing name — and anything dropped or rejected along the way is written down beside it, because losing rows silently is the failure that makes people distrust cleaned data.

What is not here

No speed claim. DataTray opens files far past what a spreadsheet will, but we have not published a number for it, because nobody has measured one on a named reference machine yet. When that happens the measured figure goes up and not before.

And it is not a spreadsheet — no formulas, no charts, no pivot tables. It operates on files.

The rest of the caveats are on the home page, stated up front rather than discovered later.