Find personal data in a spreadsheet DATASCAN

What it looks for

It reads a CSV or JSON file in your browser and scores each column by what its values look like, rather than by what the column is called. A column headed ref holding national insurance numbers is found; a column headed email holding order codes is not flagged just because of its name.

The shapes it recognises are the ones with structure: email addresses, phone numbers in common international formats, national identifiers including US Social Security numbers, payment card numbers, IBANs, IP addresses, dates that look like dates of birth, and postal addresses.

Card numbers are checked with the Luhn formula rather than by length, so a column of sixteen-digit order references is not reported as a column of card numbers. That is the difference between a check somebody can act on and a list of false alarms somebody switches off.

Where it is wrong, in both directions

• It misses anything without a shape. A column of names, of free-text notes, of job titles or of medical conditions is personal data and looks like ordinary words. This finds none of it, and the free-text note column is very often the most sensitive thing in an export.
• It misses data that becomes personal in combination. A postcode, a date of birth and a sex are each unremarkable and together identify a large share of a population. No pattern matcher sees that.
• It over-reports. A column of test data, of placeholder addresses, of IPs belonging to servers rather than people. Every finding is a column to look at, not a fact about the column.

What it is for

It is the check you run in the thirty seconds before you send an export to somebody outside: a contractor, a supplier, an analyst, a support ticket. The common accident is not a hidden database breach; it is a spreadsheet with three more columns than anybody meant to attach, because the export was easier than the query.

It is not a compliance tool and it will not tell you whether a transfer is lawful. If you need to hand over a file with the identifiers removed rather than merely found, ANON pseudonymises the columns consistently, and CLEANSHEET deals with the rows and sheets a spreadsheet hides rather than shows.