Detect PII

Analyse > Detect PII... scans the table for columns that look like personal data, so you can find sensitive fields before sharing a file.

How it works

Octa weighs two clues for every column:

• the column header (does it look like email, first_name, gender, country, birthdate, ip, ...?), and
• the cell values (how many match a known shape: email, phone, IP address, credit card, IBAN, SSN, date, postal code).

That is why fields with no give-away values, like names, gender or country, are still found from their header, while a plain number column like salary is left alone. It only reads the data; nothing is changed.

Confidence

The percentage combines the two clues:

• a strong value pattern on its own reaches at least 60%;
• a matching header on its own reaches 60%;
• the two together score highest (up to 100%).

A column is listed when its best guess is at least 50%. The Basis column shows which clue drove it: column name, values (N%), or both.

Send to Anonymise

Send to Anonymise opens the Anonymise dialog pre-filled with one hashing rule per detected column, so you can mask the sensitive fields in a couple of clicks.

Command line and assistant

Also available as octa --detect-pii (see the --detect-pii reference) and the detect_pii MCP / assistant tool, which return the same confidence, by_name and value_match fields.