Turnover de-identification is the removal of personal identifiers from an attrition dataset. UK GDPR Recital 26 excludes data that no longer points to a person, and Art. 89(1) expects safeguards such as minimisation where records are reused for statistical purposes. anonym.plus marks names and IDs on your device, so the trend lines stay usable while the individuals behind each leaver are hidden.
When this applies
Attrition work starts as a chart and ends as a score. Reusing HR records for analytics engages purpose limitation under Art. 5(1)(b), and a flight-risk model that ranks named employees is profiling: Art. 22(1) restricts a decision based solely on automated processing that has legal or similarly significant effects, and Art. 35(1) can require an impact assessment before that evaluation begins. A dataset with no identifiable rows steps outside all of it.
How anonym.plus handles it
- Open the dataset in anonym.plus on your device.
- The tool flags named leavers and IDs.
- Local OCR reads a scanned chart export.
- Keep the leave-reason and date-band columns.
- Replace each identifier with a steady label.
- Save the clean dataset locally.
What you need to provide
- The attrition export (CSV, XLSX export, PDF).
- An operator (Replace keeps the columns aligned).
- Optional allow-list for date-band columns.
PII entity types detected
| Category | anonym.plus entity type | Example |
|---|---|---|
| Names | PERSON | leaver row → [LEAVER] |
| Identifiers | NATIONAL_ID | staff no. 51140 → [STAFF_ID] |
| NRP | NRP | job title → [TITLE] |
| Organisation | ORGANIZATION | department → [DEPT] |
| Dates | DATE_TIME | left 2024-Q3 → [BAND] |
| Contact | EMAIL_ADDRESS | work email → [EMAIL] |
Compliance achieved
- Works towards the anonymity bar in UK GDPR Recital 26, which the ICO tests with its motivated-intruder approach.
- Supports the safeguards Art. 89(1) expects where HR records are reused for statistical purposes.
- Serves purpose limitation under Art. 5(1)(b) when personnel data is reused for analytics.
- Keeps flight-risk scoring clear of Art. 22(1) and the Art. 35(1) impact-assessment duty by removing the identifiable rows profiling needs.
- Reduces exposure to the re-identification offence in DPA 2018 s.171.
Anonymise turnover analytics datasets offline — see plans & start free →
Limitations & cautions
Trends are aggregate, yet a rare title plus a leaving quarter can re-identify one leaver, so band those fields too. Removing names does not retrospectively cure a model already trained on identified staff — that is an Art. 35 question, not a redaction one.
Frequently asked questions
Can attrition trends still expose a person?
Yes, when a slice holds one leaver. Band dates and group rare titles so a pattern cannot single out a worker.
Does attrition scoring need an impact assessment?
If it systematically evaluates staff through profiling, Art. 35(1) is likely engaged, and Art. 22(1) restricts decisions taken solely on that basis. Working on an anonymised set avoids both questions.
Is the dataset uploaded?
No. The app is fully offline, so it stays on your machine.