Calibration data de-identification is the removal of names from rating grids before managers compare scores. A calibration session is where bias concentrates. The Equality Act 2010 reaches it from several sides: section 13 on direct discrimination, section 19 on a practice that disadvantages a group, section 18 on pregnancy and maternity, and section 15 on treatment arising from disability. anonym.plus masks each identifier on your device, so the session weighs the work alone.
When this applies
A calibration meeting lines up scores across a team. Two people were on maternity leave and one has a phased return. You blind the grid first, so leave and health cannot quietly drag a rating down.
How anonym.plus handles it
- Open the rating grid in anonym.plus on your device.
- Built-in OCR reads a scanned score sheet.
- The app marks names and any characteristic-linked terms.
- Confirm the markings and keep the score columns.
- Mask each name with a stable code.
- Save the blinded grid locally.
What you need to provide
- The rating grid (XLSX-as-TXT, PDF, DOCX).
- The Mask operator for stable, blind codes.
- Optional alias map, sealed and kept off-grid.
PII entity types detected
| Category | anonym.plus entity type | Example |
|---|---|---|
| Names | PERSON | Aisha Bello → STAFF_07 |
| Demographic | NRP | disability status → [REDACTED] |
| Dates | DATE_TIME | DOB 1989 → [AGE-BAND] |
| Location | LOCATION | home postcode → [AREA] |
| Contact | EMAIL_ADDRESS | a.bello@example.co.uk → [EMAIL] |
| Org | ORGANIZATION | Women in Tech network → [GROUP] |
Compliance achieved
- Supports bias-blind review against Equality Act 2010 s.13 (direct) and s.19 (indirect discrimination).
- Hides the cues behind Equality Act 2010 s.15 (treatment arising from disability) and s.18 (pregnancy and maternity).
- Follows the blind-assessment thinking of the EHRC Employment Statutory Code of Practice, which a tribunal must take into account.
- Keeps the evidence you need if s.136 shifts the burden of proof onto the employer, with AES-256-GCM at rest, offline.
Anonymise calibration data sets offline — see plans & start free →
Limitations & cautions
Blinding names does not remove bias from prior scores. If a rating already reflects a protected characteristic, masking alone will not cure it. Audit the inputs, not just the labels.
Frequently asked questions
How does blinding help under the Equality Act 2010?
It lets a panel weigh work without seeing who did it, which curbs direct discrimination under section 13 and the group disadvantage section 19 targets. It is one safeguard, not a full equality programme.
Why single out disability and maternity here?
Because both distort a raw score. Section 15 covers unfavourable treatment because of something arising from a disability, and section 18 covers unfavourable treatment because of pregnancy or maternity leave. Absence-driven ratings are a classic route into both.
Does it mark protected-characteristic terms?
Yes. It marks references to age, disability, and similar traits so you can mask them before calibration.
Where does the alias map live?
Only on your device, if you keep one. For a fully blind session, leave the map off entirely.