One duty officer. Three blocks. ASB patterns surfaced in the first incident.

Vendor type

UK BTR operator

Sites covered

3 blocks

Residences

~480 across the portfolio

Camera estate

~38 cameras (lobbies + bin stores + back exits + lifts)

Duty model

1 overnight officer covering 3 sites

ICO posture

ZA-class controller; ZB-class DPIA on file

Outcome headline (template)

Repeat ASB patterns surfaced inside the first incident, not after the third.

Embedding-only ReID across the lobby + bin-store + back-exit cameras

Each detected person is reduced to a 512-dimensional embedding on the edge node at lhr1. The vector is stored in the UK embedding store with a 14-day expiry. No face image is retained beyond the 30-second on-edge buffer. Cross-camera matches across the three blocks surface in the duty officer's pane as a thumbnail strip — same vector, different cameras, different times.

ASB pattern detection — same vector, multiple incidents

When the same embedding vector reappears across multiple incident-tagged windows (door-forcing, bin-store loitering, late-night corridor entry), the system flags the pattern. The duty officer sees a single 'this vector has now been flagged three times this week across all three blocks' row — not a haystack of footage to scrub.

ZB-class DPIA stamped by the operator's DPO

We arrive with the DPIA scaffold pre-filled against the operator's existing CCTV DPIA template. The operator's DPO stamps the ZB-class control-set (legitimate-interest balancing test, 14-day vector expiry, no facial-recognition store). The residents' association gets a one-page summary in plain English.

Resident-facing posture: no face image, no name, no identification

The lobby signage names what the system does ('cross-camera matching by clothing and gait') and what it does not do ('not facial recognition; no face images stored'). The system is operable by a duty officer with no special training — the only new affordance is the 'view cross-camera trace' button on an incident.

The named customer + the DPO stamp

We will publish the operator's name (with their consent), the DPO's name and title, and a redacted-PDF copy of the stamped DPIA.

The measured outcome — incidents surfaced, time-to-pattern

We will publish the measured 'time to first pattern surfaced' on the camera estate (target: <72 hours from go-live). We will publish the count of incidents the system surfaced in the first 90 days and the count the duty team would have caught without us.

The residents' association sign-off

We will publish the residents'-association meeting minutes (redacted) in which the system was discussed and the vote was carried.

The honest miss-list

We will publish what the system missed in the first 90 days that the duty team did catch by eye. We will not publish a success-only template — that posture is what makes a real case study trustworthy.