Sensing plane · UK
Watch the perimeter. Without watching your residents.
Re-identify people across cameras using embedding similarity, not facial recognition. Sensing-plane sibling to TicketWave's decision plane. Built for UK build-to-rent operators, boutique hotels and mid-market venues — the cohort where the duty officer already knows the regulars and the security team is two people.
§ 01 · WHO IT'S FOR
Eight buyers. One wedge.
We start with UK build-to-rent because the concierge-desk economics are the worst in the sector. Boutique hotels and mid-market venues come with the same posture. Three further verticals — universities, local-authority CCTV, transport hubs — share the same ICO + PSED + JCNSS reasoning. Plus two more on the same posture: events and festivals, and purpose-built student accommodation. Not nightclubs. Not stadia. Not Heathrow.
- 01
Build-to-rent
BTR operators
L&Q, Get Living, Quintain Living and the rest of the UK BTR cohort all run 24/7 concierge desks watching screens. Embedding re-identification lets one operator cover three buildings without reviewing footage of residents going about their lives.
- 02
Boutique hotels
Hotels under 200 keys
Independent and small-group hotels in London, Edinburgh and Manchester. Track a guest's bag from lobby to lift to landing without identifying them, in a posture the ICO will read as embedding-only rather than biometric.
- 03
Mid-market venues
1,000–5,000 capacity venues
Smaller theatres, music venues and conference centres where the door staff already know the regulars and the security team is two people. We do the watching; they do the deciding.
§ + THREE MORE · SAME POSTURE
Three further buyers on the same ICO posture.
- 04
Universities
Open campuses + halls
Russell Group estates: public-access library footprint glued to private-residential halls. Lab-door tail-gating + Athena Swan + PSED, all without categorising students.
- 05
Local authorities
Council CCTV control rooms
Estate ASB, town-centre night-economy review, CCTV-operator burnout. ZA-class posture preserved; no ZB-class biometric step required.
- 06
Transport hubs
Stations · airports · depots · ports
JCNSS reasonableness, lost-property reunification, anti-stalking pattern detection — surfaced inside the first incident, not the third.
§ + TWO MORE · SAME POSTURE
Two further buyers on the same ICO posture.
- 07
Events and festivals
Outdoor and temporary environments
Gate ingress at peak, anti-tout patterns, vendor lanyard validation, after-party welfare drift. Consent signage at every ingress; event-window retention, narrower than fixed-site verticals.
- 08
Student accommodation
PBSA — purpose-built halls
Unite / Yugo / iQ-shaped buildings. Communal entrances, vandalism patterns, NUS-aligned posture and Higher Education Code of Governance proportionality, in a duty model closer to BTR than to universities.
§ 02 · WHAT IT DOES
ReID via embeddings is not a facial-recognition workaround.
The shortcut every CCTV-AI vendor takes is to call their face store an "embedding store" and hope nobody reads the schema. We don't have a face store. We compute an embedding per detection, match against the same camera's recent history, then drop the raw image inside a 30-second buffer. The vector itself is not a face — it's a hash of body shape, gait and clothing colour that becomes statistically meaningless after about 14 days.
That distinction matters because the Information Commissioner's Office has been explicit on facial recognition in private space (the Serco custody contract; the South Wales Police case; Facewatch). We're built to answer the auditor's first question — "where is the face store?" — with "there is no face store."
-
§01
Embeddings, not identities
Cameras stream to a UK-hosted edge node. We compute a 512-dimensional embedding per detected person and store only that vector. No face image, no name, no biometric template under the UK GDPR Schedule 1 special-category meaning.
-
§02
Match across cameras
When the same vector reappears on a different camera within a tunable window, we surface that as a re-identification event. Operators see a thumbnail strip; residents and guests see nothing.
-
§03
ICO-clean by construction
The Information Commissioner's Office has been explicit that automated facial recognition needs an Article 9 lawful basis. We don't do that. ReID via embeddings is a different category of processing, and we have written legal opinion to back the distinction.
-
§04
DSPT v8 ready
The Data Security and Protection Toolkit v8 audit becomes mandatory on 30 June 2026 for relevant care and adult-care-adjacent settings. Our control mappings are built for the v8 controls; we share them with you so your auditor sees a finished spreadsheet, not a half-built one.
§ 03 · WHERE IT RUNS
UK-hosted. By construction.
Edge inference at the LHR region. Embedding store in the UK region. No US transit. The full topology is on the security page.
§ NEXT
See it on one camera, in one building.
We don't run paid pilots and we don't run paper pilots. If you're a UK BTR operator, a boutique-hotel security lead, run a 1,000–5,000-capacity venue, sit in university Estates, run a council CCTV control room, a transport-hub security ops desk, an outdoor-event security operation or a PBSA building manager, the next step is a 30-minute call to see whether ReID-via-embeddings fits the camera estate you already have. Until a paid pilot lands, the case-studies hub publishes the templates we expect to swap in for the real ones — explicit-template framing, named verticals, honest commitment list.
Schedule a 20-min call → See the demo → Read the templates → Read what we believe →