TAMREVS
It never asks who you are.Only whether you're the same person who walked in.
Exit verification that reduces theft while deleting every observation the moment a customer leaves.
- ~3s
- entry pause
- 0
- identities stored
- 3
- exit outcomes
The problem
Retail shrinkage is a multi-billion-pound epidemic. Existing tools are stuck in trade-offs.
Every current approach sacrifices something fundamental: accuracy, privacy, throughput, or cost.
Electronic article surveillance
Requires every product to be tagged. Misses untagged items and bulk theft.
CCTV
Mostly retrospective. It records the crime rather than preventing the exit.
Security guards
Effective but expensive, inconsistent, and hard to scale across thousands of stores.
Receipt checks
Inconvenience every legitimate customer and still rely on human judgement.
Facial recognition
Permanent biometric databases create privacy, regulatory, and public-trust risks.
Single-signal AI
One camera or one sensor alone generates too many false positives to be operationally useful.
The journey
Three beats. No friction.
Entry, shopping, exit. The system only does what it needs to do, only when it needs to do it.
Entry
The customer pauses for about three seconds. Floor sensors, cameras, and optional depth sensors create a temporary encrypted profile — no name, no card, no phone.
Shopping
The customer shops normally. If they buy something, the POS links the receipt to the temporary session ID.
Exit
Exit sensors repeat the measurements. AI compares the profile, reconciles purchases, and produces a confidence score: green, amber, or red.
Privacy by design
What TAMREVS never collects
The system is deliberately built to need as little as possible.
What exists, briefly
Temporary encrypted observations
All data is tied to a single-session ID. It is encrypted, fused into a confidence score, and securely deleted after exit.
Decision model
Green, amber, red.
Every exit produces a confidence score and a clear operational recommendation. No black-box rejections.
Confidence is above the operational threshold. The gate opens and the temporary session is securely deleted.
Confidence is borderline, or a sensor is degraded. The exit pauses and a staff member is alerted for a quick, logged override.
A significant mismatch beyond tolerance. The session is flagged, staff are notified, and the event is recorded for learning.
The dossier
Four pillars, one method.
The full reference architecture spans thirty-five chapters. These are the four routes through it.
Verification
The session lifecycle, decision pipeline and confidence logic that turns observations into an explainable recommendation.
Governance
Privacy by design, human oversight, fairness testing and the engineering controls that keep the platform accountable.
Evidence
Prototype objectives, pilot phases, governance gates and the measures that decide whether the concept progresses.
Honesty
Assumptions, known limitations and open questions, stated plainly. Nothing here has been validated yet.
Ready to review the full dossier?
The technical architecture, privacy model, deployment map, and IP strategy are available for qualified investors and retail partners under confidentiality.