How this report is validated
The evidence behind every number, and the lines we will not cross.
What we measure
Behaviour, not opinion. We capture where shoppers look using webcam eye tracking, with no headset and accuracy of about 7 degrees. We also record what they pick up, what they put back, what they buy, and how long each step takes.
How we validated it
Virtual-shelf behaviour tracks real-world shopper behaviour at r = 0.91 across 800+ consumer tests. The protocols were validated in paid studies with Nestle, Unilever, pladis, Pescanova, AB Vassilopoulos and Konva, under our predecessor Digital Bites.
When we tell you not to decide
Below 18 completed sessions per stream, every report is stamped NOT DECISION-GRADE. That stamp appears on screen and on every export, so a thin sample is never mistaken for a decision.
Real shoppers and AI shoppers, never blended
Real shoppers and AI shoppers are reported as separate streams, never merged into one number. AI cohorts are directional, and are calibrated against human sessions only once enough paired data exists.
Privacy and security
GDPR export and deletion are available on request. Consent is collected before every session, and an independent security review returned zero findings. Data retention is set per study.
What we do not claim
We do not predict sales volume. We do not fuse human and AI numbers into one figure, and we do not claim eye-tracking precision beyond what a webcam supports.