Edge AI in retail: why in-store processing matters
Edge AI analyses camera images on a device in the store. Why that matters for privacy, speed and offline resilience, its trade-offs, what to ask.
What happens to the screens and the numbers when a store loses its internet? We unplugged the store device's cable during a test session and checked both, from the device's own log and the cloud records.

For 7 minutes 43 seconds without internet, every product shown reached the screen, and every record of it reached the cloud once the cable was back, with the time it really happened. Three showings, three on screen, three recorded. Only the sync to the cloud waited.
4 September 2026, GoTrack's test setup. Method below.
| Showing | On screen (Türkiye time) | Record reached the cloud | Waited on the device |
|---|---|---|---|
| 1 | 13:15:52 | 13:24:41 | 8 min 50 s |
| 2 | 13:16:13 | 13:25:07 | 8 min 54 s |
| 3 | 13:18:54 | 13:23:26 | 4 min 32 s |
| Online showings the same morning | 1–4 s |
The internet was out from 13:15:39 to 13:23:22: the store device logged the moment it reached the cloud again, 463 seconds after losing it.
Recognition runs on the store device, and the product content and the decisions about what to show are kept on it in advance, so a showing never waits for the internet.
Records are written on the device first and sent when the connection returns, carrying the time they happened. Reports therefore show the outage as ordinary activity, not as a gap followed by a burst.
Recognition, screens and event recording carry on in the store. Screens play their stored content, and events wait on the device and are sent with their real times when the connection returns. In our test on 4 September 2026, every showing during a 7-minute outage reached the screen and its record reached the cloud afterwards.
In the store. Images are analysed on a device in the store and are not sent to the cloud for analysis. Only product events and device health reach the cloud, which is also why recognition and screens keep working without internet.
In our test measurement on 30 September 2026, the screen acknowledged painting the product within 0.71 s of GoTrack deciding which product was in hand, in all 24 acknowledged fires: a median of 0.35 s with one product and 0.51 s with two. One of the 25 fires sent no acknowledgement. The time from the lift itself to that decision is not included and not yet measured separately. Method: the reaction time note.
Tell us about your connections; GoTrack keeps working without them.
or write to info@gotonom.com