In Biedronka stores, refrigeration unit temperatures were manually monitored by employees following established procedures. Readings had to be taken at regular intervals — for example every 2 hours — and then logged in the system. This process was time-consuming, error-prone, and offered no guarantee that the full schedule was being followed.
‍
Refrigerator and freezer failures were sometimes detected with a delay, leading to thawed goods, financial losses, and food safety risks. An additional challenge was the lack of centralized monitoring of reading completeness and no mechanism for automatically responding to out-of-range temperatures.

Escola designed and deployed a mobile app using OCR technology and Machine Learning to automatically read temperatures from the physical displays of thermometers on refrigeration units.
‍
Instead of manually entering data, a store employee scans the displayed temperature and the device code using their phone's camera. The system automatically recognizes the data and sends it to a central database. If a device is missed, the app sends a PUSH notification flagging the missing reading. If a detected temperature exceeds acceptable thresholds, emergency procedures are automatically triggered, enabling a rapid response and minimizing losses.
‍
The biggest design challenge was adapting the OCR algorithm to dozens of thermometer models and challenging in-store conditions — some displays were scratched, partially fogged, or subject to variable lighting. The Machine Learning model was trained on dozens of device variants to deliver accuracy comparable to manual reading. The solution runs on Android and was also tested on older mobile devices. The rollout was preceded by FRD, BRD, and TRD analytical workshops, in-store testing, and a phased launch.