In 2019, ground breaking research from ECR, identified that 60% of inventory records were wrong, and that when records went from wrong to right, sales grew by 4-8%.

In that report, the academics outlined three broad approaches to improve inventory record inaccuracy, the first could be technology based, such as RFID, the second could be to increase accuracy and recover lost sales by counting more and finally, the third approach could be a systems approach, such as machine learning that could detect and correct wrong records.

It was this third approach that was the focus of this session, and the academic team outlined an approach, sharing some initial results using data from the first phase to demonstrate with some good statistical confidence that machine learning does have a strong potential role to play in detecting wrong inventory records. In the breakout rooms the retailers discussed the approach, the relevance to their organisation of such a capability and the additional factors that could be considered in the analysis, such as returns or participation rates of self-checkouts. The next step of the research will be to recruit 4-5 retailers who would be prepared to share data for the development of the machine learning algorithms.

This session was one of the regular working group meetings over the year, for retailers, producers and academics only. If you would like to participate in a future session, click here. If you would like to get a recording of this session, please apply.