Story Highlight
– Study analyzes 1.3 million stock audits from retailers.
– 60% of inventory records in retail are inaccurate.
– New model identifies real errors 90% of the time.
– Machine learning catches 19% more errors than conventional methods.
– Smart counting approach proposed for better inventory management.
Full Story
A recent investigation involving 1.3 million stock audits at six prominent retailers has uncovered that most errors in inventory records can be anticipated and addressed early on.
The findings indicate that approximately 60% of stock records across the retail sector are inaccurate, leading to discrepancies between the recorded inventory and actual shelf stock.
Commissioned by ECR Retail Loss, the study introduces an innovative stock-taking methodology that successfully detects errors in 90% of cases.
Utilising machine learning technology, which relies solely on existing retail data, this new approach identifies about 19% more inaccuracies compared to traditional inventory counting methods.
Importantly, it is capable of identifying over 80% of instances where “phantom” stock—shelves that appear to be stocked according to the system’s records but are actually empty—could lead to losses from unfulfilled sales.
Traditionally, retailers have increased their counting frequency to combat these inaccuracies; however, this method is becoming increasingly time-consuming, cost-prohibitive, and complex due to expanding product ranges and staffing challenges.
The study’s authors, Professors Yacine Rekik, Aris Syntetos, and Christoph Glock, advocate for a shift from simply counting more to adopting smarter counting techniques. Their report, titled Smart Inventory Record Inaccuracy (IRI) Prediction and Management, elucidates how many stock discrepancies can be forecasted and provides a framework for addressing these issues effectively. By leveraging a transparent machine learning model based on pre-existing retailer data, this system can predict record inaccuracies and prioritise inventory checks, or autonomously correct them when there is certainty that the stock is unavailable.
