We employed our state-of-the-art Deep Learning based Lummetry LENS+ solution to predict sales at Store-SKU level and offer preventive out-of-stock (OOS) alerts taking into account the lead time.

Context

Studies say that retailers face losses up to $1 trillion due to out-of-stock issue. Current statistical based solutions fail to deliver up to the expectations and state-of-the-art research papers prove that Artificial Intelligence and Deep Learning in particular already have a great positive impact on improving the sales forecasting and OOS predictive alerts.

Reckitt Benckiser is a world-wide leader in the home and hygiene sector with a huge base of retail sellers (large and small) and a significant number of SKUs. Thus, they are negatively impacted by the OOS too, consumers choosing a product from the competition in case they do not find the product from RB that they were looking for.

Challenge

Improve the current accuracy of sales forecasting at the Store-SKU level and correlate the predictions with the initial stock and planned replenishment in order to provide better OOS preventive alerts to the merchandisers and distributors.

Solution

Lummetry LENS+ has complex Deep Learning models capable of analyzing all time-series based buying patterns of each SKU at a store level. Once the predictions are validated with the business team, by either the backtesting method or deliver predictions for the next X days and wait for the reality to happen in order to calculate the prediction error, we then add information about initial stock level at a day level and information about the planned replenishment for each SKU for the next X days. Next, Lummetry LENS+ processes all the info and triggers preventive OOS alerts according to the lead time for each SKU.

Results

Our solution managed to achieve between 10% – 50% better accuracy than internal benchmark, depending on the SKUs predictability and data availability for each timeseries.