GoodsForecast solution drives additional profit growth for Verny retail chain

The federal retailer Verny summarized the results of its new machine learning-based forecasting model developed by GoodsForecast.

Moscow, December 28, 2022. New models for predicting promotion campaigns’ effectiveness have been developed as part of the roadmap for the development of promotions management processes in the Verny chain.

The Verny retail chain has been using the GoodsForecast.Promo planning system since 2018. Thanks to machine learning technologies, GoodsForecast's solution independently analyzes key parameters of previous promotions and predicts the results of future promotions based on them.

GoodsForecast and Verny cooperate based on success fee arrangements, when the vendor’s fee is paid only if the customer benefits from favorable results, i.e., makes additional profit. In late 2021, GoodsForecast and Verny significantly revised their forecast accuracy criteria: they must now correlate with the commercial results of using the promotion forecast. As a result, a complex integral indicator consisting of a number of formulas was created. It was developed based on the proximity of the functional KPI to the financial result.

In early 2022, the figure was approved. However, the February events put the project on pause. A period of turbulence began: demand, its nature, ability to pay and prices began to change dramatically. It was impractical to check the accuracy of forecasts based on the abnormal period. However, during March and April 2022, the GoodsForecast models performed consistently well with forecast accuracy not significantly deviating from the reality.

It was decided to revert to improving forecasting accuracy without changing the target. As a result, the GoodsForecast team developed a brand-new machine learning-based forecasting model.

Now forecasts are calculated simultaneously using two models: the new one and the previous one. The results of both are then combined. This approach resulted in a 21% increase in the integral score.

“It's a unique project. The proper evaluation of forecast accuracy guaranteed the best financial effect for both the Customer and the GoodsForecast team. Although the project was implemented during a period when demand was exposed to high turbulence, which affected the results, they were nevertheless quite impressive. We are happy to state that by using the new forecasting model, Verny has improved the accuracy of forecasting the effect of promotion campaigns and, more importantly, generated additional profit,” Lev Kiselev, Head of Retail Analytics at GoodsForecast, commented.

The next phase of work to improve forecast accuracy is expected to take place in 2023. Verny is now finalizing the raw data needed to calculate forecasts. In particular, a deeper detailing of the parameters on the promotions that have already taken place is under way.