AI-model "Sberbank" predicts the demand for cosmetics by a third better than statistical models

2024: Creating a demand forecasting model

Sberbank Business Software has developed a model predicting demand based on artificial intelligence for one of the companies on the Russian FMCG Gradient market. Sberbank announced this on September 12, 2024.

Sales data was used to train the model (number of sold goods and their value), customers and warehouses, posted and planned promotional campaigns, assortment matrix and product balances.

Forecasting was carried out in 10 sections: at the level of sales of each of a specific product, by individual categories and the entire assortment, by months and weeks and so on. The greatest accuracy was shown by the forecast at the level categories. Compared to basic statistical methods, the ML model Sber Soft's business showed a 29 percent increase in forecasting. In the near future, Sberbank Business Software plans to transfer to the customer access to models for its application in industrial operation.

{{quote 'author = said Stanislav Kartashov, Vice President, Division Director "Corporate clients 360" of Sberbank. | Forecasting demand for goods is one of the most promising areas applications of artificial intelligence, because algorithms trained on huge arrays, data see more patterns than statistical models. We in Sberbank Group systematically use ML-models to optimize business processes and improve customer experience, and the competencies we have developed are widely in demand in the foreign market. So, the decision of Sberbank Business Software for the distributor "Gradient" showed high efficiency and ready for industrial implementation. Sure New the quality of the forecast will help the company improve business efficiency and increase profits,}}