The GoodsForecast team implemented the project to develop custom models for the Russian leader in ferrous metallurgy and one of the world's largest steel producers in just 1 month in 2022. New demand forecasting models demonstrated a 5% increase in forecast quality.
In collaboration with the client's Data Science specialists, the GoodsForecast team developed a demand forecasting methodology that predicts sales more accurately than the company's existing algorithms. Historical sales data and macroeconomic factors were used to train the models. During the work, various models were tested: from long-established time series models to modern neural network models. Based on the results of the experiments, a technique was chosen that showed the best quality based on the customer’s data.
The developed methodology was transferred to the customer for further testing of forecast accuracy and comparison with the performance of current algorithms. Based on the testing results, the customer decided to continue cooperation in the field of digital business transformation.