The Estima system is used by the Azymut publishing house to help in the effective forecasting of product sales. The system uses machine learning techniques to analyse past sales data, and forecasts future sales by means of sophisticated algorithms. Properly selected training and test sets ensure very high performance, enabling shortening of the entire publishing cycle, improved product rotation and reduced storage costs, while at the same time maximising sales. All this translates directly into growth in profits.
The company had sales data for hundreds of items, accumulated since 2014, in various (database) formats, divided into internal categories (group, subject, publisher, etc.). Standard human analysis (based on the previous year’s result, averages over a number of years, median values, linear regression, etc.) produced inconsistent and unsatisfactory results. The challenge was to forecast the quantity of orders (stocks) so as to enable the sale of as many copies as possible, particularly in the seasons when demand is highest. Unable to find a suitable ready-made solution, the company commissioned us to develop the Estima system.
We constructed predictive models to forecast quantities of sales based on past data. Properly selected training and test sets, in combination with high-performance technology, make it possible to obtain very good predictive performance. The system offers a user-friendly interface that enables the choice of multiple options for sales forecasting. The results are supported by attractive visualisations of the analysed data, shown in many different views. Thanks to the design of a suitable API, the Estima platform was integrated with the existing workflow. This avoided additional costs and the need for special staff training.
We can grow your profits in a similar way. Let’s talk about it!