Sales Forecast for a Store
The solution helps stores to forecast sales and prevent customers outflow, thus optimizing many business processes.Our team created a solution for retailers and online stores which makes forecasts for 5,000+ products. We used accumulated for three years data on daily sales, purchased items, average checks, loyalty card types, promotions, and so on to teach the application to identify seasonality in demand by month and day of the week, determine trends for each product group, and evaluate the impact of discounts and promotions on product sales. Based on this analysis, the program builds sales forecasts for a month, a quarter, and a year and identifies changes in demand for individual groups of customers. The solution increased the accuracy of forecasting for the next three months from 77% to 93%.
Another part of the program calculated the likelihood of customers outflow within 1-3 months. Our Python development team examined many parameters that impact customer activity: frequency and volume of purchases, number of site visits, date of the last purchase, viewed and selected products, positive or negative user experience. We identified groups that are "at-risk zone" and helped the client find out the proper offer, which will motivate customers to make purchases. For instance, we defined the most relevant goods based on the customer's previous purchases and figured out the right time for sending a message. It allowed the store to tailor the communication to each customer's needs and resulted in higher CTR (from 2.8% to 5.1%) and lower churn rate (from 9.4% to 5.6%).
Altogether, the system helped the business retain customers, boost sales through repeat purchases, and optimize business processes such as marketing communication, logistics, and storage of goods in the warehouse.
Technologies and tools: Python, ARIMA, ARCH, Recurrent neural networks, Prophet; Pandas, NumPy, Scikit-learn.
-thumb.jpg)
-thumb.jpg)
-thumb.jpg)
-(1)-thumb.jpg)
