Seamless stock anticipation: revolutionising retail with data-driven forecasting

Challenge

The customer had a tool to replenish items in their fashion retail shops. This tool worked for normal periods, but for special events, such as Christmas or Black Friday, it was unable to anticipate, due to the sales and high volume of transactions. This meant that one person had to be there during these periods adjusting the clothing replenishment, making it a manual and inaccurate process.
The customer developed a forecast algorithm but, due to their infrastructure, they were unable to process all historical data, so the results were not accurate enough.

Approach

Migrate the customer solution to a new one to be able to process large amounts of data. With this new solution, the customer could execute the algorithm with all the historical data and forecast the stock for special events.

Result

With the new solution, the data was divided into years, allowing the customer to aggregate the data according to their needs. Thus, the forecasting algorithm could be run with all the necessary historical data, but only the current data was used for normal processing. In this way, the client has now a competitive and scalable forecast solution, keeping the performance of the normal replenishment process intact.