The solution

Build an AI component capable of computing probabilities beyond human perception…

Over a few weeks of trial and error, testing different ML software and a myriad of variables we managed to come up with a new algorithm incorporated a multitude of parameters, including day, week of the month, month, payday weekend, bank holiday, promotion, and weather. Furthermore, it employed time-series forecasting, allowing it to consider the previous month's trend when forecasting for the current period. This enabled us to use all the Historical data to train the model, and not be hindered by Sales patterns for Items and Stores changing year upon year.

We eventually settled on a software called ‘A Bot Named SUE’ (ABNS), which is perfect for the implementation we chose, and perfect for the challenge facing our client. This software is unlike most other ML software’s as it is suited and optimised for small data. This allowed us to start predicting with as little as 14 data points, so 2 weeks after a new Item hit the shelves, we could start producing forecasts for it. It was also extremely quick and lightweight, which would enable us to be able to produce the number of forecasts within the timeframe available each day.

A distinctive feature of the solution was its full configurability with a user-friendly front-end application that enabled real-time analytics, error logging, model health/diagnostics, and full configurability. The process, while being completely automated, allowed granular adjustments such as the length of the forecast period, the length of the Trend, the days that go into training the model (extreme periods can be removed that may prove to be unreliable predictions for the future such as certain times during COVID). Providing the Retail company with maximum control and flexibility.

Implementation

The algorithm was applied across the range of 500 items in 1000 stores.

Each Item in each Store being individually modelled as every single combination has a unique Sales pattern, which, if not modelled individually would sacrifice accuracy. Despite the vast volume of data, the automated system could produce 7 million forecasts in a mere 2 hours, a remarkable feat that underscored the power and efficiency of the new solution.

At every turn, the solution was built with best practice, and data optimisation in mind. The sheer number of forecasts and data produced meant that it was critical that there was no redundant logic or data.

The whole process is fully automated and requires no manual intervention (though is configurable if necessary).

The results

The implementation of the new forecasting algorithm by ABNS led to significant improvements in the group's forecasting system. The enhanced accuracy in demand prediction led to:

• Optimised inventory management, leading to a decrease in costs associated with overstocking and understocking.

• Increased customer satisfaction due to better availability of products.

• Greater operational efficiency with the automation and configurability of the forecasting process.

• Improved decision-making ability due to the incorporation of real-time analytics in the front-end application.