Demand forecasting for retailers: the benefits of AI

This article is brought to you by Retail Technology Review: Demand forecasting for retailers: the benefits of AI.

By Arnaud Gauthier, President & Chief Customer Officer, EMEA at Symphony RetailAI. 

The ability to quickly and accurately anticipate demand is essential for those who want to remain competitive in the retail industry, where even the slightest variation in production volume, distribution networks, or just the weather can have an immense impact on turnover. Increasingly, technology such as Artificial Intelligence (AI) is being adopted by the world’s major retailers to enable them to better aggregate and process huge data sets to predict trends or anomalies which will affect demand. 

Demand forecasting – the challenge

Demand forecasting conducted using traditional, manual methods is no mean feat. 

Firstly, there is the process of aggregating and evaluating data. For well-established products this is extremely time-consuming due to the sheer volume of data which must be captured and analysed. And for new-to-market products there is a complete lack of historical data of any kind. This means that forecasting demand for these items is based purely on guesswork rather than facts. 

The second issue facing retailers is data quality. Retailers obtain their product data from multiple sources, whether that’s directly from suppliers or extracted at the point of sale via various consumer touchpoints, digitally or in store. The difficulty is that the data arrives in different, non-compatible formats – including Excel spreadsheets – so retailers must spend time sorting and cleaning the data before it can be analysed. This involves manual intervention which creates opportunity for human error.  

Finally, external factors such as strikes and the weather ­– which according to the British Retail Consortium is the second biggest influence on consumer behaviour after the state of the economy – also create a very specific set of challenges for retailers. The sales performance of just about every consumer product can be affected by a particular type of weather condition and the unpredictability at which these events can occur, makes forecasting calculations difficult to make without the aid of technology. 

Retailers leading the way

AI is now being used to help analysts predict demand due to its ability to aggregate and automate the processing, cleaning and categorisation of large volumes of data quickly. This makes it easier for retailers to anticipate even the slightest shifts in demand which could cause issues for the retailer and their entire supply chain - from order quantities, supplier choices, fulfilment execution, logistics, truckload ramifications and pricing. Ultimately, this could result in unhappy customers.

Brands, including the major French retailer, Intermarché have launched pilot projects to test AI’s forecasting performance, with some encouraging results. For several months, Intermarché carried out a test based on three years of data history from its smallest warehouse, which stored frozen products, and its largest warehouse with nearly 30,000 stock keeping units (SKUs). 

The aim was to automate the forecasts to obtain results at least equal to those produced using conventional demand forecasting techniques. Results exceeded expectations, achieving a forecast reliability rate of 95 per cent – 15 per cent more accurate than traditional methods. Forecast quality gains were observed across all item rotations, warehouse sizes and product categories and AI was also able to identify products that responded to seasonality – something which had in the past gone unnoticed. 

AI – data driven benefits 

The use of AI in forecasting demand produces tangible benefits throughout the supply chain, starting with improved supplies that result in fewer product shortages, overstocks and less waste.  Its use also helps to improve planning, making it possible for retailers to really optimise their storage space and their capacity to receive shipments. 

As the example of Intermarché demonstrates, AI-enabled applications – built on machine learning – are speeding up the forecasting process for retailers, evaluating trillions of customer data combinations, far quicker than any human or traditional enterprise system ever could. The adoption of AI will continue to provide retailers with valuable predictive insight that can reduce pressure on manual processes and lead to improved accuracy in demand forecasting.