Supply chain planning: using forecasts to optimise stock levels

by Julia Pott

Artificial intelligence is currently transforming many areas of the retail industry. One of AI's strengths: deriving predictions from big data. RELEX Solutions specialises in this data-driven forecasting for the supply chain. Ramona Saager explains what this means for the retail sector: using specific examples, she shows us how retail companies can use data-based predictions to automate their orders, save labour time and increase customer satisfaction at the same time.

At RELEX, you work on the supply chain. What is your goal here?

Ramona Saager: We want to optimise supply chain planning in order to achieve the perfect order quantity. Perfect means high availability with the lowest possible stock levels. Of course, customers should always receive what they are looking for. At the same time, however, inventory management is associated with costs. Write-offs should be avoided.

To know what needs to be ordered, you need to be able to look into the future.

Saager: This is exactly what we strive for. We analyse the sales volumes and stock data for each individual product in each shop. From this, we derive the quantity of each product that a retailer will need in the coming days: We create a forecast. Taking delivery times into account, we then determine every night what quantity of an item needs to be available in the shop, regional depot or central warehouse and when. This is used to calculate how much needs to be ordered from the manufacturer at what time so that everything has passed through the entire supply chain on time and is ready for sale in the stores.

However, demand is influenced by various factors. Do these also play a role?

Saager: Many different parameters can be taken into account, such as marketing promotions or seasonal events that lead to fluctuations in demand. A classic example is the weather, which plays a role for certain product ranges. Predicting how much ice cream will be bought on a Saturday is complex: just knowing that the sun will shine is not enough. How many hours of sunshine will there be? How warm will it be? How high will the humidity be? Will it be the first warm and sunny Saturday of the year or the twentieth? These are all data that play a role and can be taken into account.

Another example where external data is relevant is the travel retail sector. A retailer like Valora with shops at airports can use passenger data and visitor flows to better predict demand.

Data can come from the retail companies themselves or from external partners. Before we incorporate the data, we check whether it is relevant for the retailer and the product range.

Great progress has been made in data analysis thanks to the development of artificial intelligence. Do you also work with AI?

Saager: Yes, we use machine learning, our system is self-learning. AI is integrated into our core function, demand forecasting. These predictions can be used for process optimization in many other areas, for example for space and shelf planograms or for price and personnel management. AI is therefore essential, without it we would not be able to do what we do.

How does a retailer benefit from such a forecast?

Saager: The retailer can automate many processes and thus save working time. We have customers who order 95% of the daily order quantity without checking the forecast data again. The employees only look at relevant events, for example if alternative products need to be ordered in the event of delivery problems or an out-of-stock warning. This enables very convenient handling and noticeable cost savings. Employees can concentrate on strategic tasks such as developing the product range.

To what extent do the requirements for demand forecasting differ in different retail sectors?

Saager: We have customers from sectors such as drugstores, cosmetics and DIY, but also many from the food retail sector. In the food sector, there are some particularities such as seasonal demand for Easter bunnies or exercise books. As already mentioned, promotions, the weather or the days of the week can influence demand. Another banal example: at the beginning of the month people receive their salaries and are more generous, at the end of the month money is tight and people tend to save. Then there is the issue of freshness: perishable goods such as fruit, vegetables, fish, meat and dairy need to get through the supply chain as quickly as possible. The issue of food waste is receiving increasing attention.

With so many different factors influencing demand, an automated solution is important because a machine can process much more data than a human. This is where food retailers have a major advantage: artificial intelligence needs large amounts of data to learn. This means fast-moving items such as food provide a lot of data that can be used to make the forecast more accurate.

How can retailers react when the general conditions change?

Saager: We have a module, the Business Rules Engine, which allows retailers to create and customize their own business rules in the system. During the coronavirus pandemic, for example, there were major changes in demand for certain items such as masks or disinfectants. As a result, Rossmann shortened the demand analysis period for certain product areas and thus adjusted the forecast and the automated ordering rules.