Optimizing demand forecasts in SAP using AI

Did you like our article?

We are currently undertaking several initiatives with clients in the retail and mass consumption industry to develop solutions based on Artificial Intelligence (AI) that integrate with SAP, to improve the quality of their demand forecasts.

With these clients we have found that their current predictive models fail to adequately capture the effect of irregular peaks on demand (such as holidays, Mother’s Day, cyber Monday, or sporting events) that may vary their impact from one year to another. This forces them to apply manual adjustments to their forecasts, which depend on the intuition or experience of the planner or category manager.

There are certain product categories that have a very predictable seasonal demand, and others with high volatility associated to different events. For example, an important soccer match notoriously affects the sales of TVs and products associated with barbecues (beer, coal, grills, meat, etc.), while the demand for other products such as sugar and toiletries is not significantly altered.

With AI it is possible to properly quantify the impact of such events in the projected demand, generating more precise forecasts, and eliminating the dependance on the discretion of the expert’s “intuition”.

Basically what we do is build a demand planning model that extracts data from SAP, complements it with relevant information from other sources, uploads it in an AI engine, processes the forecasts, and then feeds them where the client’s business process requires it. This can be in Sales Operations Planning (SOP), to plan the production, or in other cases, in the Material Requirements Planning (MRP), to plan purchases and replenishment, etc.

Also, we build Business Intelligence (BI) tools that enable the business analysts to examine this information and make decisions. These tools allow the analysis of variables such as the forecasts’ accuracy depending on different factors such as category or customer segment, so as to better identify where the models may be potentially improved, helping with the continuous improvement of the planning process.

Finally, being based on AI, the forecasts will get better over time, in contrast to traditional models, where forecasts accuracy remain similar over time.

For more information about our services, we invite the SAP community to contact us.

Author Patricio Renner, Technology Manager.