SAPinsider Vegas 2023 Session: Extending SAP IBP Capabilities with AI

⇨ When it comes to off-the-shelf solutions, SAP IBP still offers some of the most advanced features available among all off-the-shelf integrated planning solutions.

⇨ As technology evolves exponentially, customers are envisioning enhanced capabilities from their supply chain planning setup.

⇨ With open-source tools available to build advanced capabilities leveraging AI and ML on top of SAP IBP, customers are looking for new and innovative ways to either streamline their existing processes or develop new planning processes.

SAP pioneered building an integrated view of an enterprise and innovated the first enterprise systems, allowing organizations to break the system silos between internal functions. While the initial avatars of the SAP ERP were complex, the advantages of having an enterprise-wide system were many for organizations. One was the capability to build visibility into enterprise-wide processes and capture enterprise-wide data. A whole ecosystem of companies evolved in the SAP technology landscape to help organizations use this data best. Eventually, SAP recognized the importance of this market and launched products that allowed customers to further use the captured data from their SAP ERP systems. SAP IBP was one such product that became SAP’s flagship product for supply chain planning. This article discusses a real-world case of pairing SAP IBP capabilities with an AI Algorithm, shared in SAPinsider Vegas 2023.

SAP IBP still offers some of the most advanced features among all off-the-shelf integrated planning solutions. But as technology evolves exponentially, customers envision enhanced capabilities from their supply chain planning setup. With open-source tools available to build advanced capabilities leveraging AI and ML on top of SAP IBP, customers are looking for new and innovative ways to either streamline their existing processes or develop new planning processes. The session “NVIDIA Leveraged AI and SAP IBP to Transform Forecasting” on day 2 of SAPinsider Vegas 2023 focused on the use case that discussed how NVIDIA leveraged AI to build advanced planning capabilities of SAP IBP.

NVIDIA already had a robust AI platform that it offered as a product to its customers. NVIDIA partnered with CloudPath, a supply chain digital transformation company, to build an AI-enabled solution that integrated with SAP IBP and helps improve external factors forecasting for the company. The solution leveraged NVIDIA’s AI platform, SAP IBP, and third-party solutions like the Gurobi solver. The solution has two components combining two types of analytics approaches — predictive and prescriptive. The prescriptive analytics component of the solution develops the initial forecast. Since this component leverages deep learning, the solution can handle large amounts of data. The solution taps into data available in SAP and other data sources to generate an unconstrained initial forecast. The forecast then acts as an input to the second component of the solution, which is a prescriptive mixed integer programming algorithm. This is not just a great example of building solutions that combines multiple analytics approaches but also illustrates how SAP technology users can enhance the capabilities of SAP IBP by leveraging advanced algorithms in tandem.

As the amount of data organizations generates explodes, it impacts traditional forecasting methods like time series forecasting. Forecasting needs have evolved from forecasting thousands to millions of related time series. The expectation of forecast granularity adds to this challenge, specifically in industries with high SKU counts or in financial engineering, where there is a need to predict the stock prices of many companies in a portfolio. Also, realistically, forecasts depend not only on historical data but also on covariates like historical features, known events in the future, and attributes for each series. That is why traditional time series forecasting methods leave room for improvement regarding large and complex data. To summarize, the advantages of deep learning for forecasting are:

• Leveraging probabilistic approach to forecasting. A properly designed deep learning forecasting algorithm will not forecast values but will forecast future probability distribution. This is a more realistic approach vs. values and helps forecasters determine quantile estimates, improving business process optimization (like a better inventory estimate).
• The covariate approach. As mentioned, deep learning-based forecasting algorithms can capture complex and group-dependent relationships embedded in the data using covariates. This helps eliminate the effort required to select and prepare covariates and heuristics leveraged with traditional forecasting approaches.
• New product forecasting. Traditional forecasting methods designed to predict one time series at a time struggle to provide predictions for items with little or no history. Deep learning-based forecasting algorithms can use learning from similar items to generate robust forecasts.