New technologies such as HANA databases, machine learning algorithms and artificial intelligence allow more and more routine tasks to be run automatically by computer systems. Accordingly, SAP is also continuously improving the planning process in its SAP IBP (Integrated Business Planning) supply chain planning tool.
This article will show you the sequence of automatic forecasting with SAP IBP, where a statistical forecast is for the most part generated autonomously by the system. The planner intervenes only in the event of critical materials or complex situations.
The process breaks down into the following main steps:
Product segmentation
The first step segments the products in IBP. Both an ABC classification of the products according to importance (e.g., price, margin or quantity) and an XYZ classification are carried out in order to distinguish between volatilities in sales performance. SAP IBP can either perform the ABC and XYZ classifications itself or adopt them from another system. Should the importance or the sales performance of the material change over time, the system will redefine the segments. But in addition, the values for special products can also be set and fixed manually, so as to rule out system-based changes in special cases.
Time-series analysis
Time-series analysis is another basis for automatic forecasting. It’s one of the most important areas of progress in SAP IBP of late. It can be used for identifying selling patterns of products and classifying them into the following categories. This information can then flow into an improved calculation of the XYZ segments, since the patterns related to trends and seasons can be excluded when determining volatility.
Automatic selection of the forecasting model
The product segments and the time-series analysis form the basis of the automatic selection of the forecasting model. The planner creates matching forecast profiles for the various segments and time series, to which one or more forecasting models are assigned. SAP IBP automatically selects the final forecasting model by means of the best-fit approach. A test period is specified in the historical sales data, and serves as a basis for comparing forecast against actual sales data. In this way, the forecast quality of the different models is assessed, and on this basis the model with the lowest forecast error can be selected.
Summary and future development
Through automatic forecasting, the planner can leave monotonous data-processing work steps to the system, and thereby gain time for attending to complex and important tasks as well as exceptional situations. In addition, optimization of the planning-process sequence – thanks to shorter planning cycles, among other things – has a positive effect on the entire company situation. Automatic forecasting in SAP IBP will also include new possibilities on a regular basis in the future, in machine learning and artificial intelligence, for example. We will of course be keeping you up-to-date about any other upcoming improvements and reliefs.
Author: Sebastian Fritzsche, Expert Supply Chain Management Consulting
