Generating a Statistical Forecast

Objective

After completing this lesson, you will be able to use different forecasting algorithms in SAP IBP and employ the use of key figures

Generate a Statistical Forecast

The purpose of cleansing sales history data is to get a solid data foundation for the statistical forecast calculation. Therefore, we cleanse sales history for outliers. For example special events, stock out situations, etc.
Description of available pre-processing algorithms, forecast models and post-processing options.

Any statistical model could fit all data within an organization. As data will have different patterns of occurrence, it will need different models to forecast. For this reason, there are several forecasting algorithms available in SAP IBP for Time Series data.

Mid- or long-term forecasting is a demand planning process that helps to cope with

This figure points out that while mid- or long-term forecasting using advanced algorithms to run traditional mathematical models and machine learning algorithms, it helps to cope with the uncertainty of the future based on trends calculated from historical data. Demand sensing tries to make these forecasts more accurate in the short-term horizon using machine learning.

Single exponential smoothing can be used for forecasting the demand for mature products with fairly stable sales numbers. The algorithm detects irrelevant fluctuations in the data and smooths them using weights that are exponentially decreasing over time (that is, older data is given progressively less relative weight). The forecast calculated by this algorithm is a constant number based on weekly or monthly history.

Single exponential smoothing is used for forecasting demand for mature products with fairly stable sales numbers. The predicted outcome is a constant number.

Outputs from Statistical Forecasting: Local Demand Plan Quantity, Demand Planning Quantity leading to the Global Demand Plan Quantity. Outputs from Demand Sensing: Sensed Demand Quantity, Adjusted Sensed Demand Quantity, Final Sensed Demand Quantity, Combined Final Demand being used as input to inventory optimization.

Running demand planning processes implies flows of information and materials. To save the information in real time within the system, SAP IBP has created a technical container to store relevant numbers which have a business meaning. In the previous slide, we see some examples of these containers called Key Figures. As per SAP Best Practices for SAP IBP for Supply, a company can use these key figures with the predefined business meaning or adjust them according to specific requirements.

A good example of key figures application is described as follows:

Demand Planners and Sales Planners have information that has not been incorporated into the statistical forecast. The statistical forecast must be adjusted to include this information, removing promotions and outliers. The adjustments can be due to a special event, promotions, and so on.

The preconfigured key figures on the slide can be used as the definition recommends on the image. As shown, key figures may represent hierarchies, aggregating or disaggregating data from one planning level to other. The use of the predefined key figures is broadly described on the SAP Best Practices for SAP Integrated Business Planning (IBP) site.

In the subsequent figure, Logical Progression of the Demand Planning Process, you will see the logical evolution of the demand planning process considering the usage of key figures and SAP Engines:

Logical Evolution: Cleanse historical data, calculate a statistical forecast, adjust a local demand plan, adjust a global sales and operations demand plan, use analytics and dashboards, control exception-based alert, calculate and release a consensus forecast.

Within the first step of the demand planning process and as part of the demonstration Cleanse Historical Data , historical records will be cleaned to remove past promotions, outliers or extemporaneous events. Within the demonstration Create a Statistical Forecast and as the second step of this planning process, assuming that past events can be studied to recognize patterns for the long-, mid- and short-term, a statistical forecast model is used to create a prognosis for the future based on past cleaned historical records. As next step, based on special information from sales and other functional areas, the statistical forecast is adjusted to match with the reality of the business (demonstration Adjust a Local Demand Planner). This local demand plan is adjusted and aggregated to be an input for higher planning levels, for instance, the global sales and operations demand plan, where all local demand plans are summed up to come to a regional or global demand plan (demonstration Adjust a Global Sales and Operations Demand Plan). To publish and to control this global plan, charts and pin boards are created to visualize real time data. What is coming next? - What do we do if things do not go as planned? - We have to plan against exceptions and atypical occurrences. Therefore, within step 6 (shown in the demonstration Control an Exception-based Alert) exception-based alerts are controlled to take proper correction actions.

As shown in the figure at step 7 (shown in the demonstration Calculate and Release a Consensus Demand), Demand Planning can be covered as stand-alone. However, This process can be integrated with other processes as well. For example, in this course demand planning triggers the next planning process as Inventory Planning and Optimization and Supply Planning. That is, demand plan values are saved as a consensus demand for supply planning. On the other hand, forecast values are delivered to Sales and Operations to make strategical and tactical decisions. Finally, considering a premise that all forecasts are wrong, a safety stock must be calculated to hedge against run out of stock optimizing costs and reaching target service levels.

Demonstration how to Cleanse Historical Data

Demonstration how to calculate a Statistical Forecast

Log in to track your progress & complete quizzes