Reducing forecast errors, especially for near-term forecasts, has been a challenge for companies and their supply chains for years. Technology has evolved from demand planning or statistical forecasting to demand collaboration to demand signal. (See the sidebar, “Evolution of Demand Planning,” for more detail.) Today, SAP offers demand management capabilities as part of five different solutions:

• SAP ERP Central Component (SAP ECC) Flexible Planning (FP): (current version: ECC 6.0)
• Advanced Planning & Optimization (APO) Demand Planning (APO DP): (current version: SCM 7.1)
• APO Forecast and Replenishment solution for the retail industry (APO F&R): (current version: F&R)
• Supply Network Collaboration customer collaboration (SNC CC): customer collaboration solution for demand and promotion collaboration (current version: SNC)
• Demand Signal Management (DSiM), the fourth-generation demand management solution from SAP, which is based on SAP HANA. The current version of DSiM (DSiM 2.0) comes free with the SAP Sales and Operations Planning (S&P) on HANA solution

Table 1 lists the pros and cons of each solution, along with the situations for which each solution is most suitable.

Table 1

Pros and cons of five demand management solutions from SAP

Sidebar: Evolution of Demand Planning

The demand management process has evolved from simple Excel-based tools to today’s highly sophisticated demand signal applications, as shown in Figure A. I group these tools in four generations, as follows:

• First-generation tools: Tools such as Excel used for demand planning. Planners used to have their own Excel spreadsheets for planning.
• Second-generation tools: Planners started using ERP tools for demand planning. In SAP ECC, Flexible Planning was used as a demand planning tool.
• Third-generation tools: Integrated demand planning tools such as APO Demand Planning emerged. For the retail industry, SAP came up with a separate demand planning tool, Forecasting and Replenishment (F&R).
• Fourth-generation tools: More sophisticated tools for demand collaboration and demand signal management emerged. SAP introduced SNC customer collaboration tools for demand collaboration and DSiM tools for demand signal detection and management.

 Figure A shows how demand management solutions have evolved over time.

Figure A

Four generations of Demand Management solutions

Statistical Forecasting and Traditional Demand Planning Applications

Demand planning solutions are widely popular across industries for managing the end-to-end demand management process. Advanced demand planning applications such as SAP APO cover a wide variety of statistical forecasting algorithms that help organizations forecast demand for future time buckets. Statistical forecasting algorithms can forecast with acceptable accuracy for products with a history of at least three years, but they have trouble producing accurate forecasts in the following cases:

• New stock-keeping units (SKUs): There is little history available for new SKUs, making forecasting difficult. Demand management applications support some planning tools, such as modeling, to forecast new products. However, these are not effective for forecasting a new SKU that does not closely mimic the sales behavior of an existing SKU.

• Seasonal SKUs or SKUs sold for a very short time: Seasonal SKUs are not new SKUs, but regular SKUs that are sold during particular seasons. Typically, forecasting of SKUs that are sold for a very short period, such as near Christmas, is kept outside the scope of statistical forecasting because these products do not have a regular history.

• SKUs that are highly promoted: Statistical tools are good for forecasting a baseline. They cannot forecast a promotion.

• When past sales are not a good predictor of future sales: Time series techniques create a forecast based on prior sales history and draw on several years of data to provide insights into predictable seasonal patterns. However, time series models are disconnected from events that can affect demand in unpredictable ways, such as a financial downturn or recovery or changing weather patterns. Time series models are ill suited for volatile markets.

• Forecasting that requires a huge amount of data: Testing the impact of weather on forecasts and forecasting based on downstream data (such as retail POS data) require a huge amount of data, and that can create performance problems for traditional data warehousing applications such as SAP NetWeaver BW.

These issues present major challenges for industries such as consumer goods and consumer durables. In these types of industries, promotional volume can be as high as 30 percent of overall yearly sales. Today, new product proliferation has become the norm and, for certain industries, new products account for 25 percent to 40 percent of yearly sales volume. In the smart phone industry, for example, half of the items have fewer than two years of history, making statistical forecasting difficult.

DSiM

DSiM helps boost sales by predicting and reducing stock-outs with demand patterns and trends recognition, predicts demand by tapping data from enterprise demand systems and downstream demand signals, and lowers expediting and repositioning costs through improved short-term forecasts. DSiM can talk with multiple sources, including planning and scheduling applications, ERP applications, external data sources, and Excel files.

DSiM leverages new mathematical algorithms and near-real-time information to create a forecast based on factors currently affecting the supply chain, such as market shifts, weather changes, natural disasters, and consumer buying behavior. Thus, DSiM uses a much broader range of demand signals than traditional statistical forecasting, which only looks at historical sales data.

Using such a wide variety of demand signals requires the processing of a large amount of data associated with hundreds of thousands of items and location combinations every day in a very small processing window. For a consumer goods company, when retailer data, such as promotion data (items, prices, sales), launch data (specific items to be listed/delisted, ramp-down plans), and inventory data (stock levels per warehouse, sales per store), is included, the amount of information to be analyzed can grow exponentially.

The DSiM application supports all three major demand signals: internal demand signals (SAP suite data such as orders, shipments from SAP ECC, demand plan from APO, promotion from APO), downstream demand signals (such as POS), and external demand signals (mainly two external demand signals are supported now: social media and market research data).

Accessing massive amounts of POS data, syndicated market research data, and social sentiment indicators was more difficult before the arrival of SAP HANA. With the enormous data-crunching ability of SAP HANA, DSiM can quickly spot market trends and deviations, and drill down into massive amounts of data to develop market insights and understand consumer trends. Integration with all back-end SAP solutions helps DSiM respond faster to demand fluctuations and accelerate downstream processes.

DSiM is tightly integrated with many solutions from the SAP solution suite, including APO, SAP Demand Planning, SNCC, Sales and Operations Planning, and retail solutions such as F&R, ERP, CRM, and Trade Promotion Optimization. DSiM can take inputs from these applications or can feed these applications with data.

DSiM is relevant for multiple business functions, including sales, marketing, and supply chain (Figure 1).

Figure 1

The role of DSiM in data integration, data repository, and data consumption

For sales, it helps visualize sales performance, avoid lost sales, make promotions successful, and increase price effectiveness. It lowers expediting and repositioning costs through improved short-term forecasts. It can talk with multiple sources, such as planning and scheduling applications, ERP applications, external data sources, and Excel files.

For marketing, DSiM helps avoid new product launch failure and defend an existing customer base. For the supply chain, it helps improve forecast accuracy, avoids stock-outs with demand pattern and trend recognition, predicts demand by tapping data from enterprise demand systems and downstream demand signals, and optimizes replenishment plans for retailers.

How DSiM Is Different from Traditional Demand Planning

Demand Signal applications are different from traditional demand planning applications in a number of ways.

• Demand Signal applications can consider a variety of demand signals, such as internal demand signals (orders, shipments), downstream demand signals (retail, POS) and external demand signals (social media, weather, or market research). Traditional demand planning considers only internal demand signals (orders or shipments).
• Demand Signal applications are more suitable for short-term forecasting. These applications update the forecast daily, whereas traditional forecasting updates typically happen monthly.
• Demand Signal applications use a different set of algorithms than statistical forecasting applications. These applications use what are known as self-learning or self-tuning algorithms. They learn from the most recent data and tune the forecast continuously to keep it up-to-date. They consider not only current orders but also previous order patterns, consumption logic, and other data that was not available during the original forecasting process. These applications use specialized order-pattern recognition algorithms and optimal forecast-blending techniques to predict the future.

Many companies use DSiM and demand planning applications together to achieve high forecast accuracy. The demand planning application provides raw forecast data to DSiM, which then considers more recent data in calculating the sensed numbers, and it feeds this back to the demand planning or forecasting application.

In cases like this, the forecasting application simply extrapolates the past into the future: trends in past sales, past seasonality, and past promotional impact are carried forward to produce a calculated estimate of future demand. DSiM takes that traditional forecast as an input and adds to the mix real?world events such as market shifts, weather changes, changes in consumer buying behavior, social network sentiment, and real-time POS data to fine-tune the short-term estimate.

Figure 2 shows the difference between demand planning and demand signal applications.

Figure 2

Demand Planning vs. Demand Signal Management

Rajesh Ray

Rajesh Ray currently leads the SAP SCM product area at IBM Global Business Services. He has worked with SAP SE and SAP India prior to joining IBM. He is the author of two books on ERP and retail supply chain published by McGraw-Hill, and has contributed more than 52 articles in 16 international journals. Rajesh is a frequent speaker at different SCM forums and is an honorary member of the CII Logistics Council, APICS India chapter and the SCOR Society. 

You may contact the author at rajesray@in.ibm.com.

If you have comments about this article or publication, or would like to submit an article idea, please contact the editor.