Expand +



How to Reduce the Bullwhip Effect by Following a Demand-Driven Supply Chain Strategy

by Izak Pierre Erasmus, Program Lead and Consultant, SAP

June 8, 2016

Learn how to integrate consumer demand signals to achieve a demand-driven supply chain that is used to optimize short-term forecasts and increase demand network visibility. From a demand planning perspective, planners can now compare the traditional consensus demand forecasts with a point-of-sale (POS) statistical forecast and therefore apply corrections based on true consumer demand. From a demand network perspective, demand managers can now visualize the entire demand network, including the retailer distribution center, store stocking, and consumption points, and use this information to increase visibility to reduce the bullwhip effect. Extending their demand network visibility beyond their own supply chains allows organizations to optimize inventory, reduce out-of-stock lost sales opportunities, and increase profitability.


With an increasingly competitive market, manufacturers are competing for market share through brand loyalty, pricing, product innovations, and their ability to streamline their supply chain operations. Within mature markets manufacturers often have to optimize their supply chain operations by minimizing out-of-stock lost sales situations to increase revenue while decreasing waste overstock to reduce cost and ensure profitability. Research has shown that a constant out-of-stock situation affects a manufacturer not just in terms of lost sales opportunities and loss in revenue but also loss in market share as the consumer easily moves to a similar or competing brand. I provide insights on how to reduce the bullwhip effect by increasing demand visibility.

Demand visibility is obtained by extending the traditional view of supply networks beyond the manufacturer’s own supply chain to include its customers’ supply chain consumption and stocking points into its own supply network. This strategy increases extended demand visibility and provides a further opportunity to correct and improve the accuracy of supply chain planning, ideally reducing the bullwhip effect.

Traditionally, demand planners created a consensus demand plan with a forecast based on internal data, such as sales and shipment orders. This forecast focused on the assumed demand based on shipment to the retailer or distributors’ warehouses or distribution centers. If you revisit the traditional approach of forecasting within demand planning, you soon realize the limitations and issues related with this approach, described as the bullwhip effect.

You can reduce this bullwhip effect by increasing demand visibility and correcting or aligning your demand plan based on the current and short-term prediction of “true” consumer demand that could be achieved with an integrated, demand-driven supply chain strategy.

I show you how demand visibility can be achieved by integrating SAP Demand Signal Management with SAP Integrated Business Planning while using SAP Supply Chain Control Tower analytics for visualization.

I focus on how to derive harmonized point-of-sale (POS) data from SAP Demand Signal Management into SAP Integrated Business Planning for Demand and the SAP Supply Chain Control Tower. The SAP Supply Chain Control Tower is the analytics module that is part of SAP Integrated Business Planning. I also show you how to use these demand signals in SAP Integrated Business Planning for demand to create a POS statistical forecast and to use this to correct your consensus demand forecast. I demonstrate how to create a demand network to obtain an end-to-end visibility in the Supply Chain Control Tower.

Problem Statement

The bullwhip effect is causing continuous out of stocks that lead toward a decrease of market share for manufacturers.

With a demand-driven supply chain strategy, you can increase demand visibility and react in time to rapid market fluctuations.

Using SAP Demand Signal Management, you can generate demand signals and integrate the downstream demand data from your consumer demand. Using the Supply Chain Control Tower, you can increase your demand visibility via a demand network. Using SAP Integrated Business Planning alerts, you can react to out-of-stock situations before they happen. Using SAP Integrated Business Planning Demand, you can correct your short-term forecasts using the downstream demand signals.

Integrated Supply Chain Architecture

With an integrated supply chain architecture organizations can leverage synergies across SAP solutions, departments can share data using a single source of truth, and supply chain processes can be optimized. Collectively, this also equips organizations to communicate more effectively while creating transparency between functional and business roles. With an integrated supply chain architecture, business planning becomes less cumbersome. You can reduce efforts for preparing and sharing data across roles and for simulations.

Figure 1 shows what an integrated demand-driven supply chain architecture could look like. SAP Demand Signal Management is used to upload and harmonize POS data from retailers. The harmonized POS data is then used by downstream planning systems, such as Trade Promotion Optimization (TPO), SAP Advanced Planning and Optimization (SAP APO), and SAP Integrated Business Planning, to optimize planning forecasts to align with changing consumer demand. Additionally, the SAP Supply Chain Control Tower is used to create visibility, combining both internal data such as shipping and sales orders with external demand signals coming from SAP Demand Signal Management.

Figure 1
SAP Demand Management architecture

SAP Demand Signal Management is an on-premise solution, whereas SAP Integrated Business Planning and the Control Towers are cloud-based solutions. As shown in Figure 2, all data is transferred with an SAP HANA Cloud Integration agent via HTTPS from SAP Demand Signal Management to SAP Integrated Business Planning. 

Figure 2
SAP HANA Cloud Integration

SAP Demand Signal Management

SAP Demand Signal Management is positioned as a toolset to harmonize and enrich external demand data. Figure 3 shows how SAP Demand Signal Management is used as an automated solution facilitating the entire process to upload external data such as POS, market research, distribution, social media, weather, or a planogram. The data is then harmonized so that the downstream planning systems such as SAP APO or SAP Integrated Business Planning can consume the data. The SAP Demand Signal Management enrichment process also generates new key performance indicators (KPIs), such as out of stock, out of shelf, and lost sales, all of which can then be sent to the downstream planning systems. SAP Demand Signal Management also has its own analytics that can be used for drill-down data analysis.

Figure 3
SAP Demand Signal Management architecture

As shown in Figure 4, the SAP Demand Signal Management process includes several subprocesses to prepare and upload the data, perform quality checks, harmonize external master data against internal data, and enrich data with additional insight such as calculating out-of-stock signals.

Figure 4
The SAP Demand Signal Management process

The SAP Demand Signal Management harmonization process creates a global harmonized object for products, called Global Product Identifier (GPID), and harmonized objects for locations, called Global Location Identifiers (GLID). The GPID is created by uploading the manufacturer’s internal master data such as 0Material and 0Customer. The GPID, for example, then has harmonizable attributes such as a Global Trade Item Number (GTIN) or Universal Product Code (UPC) that can be used as common denominators by which to harmonize. External products and locations, meaning the number ranges provided by the institute providing the data coming from several different sources such as retailers, distributors, and market research data, as shown in Figure 5, can all be harmonized against the manufacturer’s own SAP ERP Central Component (ECC) master data. Now there is a relationship established between external data and internal master data.

Figure 5
The harmonization process

The SAP Demand Signal Management user interface is a fully personizable Fiori Launchpad and is grouped by roles (Figure 6). Tasks for each role can be launched via a tile. Here you check the Process Monitor for the person responsible for downstream data upload—for example, to upload POS data from retailers.

Figure 6
The SAP Demand Signal Management Fiori Launchpad

In the Process Monitor screen (Figure 7), you can view the details of all the processes running in an automated fashion and their statuses. On the bottom of the screen you can see the details for the selected process and the status of each sub-step, such as data upload, quality validation, harmonization, and enrichment. The traffic light gives a visual indication of the status.

Figure 7
The SAP Demand Signal Management process monitor

The cleansed, harmonized data can be integrated into the demand planning process. Here you release the POS data from SAP Demand Signal Management to SAP Integrated Business Planning for demand and the Supply Chain Control Tower. 

Harmonization in SAP Demand Signal Management is once again critical because the POS data shared by the retailer is not in the internal product nomenclature and the product names used in the demand planning process. To leverage this information it needs to be put in the same language as in the demand planning process.

All data deliveries are represented by a data delivery agreement that is set up in SAP Demand Signal Management. The Plan Data Deliveries screen (Figure 8) shows all plan deliveries for each data delivery agreement in a calendar view. The legend at the bottom shows the current date, the holidays, and the expected delivery dates. 

Figure 8
SAP Demand Signal Management Plan Data Deliveries

The cleansed, harmonized data can now be integrated into the downstream SAP Integrated Business Planning Demand Planning process. To enable that, the harmonized data is first prepared for integration on the correct planning levels. For example, the data is converted from a daily level to weekly.

Next, you release the POS data from SAP Demand Signal Management to SAP Integrated Business Planning for demand using the SAP Integrated Business Planning Integration User role (Figure 9).

Figure 9
SAP Integrated Business Planning Integration User role

The release data for the SAP Integrated Business Planning application (Figure 10) allows you to select and release aggregated and harmonized POS data. Additional insight is possible by selecting statistics to make sure that the sales and stock volumes are correct before releasing the data to SAP Integrated Business Planning. The application also enables you to filter the data by selecting the desired POS data dimensions that include customer, location, and product dimensions.

Figure 10
SAP Integrated Business Planning data release

SAP Integrated Business Planning

You are now in SAP Integrated Business Planning. Before you can integrate POS demand signals, you need to create the “containers” for master data and key figures in the SAP Integrated Business Planning planning areas. Select the Configuration tile (Figure 11).

Figure 11
SAP Integrated Business Planning Administrator role

The planning area in SAP Integrated Business Planning (Figure 12) contains all the required master data, attributes, and key figures.

Figure 12
SAP Integrated Business Planning planning area

In this case additional key figures need to be created on top of the SAP demo SAP6 or Rapid Deployment Solution (RDS). The key figures listed in Table 1 are required.

Key figure


Data provider


Inventory/stock on hand

ECC manufacturer internal data


POS out of stock

Calculated KPI – SAP Demand Signal Management enrichment


POS units

Retailer POS data


POS stock on hand

Retailer POS data


POS out-of-stock rate

Calculated KPI – SAP Demand Signal Management enrichment


POS promotional sales

Retailer POS data


Weather data - precipitation

Weather data provider


POS overstock demand signal

Calculated KPI – SAP Demand Signal Management enrichment


POS phantom stock signal

Calculated KPI – SAP Demand Signal Management enrichment


Weather data – average temperature

Weather data provider


Category market share

Retailer POS data


Store distribution

Calculated KPI – SAP Demand Signal Management enrichment


Lost sales

Calculated KPI – SAP Demand Signal Management enrichment

Table 1
SAP Integrated Business Planning key figures

By going to Key Figures as shown in Figure 13, all the newly defined key figures can be created. As guidance you can create the key figures one at a time as listed in Table 1. Follow the same process as per the example below. Add the key figure Name represented by column 1 and Description presented in column 2 as shown in Table 1. The Base Planning Level is on the WKPRODLOC (Week Product Location) combination. Select the Stored check box to indicate that the key figure should store the values loaded in the key figure.

Figure 13
Configuration key figures

After the key figures have been added, the planning area needs to be reactivated. By selecting the green icon with the arrow  as shown in Figure 14, you can activate the planning area.

Figure 14
Planning area activation

To manually upload all the newly created key figures, you need to select the Data Integration application (Figure 15). Note that the process to upload data into the key figures can be automated using SAP HANA Cloud Integration once it is tested.

Figure 15
Data integration

It is always a good practice to first get the data structure so that you can download a template of the planning area with all the key figure structure in it. The key figure structure contains all the key figures defined in the planning area. Selecting the Get Template button provides you with the entire structure. Figure 16 shows you where to select Get Template.

Figure 16
Select Get Template

As shown in Figure 17, select the Key Figure radio button and then select the planning area in which you created it. Select any of the key figures to get the complete structure and click the Get Template button.

Figure 17
Get Template key figures

The template opens up as an Excel file as shown in Figure 18. It’s usually advisable to upload the file manually before connecting it to SAP Demand Signal Management via the SAP HANA Cloud Integration HANA Cloud interface.

Figure 18
Data structure

You can update the Excel file obtained from the Get Template download shown in Figure 17 with the newly created values in the key figures columns. Save the Excel file and select Create New Import Job as shown in Figure 19. Browse to the path of the Excel file, select Insert/Update as the Operation Type, and click the OK button.

Figure 19
Import job

SAP Integrated Business Planning for Demand (Forecasting)

The traditional approach of creating a demand plan with a forecast based on internal data often created a supply chain strategy known as fix the mix. This approach results in a demand plan that could be used by production and inventory managers to plan accordingly, but the problem with this strategy is that you plan on top of historical plans and do not specifically include market fluctuations. The rippled effect of consumer demand is only reflected downstream once it’s too late to react to the changing demand. This can result in either out of stocks or too much stock, which might result in waste or increased costs.

The resulting bullwhip effect is explained as a distribution channel phenomenon in which forecasts yield supply chain inefficiencies. It refers to increasing swings in inventory in response to shifts in consumer demand as you move further up the supply chain.

In this article I show how this can be avoided by introducing a true demand-driven supply chain by including consumer demand as part of daily planning processes. The ideal demand plan is a plan that balances between reducing waste (minimizing cost) and supplying enough products to avoid lost sales (out-of-stock situations).

SAP Integrated Business Planning for Demand is a toolset that can be used to include the following influences in driving improved forecasting. Using the Multiple Linear Regression (MLR) algorithm you can introduce the following demand signal influences to improve and train an enhanced prediction:

  • POS demand signals
  • Historical weather data and predictive temperature and precipitation
  • Historical promotional data and planned promotion indication
  • Historical prescriptive and predictive out of stock 

Time series forecasting is the use of a model to predict future events based on known past events. Forecasting means predicting data points before they are observed and measured. In this case you are predicting future values for a desired time range.

The forecasting process includes the following:

  • Identify any trends
  • Analyze the fluctuations
  • Choose the best model, which is a combination of trends, cycles, and fluctuations
  • Predict future values for a desired time range 

Additional predictable variables as shown in Figure 20 can contain additional information that can be extremely useful for better detecting trends, cycles, or specific events (for example, promotions, temperature, and precipitation).

Figure 20
POS statistical forecast model

The forecast shown in Figure 21 shows how variables such as promotions, temperature, and precipitation can influence the forecast in terms of an upswing or downturn. For example, promotion, high temperature, and high precipitation all result in an upswing in the forecast for the product. Another variable such as out of stock might result in a downturn in the forecast.

Figure 21
Variable influences on a POS statistical forecast model

The effect that additional variables have on a forecast can also be determined using metadata logs. The effect that each variable has on the forecast model could be saved in a metadata log. For example, the influence of variables can be determined using variable contribution and variable weight. In Figure 22 you can see that the promotion variable has the highest positive variable weight in terms of influencing the POS forecast.

There are mainly two different measures of a variable to determine its relevance:

  • Variable contribution, which is the relative importance of each variable in the built model.
  • Variable weight, which is the weight (in the final polynomial) of the normalized variables. This is an expression whereby the normalized weight is determined in terms of the coefficient, and therefore, before multiplying the variable in an algebraic expression.

In Figure 22 I show a graphical example of the variable weight of the variables I used in Figure 21.

Figure 22
Variable weights in the POS forecast model

The following steps explain how to set up a POS statistical forecast in SAP Integrated Business Planning Demand.

The first step is to go to the Demand Planner role and select Manage Forecast Models option shown in Figure 23.

Figure 23
SAP Integrated Business Planning Demand Planner

The Manage Forecast Models application allows a user to set up a forecast. In Figure 24, select a planning area and define the forecast model in terms of periodicity and the type of algorithm used. In this case I select an MLR algorithm.

Figure 24
Manage forecast models

Part of the POS forecast definition is to set up the forecast following a process wizard. The four sections include general definition, preprocessing steps, forecast steps, and postprocessing steps. 

The GENERAL tab allows the user to define the historical periods on which the forecast is based. Also it allows the user to define how many periods to forecast. In my example shown in Figure 25 I selected 104 Historical Periods refering to 104 weeks of historical data and a forecast based on eight periods into the future. Best practice is to use two to four years of historical data and a forecast of six to eight weeks (six to eight periods) into the future. This usually alows seasonalilty to be determined by the forecast model from one year to the next.

Figure 25
The forecast model process

The PREPROCESSING STEPS tab includes all the activities for preparing the data for forecasting. Three of the general algorthims are substitute missing values, outlier correction, and promotion sales lift elimination. These preprocessing steps are to treat the data before running the forecast model. For example, outlier correction is used to detect and correct outliers, substitute missing values, and correct gaps in the data where applicable. Promotion sales lift elimination eliminates promotional outliers if the spikes in the data appear to be infrequent and are a one-off or infrequent event. In my example shown in Figure 26, I chose Substitute Missing Values as my preprocessing algorithm, key figure POSQTY (POS quantity) as input, and Actuals Qty as the output key figure.

Figure 26
Forecast preprocessing

Figure 27 shows the three types of preprocessing algorithms selectable when creating a preprocessing algorithm to your model.

Figure 27
Preprocessing algorithms

The FORECASTING STEPS tab allows you to define the input and output key figures used for the forecast. It also allows you to select the algorithm. In this case, you add the forecast algorithm of choice by clicking the plus icon  to the right of the Algorithms tab. In my example, as shown in Figure 28, I select the MLR algorithm with POS sales quantity, average temperature, precipitation, out-of-stock signals, and promotions sales as inputs for the variable selections.

Figure 28
Forecasting configuration steps

The POSTPROCESSING STEPS tab (Figure 29) allows you to define the steps after running the forecast. In this case I select Mean Absolute Percentage Error (MAPE) to determine the relevance of the forecasting results.

Figure 29
Forecast post processing

The second application for the demand planner role is to assign filters to the forecast model. Figure 30 shows how I assign the UPA_Triple Exponential Smoothing algorithm defined in the steps above to the Customer – Product combination for which I want to run this forecast.

Figure 30
Assign the forecast model

SAP Supply Chain Control Tower

You are now in the SAP Supply Chain Control Tower, which is a module within SAP Integrated Business Planning. By selecting the General Planner role in the SAP IBP Fiori Launchpad (Figure 31), you can select the SAP Control Tower dashboard tile to look at the data brought in from SAP Demand Signal Management.

Figure 31
The SAP Supply Chain Control Tower dashboard

In the SAP Supply Chain Control Tower dashboard you can find a series of Demand Network Analytics and KPIs running on top of SAP Demand Signal Management POS data. In Figure 32 on the top left, you can see the Demand Network analytics, which contain POS sales consumption and retailer stock data.

Figure 32
Demand Network analytics displayed in the SAP Supply Chain Control Tower dashboard

In Figure 33, exploring the DemandNetwork screen you find the entire manufacturer’s supply chain on the left (plant and warehouses). The retailers’ downstream distribution center (DC) and store external demand network are on the right.

Figure 33
Analytics displayed in the DemandNetwork screen

In Figure 34 by selecting any of the retailer stores on the right you find the granular KPIs, such as POS Stock, POS out of stock, and sales units.

Figure 34
DemandNetwork KPIs at the store level

Figure 35 shows how the different stores are supplied by a primary retailer DC. On this level you find the consumption sales and stock units, which can then be compared with the DC stock units. 

Figure 35
DemandNetwork at the DC level

In Figure 36, further to the left you find the selected Manufacturers Warehouse (shown as location node TI007_at_MANF), which is the supplying warehouse for the retailer’s DCs. Now you see the internal stocking and sell in sales and stocking points. In this case sell in sales is the sales orders from the manufacturer to the DC. Sell out sales is the sales from the retailer to the public.

Figure 36
DemandNetwork manufacturer warehouse

In Figure 37 all the way to the left you find the selected last node as the manufacturing plant. This node usually represents a smaller stocking point and additional KPIs such as production capacity and rates.

Figure 37
DemandNetwork manufacturer plant

When you go back to the DemandNetwork dashboard (Figure 38) and scroll down to the second part of the dashboard, you find different analytics including Phantom Stock, Out of stock by Location, POS Stock, and Over Stock Signals derived from SAP Demand Signal Management. This is the data that was transferred via the automated SAP HANA Cloud Integration process or manual upload via the Data Integration application.

Figure 38
Analytics displayed in the DemandNetwork dashboard

In Figure 39 you can drill down from the DC to the store level Out of Stock. The bigger the bubble, the larger the location for the out-of-stock rate.

Figure 39
Out-of-stock rate on the GEO map

Scrolling down in the Supply Chain Control Tower Dashboard (Figure 40), you see more analytics containing a combination of different external data, including weather and distribution calculations, all derived from SAP Demand Signal Management.

On the top left you can see the combination of POS Qty and distribution. The graph demonstrates that as sales increase at flagship stores, distribution drops. This means that the flagship stores withdraw most of the retailer DC stock when they expect higher sales volumes, which leaves many of the other stores with an out-of-stock situation. On the top right you can see as distribution drops, lost sales increase, directly related to out of stocks.

On the bottom left you can see as precipitation increases, sales increase for a specific product set in the filter, in this case chocolate ice cream. On the bottom right you can see as the temperature increases, sales increase for chocolate ice cream, and when both precipitation and temperature are high, sales increase dramatically.

Figure 40
More analytics displayed in the DemandNetwork dashboard

SAP Integrated Business Planning Alerts

The demand manager can now either jump to alerts from the dashboard or via the SAP Integrated Business Planning Alerts role shown in Figure 41.

Figure 41
SAP Integrated Business Planning Alerts role

A case or specific type of task containing a series of alerts can be assigned to a specific group or user as shown in Figure 42.

Figure 42
Forecast correction task

In this view (Figure 43) you can now see which alerts are high versus low priority. 

Figure 43
Alerts overview

You can now view all the alerts and see the nature of each alert as shown in Figure 44. In this case the POS statistical forecast is higher than the Consensus Demand Plan, and therefore, the demand planner is alerted to make a manual correction.

Figure 44
Custom alerts in the consensus demand plan forecast correction

SAP Integrated Business Planning for Demand

In SAP Integrated Business Planning for Demand, the demand planner can see a tabular form of all the data in the dashboard by scrolling over to where the forecast is (Figure 45).

Figure 45
SAP Integrated Business Planning for demand Excel add in—POS Data Analysis

The demand planner can see the projected POS sales for the stores for the next few weeks and also have a forecast of what is expected to sell. Given a gap between the POS statistical forecast versus the consensus forecast over the last few weeks, the demand planner needs to make some manual corrections and adjustments as shown in Figure 46.

Figure 46
The consensus demand forecast versus the POS statistical forecast

In Figure 47 the demand planner decides to bump up the forecast to 3,000 units for both DCs.

Figure 47
Consensus demand forecast correction

Now the demand planner saves the consensus demand plan as shown in Figure 48. The system wants the demand planner to enter a reason for the adjustment to track why the changes were made. 

Figure 48
Save correction

The data is now updated as shown in Figure 49. The demand planner has been able to analyze the recent patterns in sell-in and sell-out data along with the POS statistical forecasts and react to a potential continuing problem by increasing the consensus forecast to have stock available for an anticipated increase in demand.

Figure 49
Updated consensus forecast

An example of the process flow for using the POS Forecast Corrections and Demand Network analytics is presented in Figure 50.

Figure 50
POS forecast correction and demand network process flow

(For more information on demand sensing and how to counteract the bullwhip effect, read "Real-Time Supply Chain Planning with Demand-Driven MRP in SAP S/4HANA" and "A Look at Demand Sensing for SAP Customers.")

An email has been sent to:


Izak Pierre Erasmus

Izak Pierre Erasmus is practicing as program lead and consultant at SAP. He specializes in supply chain planning, demand management, solution management, and architecture. As a supply chain consultant, he has led several SAP Demand Signal Management implementations across the United States and Europe. The implementations were often based on the integration of external demand signals derived from data such as point of sale (POS), Nielsen, IRI, social media, weather, and planograms. He also acts as the global program lead for the SAP Model Company, a program designed to move strategic on-premise solutions to the cloud with a subscription price model, whereby the components are preinstalled, associated with best-practice configuration and integrated. The design is focusing on integration between multiple solutions to derive synergies between different functional departments, often hidden in silos of data. Today, there is a hybrid focus of mixing on-premise solutions with cloud solutions. Pierre took an architect and leadership role within SAP around the integration of demand-driven supply chain, advance forecasting, and demand network topics. He has a business intelligence background that includes experience with SAP Business Warehouse (SAP BW), SAP HANA, BusinessObjects Data Services and Analytical (BusinessObjects dashboards, Design Studio, and Lumira). Other areas of interest include research on phantom stock detection, trade promotion optimization, integrated business planning, shelf management, and advance forecasting algorithms.


Please log in to post a comment.

No comments have been submitted on this article. Be the first to comment!