Demand Management Effectiveness

In order to effectively design and implement APO Demand Planning for the business, it is important have a clear working framework. Critical aspect is that Demand Management is an approximation; it is never a pure science. The key objective is to achieve an efficient Demand Management system that is user friendly ( different user profiles) and is a critical enable to achieve primary objective of business throughput taking into consideration business complexities:

• Supply chain pattern; make to order, make to stock, export markets, VMI ect..

• Product life cycle

• Proliferation of products

• Different roles in the business impacted by Demand Planning

SAP APO Demand Planning is one of the most un-structured solution in SAP; un-structured implies that the whole Demand Planning solution has to be built with the provided technical framework. It is purposely un-structured in that it has to be built to satisfy business requirements with respect to data views, data aggregation and data manipulation.

In order to effectively manage Demand Planning it is critical to understand the following:

• The technical framework

• Key elements in the technical framework

• The Demand Planning Process


DP TECHNICAL FRAMEWORK

This framework I provided standard in the APO Demand Planning system

This framewrok is provided by SAP to build the DP solution and consists of:

•  Planning Area where data is stored and manipulated
• Planning book’s and data view; the user front end for managing Demand Planning. Characteristics and key figures. This is the most critical area for user management.
•   Info Cube in the Data warehousing system needed for feeding data to the Demand Planning solution

The above must be specifically set-up to satisfy the business requirements.

ELEMENTS IN THE FRAMEWORK

Within the framework there are additional elements provided by SAP DP framework, these are:

• Macro’s for manipulating data and presenting data in user friendly-way (example red cell for exception )
• Data aggregation management; critical for data viewing and data consistency
• Standard forecasting models
• Standard forecast error calculation formula
• Tool-set for phase-in, phase-out, interchangeability
• Characteristics based planning
• Ability to upload data back to Data warehousing info cube
• Ability to integrate Demand Planning data with other Supply Chain tools such as Supply network planning, production , Sales and Operations Planning
• Authorization control
• Exception and alert management

The above all play a critical role in setting up an effective Demand Management solution and need to be carefully addressed.

THE DEMAND PLANNING PROCESS

The Demand Planning in most cases consists of a number of steps, different resources, different data granularity for each process, business and supply chain constraints. Therefore critical to understand the process so that correct and effective Demand Planning framework is set-up. In certain cases too much time is wasted in addressing a forecasting formula or forecast accuracy formula instead of understanding clearly the DP process. The understating of the process and exploiting the technical framework ensure the correct level of user-friendliness and desired objective.

The Demand Planning process consists of:

• History Management or data preparation for actual forecasting process
•  The actual forecasting and forecasting review process
• Consensus Management with different role players; marketing and sales , manufacturing
• Alert and exception management applicable to all three of the above process to ensure a more efficient data management. Critical for forecast accuracy, data manipulation (copying from one cell to another, mathematical calculation) and data presentation (red cell for phase pout period)

Furthermore the process is controlled / constrained by organization procedure, market behaviors and supply chain patterns.

The process is also managed by different resource that require unique data granularity; consensus forecasting with sales and marketing require data to be aggregated by brand, my markets, channel ect..

Understanding the above then determine how to exploit the provided framework.

HISTORY MANAGEMENT

The purpose of history management is to provide clean base history data that will be the input to generating the statistical forecast.

This data view must show current and prior year demand such as shipment, order , promotional data .

The data view must provide the level of detail needed to generate a fairly usable Adjusted History Base. Marco’s will help to identify outliers for user to understand how to address this aspect. Critical that user is able to view data by product grouping, brand, markets ect..The macro’s must also clearly provide exceptions allowing the user to prioritize their actions. Zero exception are also critical for user to analyses and understand.

This data view is one the most fundamental in that it provides the baseline data for statistical forecasting. Forecast formula’s , forecast accuracy formula’s all become irrelevant if the baseline is of little value.

STATISTICAL FORECAST MANAGEMENT

This data view must contain all the required data to manage the forecast by the planner. It must provide required key figures that are needed for the planner to have clear view on how the forecast should be managed. 

It must clearly show how well forecast is progressing and must provide all required data to manipulate and change forecast models and factors. It must also provide historical forecast accuracy performance data to indicate how well forecasting is progressing.

• It must have historical data for comparison purposes
• It must have forecast accuracy data to determine trends
• It must have alerts to efficiently manage exceptions with respect to forecast accuracy , outliers ect..
• It must allow focused data management; example flagging products that are phasing out and phasing in
• Monitoring and managing alpha, beta and gamma 

The above is then critical for the user to review forecast results, carry-out the necessary changes such as changing the forecast model or

Additional key figures that show bias values are also critical. These can help managed data with alert threshold for planner to rapidly review forecast data.

CONSENSUS FORECASTING

Once forecasting is completed, there is normally some form of consensus forecasting done with other entities like sales and marketing , production.

Data aggregation is critical when reviewing data with the specific business entity like marketing. Furthermore not only is data aggregation critical but also time disaggregation.

Depending on target for consensus, it could be that both monthly and quarterly data will be required:

Monthly View

Quarterly View

This is critical for data view and placing data granulality to suit the end target.

The data must be specific and uncluttered with unnecessary data.

As shown above, focused data for marketing review, data aggregation of data is critical; example customer group, brand ect…, and must clearly control how data at lower level is re-determined.

LEAN VIEWS:

Depending on target, helpful to have multiple data view including lean views. Lean views mean that only have basic data containing limited key figures at aggregated view. Example would be reviewing data with marketing team, different data elements when reviewing data with manufacturing.

The above defines  the minimum data needed to review data with specific business entity.

ALERT MANAGEMENT

The process needs to be supported by robust alert and exception management to help relevant planner to address specific data results. Alert must be process relevant.

Macro’s play a key role in managing alerts in the relevant data view. Sometimes alerts have to redetermine certain forecast error calculation. Typical consideration:

The mean absolute percentage error (MAPE), also known as mean absolute percentage deviation (MAPD), is a measure of accuracy of a method for constructing fitted time series values in statistics, specifically in trend estimation usually expresses accuracy as a percentage, and is defined by the formula:

Although the concept of MAPE sounds very simple and convincing, it has a major drawbacks in practical application ; If there are zero values (which sometimes happens for example in demand series) there will be a division by zero. This is an area where a custom macro could help to build own logic. Note; SAP provides standard macro function, but nothing stops one from creating totally new function (custom function module) with own logic.