Setting Up a Planning Area for Demand Sensing
Before you can use demand sensing, a dedicated planning area must be set up in SAP Integrated Business Planning (SAP IBP). Create a planning area in the Configuration app as a basis for your demand sensing and demand planning processes.
You may copy one of the sample planning areas delivered by SAP, such as the SAP6 sample planning area for demand, and adjust it as needed.
You can either have a separate planning area for demand sensing or you can share a planning area with other applications of SAP Integrated Business Planning. In both cases, the planning area must contain the following items:
Master Data Types
Demand sensing only runs at the Product ID –Location ID –Customer ID calculation level. The following master data types are mandatory for demand sensing:
- IBPPRODUCT
- IBPLOCATION
- IBPCUSTOMER
- IBPLAG
- IBPUOMTO
- IBPUOMCONVERSIONFACTOR
The following master data types are optional:
- IBPPROMOTION
- IBPDSSETTINGS
Time profile level: Days and Weeks
The demand sensing process is currently restricted to days and weeks. The key figure defined for the consensus demand plan must be available in weekly buckets. If the data is available in months, it must be disaggregated into weeks.
Planning levels
- Product / location / customer / weeks and a key figure for the consensus demand plan (transferred from SAP IBP Demand Planning)
- Product / location / customer / days and a key figure for the result of Demand Sensing (transferred to SAP IBP Supply)
Key Figures
You need to create key figures to store the following types of data:
- The Sales signal in the past, for example, ADJDELIVQTY (Delivered Qty Adj)
- The Open Order signal, for example, FUTUREORDEREDQTY (Open Sales Order)
- Daily disaggregation profile, for example, DAILYDISAGGPROFILE
- Optional data such as promotion plans and downstream signals
Additional Key Figures When Using Demand Sensing
The following table shows the mandatory naming conventions for key figure IDs:
| Example Key Figure | Definition | Business Meaning |
|---|---|---|
| FORECASTBIAS | Bias Adjustment Detail Key Figure | Short-term forecast bias |
| FUTUREORDERQTYPROFILE0 to | Open Order Adjustment Detail Key Figures | Forecast Adjustment Factor for Open Order |
| FUTUREORDERQTYPROFILE6 | Daily Disaggregation Profile Detail Key Figures | Open Order Profile Day 1 to Day 7 |
Note
Daily disaggregation is done using the profiles PROFILE0 to PROFILE6. Seven numbers in total represent the different days of the week.
Intermediate Key Figures That Help You Understand How the Sensed Demand was Calculated
Some of these key figures are optional and not delivered with the sample planning areas. To see how you can configure them, choose the titles of the table.
Optional Weekly Intermediate Key Figures
| Example Key Figure | Definition | Planning Level | Business Meaning |
|---|---|---|---|
| WEEKLYOPENORDER | Current open orders where the order creation date is before the base period | Product/Location/Customer/ Week | Intermediate Results - Weekly Open Order |
| WEEKLYOPTIMIZEDSD | Optimized sensed demand results from the machine learning algorithm before any post-processing | Intermediate Results - Weekly Optimized Sensed Demand | |
| WEEKLYCAPPEDSD | Capped sensed demand according to maximum increase and decrease settings in the forecast model, which are applied to optimized sensed demand | Intermediate Results - Weekly Capped Sensed Demand | |
| WEEKLYUPLIFTBALANCEDSD | Sensed demand results after planned promotion uplifts are balanced and added back to capped sensed demand | Intermediate Results - Weekly Uplift Balanced Sensed Demand | |
| WEEKLYBASEDBALANCEDSD | Sensed demand results after base balancing and open order matching steps are executed on the uplift balanced sensed demand | Intermediate Results - Weekly Base Balanced Sensed Demand | |
| WEEKLYCALBALANCEDSD | Sensed demand results after sensed demand values have been moved from all holiday periods to the preceding workdays | Intermediate Results - Weekly Holiday Balanced Sensed Demand |
Optional Bias Adjustment Key Figures
| Example Key Figure | Definition | Planning Level | Business Meaning |
|---|---|---|---|
FCTBIASWEIGHT0 | Bias horizon coefficient 0: machine learning weight. Optional – will only be populated if a key figure with the right business meaning and compatible planning level exists in the planning area. | LOCPRODCUSTLAGCALWEEKLY | Forecast Bias Horizon 1 Weight |
FCTBIASWEIGHT1 | Bias horizon coefficient 1: machine learning weight. Optional – will only be populated if a key figure with the right business meaning and compatible planning level exists in the planning area. | Forecast Bias Horizon 2 Weight | |
FCTBIASWEIGHT2 | Bias horizon coefficient 2: machine learning weight. Optional – will only be populated if a key figure with the right business meaning and compatible planning level exists in the planning area. | Forecast Bias Horizon 3 Weight | |
FCTBIASWEIGHT3 | Bias horizon coefficient 3: machine learning weight. Optional – will only be populated if a key figure with the right business meaning and compatible planning level exists in the planning area. | Forecast Bias Horizon 4 Weight | |
FCTBIASWEIGHT4 | Bias horizon coefficient 4: machine learning weight. Optional – will only be populated if a key figure with the right business meaning and compatible planning level exists in the planning area. | Forecast Bias Horizon 5 Weight | |
| FCTBIASWEIGHT5 | Bias horizon coefficient 5: machine learning weight. Optional – will only be populated if a key figure with the right business meaning and compatible planning level exists in the planning area. | Forecast Bias Horizon 6 Weight | |
| FCTBIASWEIGHT6 | Start of quarter bias weight for machine learning. Optional – will only be populated if a key figure with the right business meaning and compatible planning level exists in the planning area and the time profile includes WEEKOFQUARTER and MONTHOFQUARTER data. | Start of Quarter Bias Weight | |
| FCTBIASWEIGHT7 | Middle of quarter bias weight for machine learning. Optional – will only be populated if a key figure with the right business meaning and compatible planning level exists in the planning area and the time profile includes WEEKOFQUARTER and MONTHOFQUARTER data. | Middle of Quarter Bias Weight | |
| FCTBIASWEIGHT8 | End of quarter bias weight for machine learning. Optional – will only be populated if a key figure with the right business meaning and compatible planning level exists in the planning area and the time profile includes WEEKOFQUARTER and MONTHOFQUARTER data. | End of Quarter Bias Weight |
Optional Forecast Error Improvement Key Figures
| Example Key Figure | Definition | Base Planning Level |
|---|---|---|
TOTALMAPEIMPROV | Total forecast error improvement projected due to all demand sensing short-term forecast adjustments. Optional – will only be populated if a key figure with this exact name is added to a planning area at the right planning level. | LOCPRODCUSTWEEKLY |
Note
- Helper key figures that are needed to support currency or Unit of Measure conversion
- A key figure to store the information that a historical period should be disregarded during the learning phase of the Demand Sensing process
Demand sensing can be run only on the base version.
Definition and Usage of Snapshots
Lags in demand sensing represent the number of weeks into the future that you want to forecast. The current week would be the revision date.
Given a certain forecast interval, the so-called Lag N forecast is a forecast calculated in a certain period x for period (x+N).
Using Lag-Based Snapshots for Demand Sensing
A snapshot is a static copy of the values for a key figure within a specific time range. Lag-based snapshots are taken with predefined lags and saved in a key figure that has Lag as a root attribute in its base planning level. Demand Sensing needs lag-based snapshots of the consensus demand to analyze how the forecast calculated in the past as various lags compares and correlates to actual sales at each lag. This helps in optimizing the forecast for the future.