The Analyze Promotions app allows you to calculate the success of past promotions. Promotion success is calculated in the root periodicity of the promotion level and can only include data that was loaded before the current period.
Before you trigger the calculation, you must select the key figure that contains the sales data you need to determine the average number of sold items. By default, the key figure with the Actual Sales business meaning is selected. If you select another key figure instead, the next time you trigger a calculation, this key figure displays.
The result of the success calculation is displayed in the Promotion Success column. By clicking the percentage value, you can navigate to the details.
For promotions based on key figures with the Promotion Total business meaning, the promotion success is calculated as follows:
Average number of sold items in the promotion periods divided by the average number of sold items in the periods without promotions
For promotions based on key figures with the Promotion Uplift business meaning, the promotion success is calculated using the following formula:
Average number of sold items in the promotion periods minus the average number of sold items in the periods without promotions divided by the average planned uplift.
These formulas do not require a baseline calculation.
A period without promotions is a period where no promotion exists for the relevant objects, for example, for product and customer if product and customer are the root attributes of the promotion level. Periods directly before and after promotions are not considered as periods without promotions because of possible pre-dip and post-dip effects.
Promotion success is negative if fewer items are sold in promotional periods than in periods without promotions.
Impact of Seasonality
To consider the seasonal behavior of the sales of a product, you select a specific number of periods per season. The following prerequisites apply:
You specify the number of periods per season in the root time level of the promotion level, that is, the time level in which the promotion is loaded.
If the number of periods per season is 0 or 1, seasonality is not considered.
If the number of periods per season is greater than 1, only those past periods in the sales key figure are used for the calculation of the average number of sold items in un-promoted periods that fit to the given seasonality.
Impact of Trends
To calculate promotion success, the system uses an algorithm with exponential smoothing for calculating the average number of sold items in periods without promotions. Only the periods before the promotion in question are considered for calculating the average. Using this strategy, you take trends into account that could be observed before the promotion took place. The periods closer to the promotion are given more weight in the exponential smoothing algorithm than the periods further in the past.
The smoothed average is calculated as follows:
- At = At-1 + (SF * (Pt - At-1))
- SF = 2 / (n+1)
At is the average at a certain point in time.
Pt is the sales value at a certain point in time.
SF is the smoothing factor.
Note
The assumption in the SAP6 modeling of promotions is that the planned Uplifts are integrated/loaded into SAP IBP. In the case that the total planned quantity and not the uplift is integrated, the SAP6 model needs to be adapted accordingly.
The assumption in the SAP6 modeling of promotions is that the promotions are integrated/loaded into SAP IBP at the Product/Customer/Calendar Week level. In the case that the promotions are at a different level (for example, product group), the SAP6 modeling needs to be adapted accordingly.
Promotion Sales Lift Elimination
Another possible option related to promotions in SAP IBP is the management of past promotions. You can do this with the help of the pre-processing step in the Manage Forecast Model app.
The algorithm is called promotion sales lift elimination.
This algorithm can be used to identify positive outliers (sales lifts) associated with promotions, and to remove them from the sales history. For example, you can set up a forecast model in which promotion sales lifts are eliminated from the sales history before the Demand Sensing algorithm calculates the short-term forecast.
Note
Process
The system executes the following steps to eliminate promotion sales lifts:
Identifies outliers in the sales history key figure that you selected as the input for the algorithm.
The outlier detection logic is the same as the one used by the outlier correction algorithm, while the multiplier may be different.
Identifies sales lifts caused by promotions in the sales history key figure that you selected as the input for the algorithm.
Searches for correlations between the two types of input, therefore identifying the periods when the outliers were caused by promotions.
Removes the outliers associated with promotion sales lifts from the sales history using the following formula:
- Key Figure to Save Result In = Sales History * [Consensus Forecast / (Consensus Forecast + Planned Sales Lifts)]
Looks for the closest matches to adjust the values proportionally whenever there is no exact correlation between the promotion sales lifts and the outliers.
Note