The time series data analyzed by SAP Analytics Cloud (SAC) Predictive Planning are measured on dates based on a Gregorian calendar. The granularity comes from the planning model. Comparing two consecutive years with such calendar is difficult. The reason is that weeks and weekends are not aligned. This is a disadvantage for many organizations in the retail and CPG industries. Firstly, their data are measured on a weekly basis and secondly, they need to see the evolution from year to year and at the same periods.
A business calendar consists of dividing a year into periods, all based on the same pattern. For example, in a 4-5-4 calendar, each quarter is divided in three periods of 4 weeks, 5 weeks and 4 weeks. All weeks contain seven days from Monday to Sunday. If you want to learn more, I encourage you to read the excellent blog Introducing Flexible Time in SAP Analytics Cloud from Scott Godfree. It explains how in Q2-2022, this kind of calendar was introduced in SAC Planning.
Now, SAC Predictive Planning uses these calendars to generate predictive forecasts on a weekly basis. The objective of this blog is to explain how it works and what are the advantages.
What kind of Business calendars?
When the data source is a planning model, the accepted calendars are:
- Gregorian calendar,
- Custom calendars.
The following figure shows a planning model where the time dimension is based on a business calendar with a 4-5-4-week pattern.
Figure 1: Setting of a 4-5-4 business calendar
This planning model presents the sales revenue of a retailer per country between 2018 and 2022. The figure below shows how the years are split into periods following the 4-5-4 pattern.
Figure 2: Story showing the weekly sales on a 4-5-4 calendar
A year is divided in quarters. All quarters contain three periods prefixed by ‘Z’ as specified in the settings of Figure 1. The first and third period are split in four weeks, while the second period is split is five weeks.
Effects of business calendars on SAC Predictive Planning
I continue with my retailer example to explain how business calendars are managed by SAC Predictive Planning. I create a Predictive Scenario to get an estimation of my sales in the eighteen next weeks for each country. I also want to compare week by week the evolution of the sales revenue between years.
In the setting panel of the predictive model, the time granularity is the same as the granularity of the planning model. This means that SAC Predictive Planning handles now a week time granularity, as shown in the figure below.
Fig 3: Settings of the predictive model
I train the predictive model with these settings, and I get these results for each value of entity Country.
Fig 4: Performances of the predictive model
One word about the predicted dates. In our training dataset, the last observation is on the end of December 2022. In the settings of the predictive model, I requested to predict for the next eighteen weeks. This means that the first prediction will be for the first week of 2023 of the business calendar until the first week of the second period of Q2-2023, as shown below.
Fig 5: Predictive forecasts for country France
Now, I want to compare the evolution of the market on the same periods. For this, I save the predictive forecasts into a private version of my planning model, as shown below.
Fig 6: Save predictive forecasts
The story is updated with the predicted forecasts.
Fig 7: Story with the predictive forecasts
The predictive forecasts are generated at dates corresponding to the granularity of the data source. Weeks in the example. Each predictive forecast is assigned to the corresponding week of the business calendar used in the planning model.
As a retailer, I would like to study the evolution of my sales revenue and compare it on period:
- of the same duration and
- which are at the same date from year to year.
This is the benefit of business calendars. I add to the story with a chart controlled by two input controls: one to select a country and one to choose the periods. The figure below shows a reduction of sales revenue on the first period of each year. But a small bounce is predicted for the first period of 2023.
Fig 8: Story with comparison of sales year to year
Let see on the same data what the retailer will get with a 13x4 business calendar as the one shown below.
Fig 9: Date settings of the planning model with a 13x4 business calendar
The story below shows that the sales revenue by country reveals that a year is split into thirteen equal period: Z01 to Z13. First, second and third quarter contains three periods while the fourth contains four periods.
Fig 10: Story showing the thirteen periods of 2020
The periods are split into four weeks of 7 days.
Fig 11: A period is split in four weeks
In my predictive scenario, I create a new predictive model based on the 13x4 planning model. The goal is the same: generate predictive forecasts for the next eighteen weeks.
Fig 12: Predictive forecasts per country for the next eighteen weeks based on a 13x4 business calendar
The predictive forecasts are the same because it is the same data source as before. This time, the predictive forecasts are distributed on weeks corresponding to the 13x4 business calendar.
Fig 13: Distribution of the predictive forecasts in the periods of the 13x4 business calendar
Conclusion
What was shown in this blog is the benefits of business calendars and how SAC Predictive Planning handles them in the generation of the predictive forecasts, in the explanation pages and when predictive forecasts are saved into a story.
I hope that reading will make you more comfortable with business calendar and how to use them with SAC Predictive Planning. Feel free to test it, and I would be delighted if you share your experience with me.
Resources to learn more about SAC Predictive Planning.
- SAC Predictive Planning playlist
- Introducing Flexible Time in SAP Analytics Cloud
- Planning on data on a weekly basis
- Getting Started with SAP Predictive Planning
Finally, if you enjoyed this post, I’d be grateful if you would help spread, comment and like. Thank you!
- SAP Managed Tags:
