Blog updated last: December 6th, 2023 (influence request list & number of votes)
2022 is getting to an end. It's time to look back for a year that was full of predictive updates!
Right now a lot of SAP Analytics Cloud customers are using the integration into multi actions to ease the integration of predictions in their stories and their planning
You can learn more to this in the blogs here and there.
It is now possible to create predictive forecasts when business calendars are used in the underlying planning model. Example of such business calendars include 4-4-5, 13x4, or 5-5-6.
The blog here is making more concrete what the new capability means in particular for retail customers, or in general anyone that's doing planning on a weekly or periodic level.
This other blog explains the change from a planning standpoint.
A new visualisation "Impact of Numerical Influencers" was added in the predictive scenario Explanation area to better understand the impact of each influencer and the role they play when computing the predictive forecasts.
This model is forecasting bike hires in London - we can notice different influencing factors playing a role, for instance the maximal outside temperature on the given day.
The new visualization digs deeper for us to understand the impact of the temperature - as you can see the more the temperature increases on a given day (X axis) the more the impact to the predictive forecasts increases too.
A temperature of 31,20°C will increase the number of bikes hired by 14829 while a temperature of 2,1°C will decrease the number of bikes hired by 11905.
Back to the London Bike Hire example, you might have noticed that cycles are also playing a role in the predictive model. Typically the day of the week will influence if people rent more or less bikes.
Now we are curious to understand what's the specific impact of each day. This is what the new visualization "Impact of Cycles" offers.
Basically it says people rent more bikes on Fridays and Sundays compared to the rest of the week.
We do not rest on our laurels and have already enhancements planned for Q1 2023.
It is important for some customers to consume not only predictions for future periods (well... predictions!).
They also want to write back the predictions for past periods to compare them to actuals through their own story calculations as a way to gauge the accuracy of the predictive model.
Today it is possible for these customers to do it via the predictive scenario user interface but not through multi actions - we'll be bridging this gap with the new capability.
This capability is planned to be available in the wave 2023.01 and in the Q1 2023 release.
Beyond this, we of course have plans cooking but I now would want to hear from you too!
The enhancement requests have been created in the customer influence portal for SAP Analytics Cloud.
Without further due, here are the existing requests (sorted from the most voted to the least voted).
#1: Predictive Scenario Data Entity limitation to be extended
Do you want to be able to create a time series forecasting model with more than 1000 entities in one go? Then vote this idea from Anushankari Chandra with 18 votes to date.
#2: Selecting leaf members in an Entity filter
When one is defining the base element or combination to forecast on (aka the Entity) the hierarchical selection of the leaves could be eased. This is the focus of this enhancement request, which received 14 votes to date.
#3: Ability to save the confidence intervals from time series forecasting into planning model
Planned to be delivered in Q2 2024 QRC
Today one can save these confidence intervals when using datasets as a source but not planning models. Thanks adamnekola for raising this. It got 12 votes so far.
#4: Show ID and Description option not available in Predictive Scenarios
Another idea to ease the selection of an entity by danielweiskircher with 12 votes to date.
#5: Regression or Classification (Predictive Scenario) using Model
Today predictive planning is about doing time series forecasting on top of planning models. What if you could use regression too? This is the focus of this idea, that got 11 votes to date, from kassidyroussel
#6 Filters in Predictive Scenarios decoupled from the entities
danielweiskircher strikes again - this idea is about enriching the filtering possibilities when it comes to the entity definition (9 votes to date).
#7 Ability to select hierarchy in predictive scenarios
Roberts Veics raised this idea to ease filtering possibilities when creating forecasting scenarios on aggregated data. The idea has 8 votes so far.
Beyond all these great ideas, there are other tracks that one can think of:
Thanks for your attention, this is #predictiveplanning blog #49 and I am sure you can find interesting information in the other blogs.
2022 is getting to an end. It's time to look back for a year that was full of predictive updates!
In Q1 2022 predictive planning was integrated into multi actions, giving the possibility to automate the refresh of predictive forecasts on a regular basis.
Right now a lot of SAP Analytics Cloud customers are using the integration into multi actions to ease the integration of predictions in their stories and their planning
You can learn more to this in the blogs here and there.
Q3 2022 was a significant release with two incremental deliveries, the support of business calendars and increased visibility of the influencers' impact.
It is now possible to create predictive forecasts when business calendars are used in the underlying planning model. Example of such business calendars include 4-4-5, 13x4, or 5-5-6.
The blog here is making more concrete what the new capability means in particular for retail customers, or in general anyone that's doing planning on a weekly or periodic level.
This other blog explains the change from a planning standpoint.
A new visualisation "Impact of Numerical Influencers" was added in the predictive scenario Explanation area to better understand the impact of each influencer and the role they play when computing the predictive forecasts.
This model is forecasting bike hires in London - we can notice different influencing factors playing a role, for instance the maximal outside temperature on the given day.
The new visualization digs deeper for us to understand the impact of the temperature - as you can see the more the temperature increases on a given day (X axis) the more the impact to the predictive forecasts increases too.
A temperature of 31,20°C will increase the number of bikes hired by 14829 while a temperature of 2,1°C will decrease the number of bikes hired by 11905.
In the Q4 2022 release, a new visualization helps understanding the impact of cycles to compute the predictive forecasts.
Back to the London Bike Hire example, you might have noticed that cycles are also playing a role in the predictive model. Typically the day of the week will influence if people rent more or less bikes.
Now we are curious to understand what's the specific impact of each day. This is what the new visualization "Impact of Cycles" offers.
Basically it says people rent more bikes on Fridays and Sundays compared to the rest of the week.
Wait, what's coming in 2023?
We do not rest on our laurels and have already enhancements planned for Q1 2023.
It is important for some customers to consume not only predictions for future periods (well... predictions!).
They also want to write back the predictions for past periods to compare them to actuals through their own story calculations as a way to gauge the accuracy of the predictive model.
Today it is possible for these customers to do it via the predictive scenario user interface but not through multi actions - we'll be bridging this gap with the new capability.
This capability is planned to be available in the wave 2023.01 and in the Q1 2023 release.
Beyond this, we of course have plans cooking but I now would want to hear from you too!
Vote Existing Predictive Planning Enhancement Requests... or create yours!
The enhancement requests have been created in the customer influence portal for SAP Analytics Cloud.
Without further due, here are the existing requests (sorted from the most voted to the least voted).
#1: Predictive Scenario Data Entity limitation to be extended
Do you want to be able to create a time series forecasting model with more than 1000 entities in one go? Then vote this idea from Anushankari Chandra with 18 votes to date.
#2: Selecting leaf members in an Entity filter
When one is defining the base element or combination to forecast on (aka the Entity) the hierarchical selection of the leaves could be eased. This is the focus of this enhancement request, which received 14 votes to date.
#3: Ability to save the confidence intervals from time series forecasting into planning model
Planned to be delivered in Q2 2024 QRC
Today one can save these confidence intervals when using datasets as a source but not planning models. Thanks adamnekola for raising this. It got 12 votes so far.
#4: Show ID and Description option not available in Predictive Scenarios
Another idea to ease the selection of an entity by danielweiskircher with 12 votes to date.
#5: Regression or Classification (Predictive Scenario) using Model
Today predictive planning is about doing time series forecasting on top of planning models. What if you could use regression too? This is the focus of this idea, that got 11 votes to date, from kassidyroussel
#6 Filters in Predictive Scenarios decoupled from the entities
danielweiskircher strikes again - this idea is about enriching the filtering possibilities when it comes to the entity definition (9 votes to date).
#7 Ability to select hierarchy in predictive scenarios
Roberts Veics raised this idea to ease filtering possibilities when creating forecasting scenarios on aggregated data. The idea has 8 votes so far.
Beyond all these great ideas, there are other tracks that one can think of:
- Would you want SAP Datasphere and SAC Planning models to be coupled closer together, giving predictive planning additional capabilities when it comes to data management?
- Do you want every predictive forecast to be explained (what impact is coming from the trend, cycles, influencers etc) leading to this exact prediction? see here!
- If you are doing "what-if" hypothesis with influencers and generate multiple predictions from this, would you want to automate this logic? Planned in H1 2024!
- when you have many entities, is one indicator enough to assess the accuracy of the model? delivered in Q3 2023! see here.
- <YOUR VERY OWN & SPECIAL IDEAS THERE OR BETTER IN THE INFLUENCE PORTAL>
Thanks for your attention, this is #predictiveplanning blog #49 and I am sure you can find interesting information in the other blogs.
- SAP Managed Tags:
