Automate Forecast Update With Predictive Planning
Updating a forecast is a complex activity. It requires coordinating several distinct steps. A forecast update in SAP Analytics Cloud might contain the following steps:
- Prepare the Actuals using a Data Action
- Train a Predictive Model
- Write the forecast to a private version
- Perform some allocation
- Publish the forecast
You would have to run all these steps every month. This is time consuming and there is a risk of doing things that are not entirely consistent.
In wave 2022.01 (first quarterly release of 2022) SAP Analytics Cloud introduces the Multi Actions Predictive Step, a new feature that enables the planning users to control Predictive Planning from a Multi Action. All the steps required for the forecast update are now coordinated by a Multi Action: there is no more risk of inconsistency and as Multi Actions can be scheduled the forecast is updated automatically.
I suggest you read this blog post if you are not familiar with Multi Actions yet.
Let’s consider a simple scenario: you are working for a company selling Champagne. Every month you generate an updated forecast for the next 3 months based on the latest Actuals.
You want the planning version where the forecast is written to be automatically published so the updated forecast is immediately visible in the stories.
In this blog post, you will learn how to create a Multi Action that will contain the steps to retrain a Predictive Model, write the forecast and finally publish it.
In a more realistic scenario, you would also want to add some data preparation before the forecast generation and maybe some allocation after the forecast generation. We will skip these steps to keep this post short and simple but know it’s something that would be possible thanks to the Multi Actions.
In the rest of this post, we will assume that you are working with a Planning Model called “champagne sales” that contains your Actuals and will receive the forecast.
You can download this dataset if you want to recreate this example.
Step by Step
The Predictive Model
As a starting point you need a Predictive Model to be used in the Multi Action. The point is that the Multi Actions Predictive Step is not meant to be a replacement for Predictive Planning, but a mean to automate Predictive Models that you have designed using all the Predictive Planning capabilities.
I won’t detail step by step how to create a Predictive Model as I assume you are already familiar with Predictive Planning. Though, I’d like to highlight a few settings that have a specific importance in the context of the Multi Actions.
First, it’s particularly important that you always use the setting Until: Last Observation. It’s the setting that will allow your predictive model to always use the data up to the last known actual. If you use the option Until: User Defined Date, then as the selected end date is fixed, the training data source will be the same for each retraining and therefore the forecast will be same.
The second setting to consider is the Predictive Model description. If you have created several Predictive Models in your Predictive Scenario, I strongly recommend that you provide a description for each Predictive Model. This description will help you to select the right Predictive Model in the Multi Action later. If no description is available then the default Predictive Model name is displayed (“Model 1”, “Model 2”…).
In the rest of this post, we will assume you have created at least one Predictive Model into a Predictive Scenario called “sales forecast.”
The Multi Action
Multi Action Creation
In the SAC left toolbar, click the Multi Actions button, then click the Create New Multi Action button.
Provide a name and a description for the Multi Action in the right-hand panel.
Make the Output Version Dynamic
It’s not strictly necessary for our scenario but I’d like to show you how to use Parameters, so it’s possible to specify later where the forecast must be written. It’s usually handy because as you are creating the Multi Action the output version may not exist yet and therefore cannot be selected.
Click the Parameter button then in the right-hand side panel click Add Parameter.
Setup the Parameter as specified in the screenshot below. Selecting the Dimension called Category will allow you to select a Version.
Predictive Step Setup
Now we will create the Predictive Step (the action in the Multi Action that will run the predictive operations). In the top toolbar click the Add Predictive Step button (binocular button).
A node is added to the graph.
In the right-hand side panel, setup the Predictive Step as specified in the screenshot below:
Here is in details how to select the Parameter for the output version (Save Forecast To parameter):
The configuration is straightforward as all the predictive settings are pre-defined in the selected Predictive Model. You only must select the Predictive Model to be used and the forecast output version.
If you have provided a description for your Predictive Model you will see this description when selecting the Predictive Model, otherwise you will only see the Predictive Models name (Model1, Model2…).
Note that you don’t have to manage a Predictive Model retraining step and a “Save Forecast” step separately: The Predictive Step (as highlighted in the Predictive Action field) will systematically retrain the Predictive Model and then immediately write the forecasts to the output version.
Version Publication Setup
In our scenario we want the forecast to be published automatically after it has been written to the output version. To do so you must add a Version Management Step.
In the top toolbar click the Add Version Management Step button (post sign button).
Then select the version to be published as shown below:
The predictive Multi Action is now ready to be used. Don’t forget to click Save.
Scheduling the Multi Action
Finally, you must setup your Multi Action to run every first of the month each. This is done in the Calendar.
Click Add Multi Action Task.
Setup the execution schedule as shown below:
Click Create to create the schedule.
In the right-hand side panel scroll down to the Multi Action section. You must select the Multi Action to be scheduled and the output version for the forecast.
Trigger the Forecast Generation from a Story
Sometimes, scheduling the forecast update may not be desirable or workable. It could be that this month there were some hiccups in your data collection process and the actuals were available on the 15th of the month instead of the 1st of the month. In such cases you may want to make it possible for the story users to trigger the Multi Action manually
To do so, in the story toolbar click the Add button (“+”) and then click the Planning Trigger menu item.
Then click Multi Action Trigger in the dialog box.
Fill the Multi Action parameters as shown below:
That’s it, the Multi Action Trigger is now ready. Any user of the story can click the trigger button and select an output version where he wants the forecast to be written:
After the Multi Action has finished, the story is automatically updated and displays the forecast values.
Version Management Recommendations
As a rule of thumb, if your Multi Action contains a publication step, always write the forecast to a public version. Private versions are deleted after publication, meaning that you will have to manage their recreation after each run of the Multi Action.
In this blog post, you learned how to create a Multi Action to package a consistent and repeatable predict process for your planning. You also learned how to schedule the Multi Action so the forecast is updated automatically and how to add a Multi Action Trigger to a Story so any story user can generate an up-to-date forecast.
I hope this blog post was helpful to you. If you appreciated reading this, I’d be grateful if you left a comment to that effect, and don’t forget to like it as well. Thank you.
Do you want to learn more about Predictive Planning?
- You can read our other blog posts.
Find all Q&A about SAP Analytics Cloud and feel free to ask your own question here: https://answers.sap.com/tags/67838200100800006884
Visit your community topic page to learn more about SAP Analytics Cloud: https://community.sap.com/topics/cloud-analytics/augmented-analytics