In this post, we will talk about the basics of Predictive Planning but elaborate to discuss the recent enhancement of Auto Predict which takes these capabilities to a new level with regards to EPM machine learning.
Planning has a functionality “Predictive Planning” that helps planners improve forecast accuracy based on the historical data.
The planner can perform the prediction using Data forms and through an Ad-hoc grid in Smart view.
When a planner runs a prediction, historical data for each member is retrieved and then analysed using time series forecasting techniques to predict these members’ future performance.
We can compare and validate plans and forecasts based on predictions. For accurate forecasting, we can copy the prediction values and paste them into a forecasting scenario for the plan.
Predictive Planning works with EPM Standard and EPM Enterprise applications for Custom and Module application types. For legacy applications, Predictive Planning works with Standard, Enterprise, and Reporting application types.
To perform the predictive planning on the valid data forms, perform the below steps:
- From a form click Actions ->Predictive Planning (Predictions runs and a result area opens below the form with a chart and detailed information) as shown below.
We can check the predictive planning for the different members present in the rows, as shown below:
We all know that Predictive planning was helpful to improve the forecast analysis using the historical data. Still, it was a manual process to run using the data forms or ad-hoc grid.
Unlike Predictive planning Auto predict is limited to data forms, but we can also predict values for thousands of cells, even schedule the prediction and automatically write the values to the Scenario/Version in the same cube or a different cube from historical data.
Using Auto predictions, we can automate the prediction process and write the prediction data to the cubes.
Prediction results are more accurate, the more historical data you have. There should be at least twice the amount of historical data as the number of prediction periods.
We can easily predict data at both weekly and monthly levels in the same application.
We can also include Best case and Worst-case prediction results.
Auto predict is helpful in the below scenarios:
- When a planner/analyst has a substantial amount of data to predict.
- To start the forecasting process by pre-filling a prediction scenario with predictions based on historical data. Then, planners can compare predictions to forecasts.
- For variance analysis.
- To pre-populate a Forecast or Plan scenario before starting a forecast cycle.
- To keep predictions up to date as actuals come in by scheduling Auto Predict jobs.
We can use these Auto predicted results in Forms, Dashboards, Charts, Ad-hoc grids and SmartView.
Auto Predict works with EPM Standard and EPM Enterprise applications for Custom and Module application types when Hybrid Essbase is enabled. Auto Predict works with the Enterprise application type for legacy applications when the Essbase version is upgraded to the version that supports Hybrid Essbase, and Hybrid Essbase is enabled.
To work with auto predict:
- Enable the Standard or EPM planning application to hybrid mode.
- Navigate to Application -> Overview -> Actions -> Auto Predict.
- Click on Create
- Enter a name and description for auto prediction and select the source cube and a prediction cube If you would like to predict data in one cube and Load the predicted data into another cube as shown below:
Or, if you would like to Predict the data within the same cube select option as shown below.
In the below example “Prediction” I am trying to predict the data in the same cube “OEP_FS” from
“OEP_Actual” ->” OEP_Forecast” scenario and “OEP_working” -> “Most Likely” Version.
and select the other dimension members and “Best case”, “worst case” (By selecting “+” icon from the right-hand side).
The system will analyse a maximum of 5 years of historical data and predict the data based on the Period and year selected for the prediction.
We can select the Start and End period for the prediction as shown below:
Once all the dimensions and time period is selected click on “Save” to save your Auto predict.
Once we create Auto predict, we can run it manually, or schedule a job to automatically run.
Go to Auto Predict Page -> select the job and click on the dots to run the Auto Predict Job as shown below:
This job will be running in the Job console, to get the status of the job click on the refresh button.
Once the job completes, the status of the job will be updated to “Completed”.
If you click on “completed” status, we can view the job details as shown below:
Now, Auto predict report is ready for download.
Once you click on the Download Report, A zip (AP_Prediction.zip) folder will be downloaded in the Inbox/Outbox explorer under Application, as shown below:
Auto Predict detail Report
December 2020 EPM update provides an option to download an Auto predict detailed report in an excel format.
This helps the business users/Planner/Analyst review details about prediction results and validate prediction accuracy. Allows further ad hoc analysis of prediction results, for example, to inspect results where accuracy is low, or to sort so you can quickly review data such as seasonality by series.
NOTE: planner can also perform ad hoc analysis on the prediction results using Smart View.
This detailed report is available once we run the Auto Predict report from Auto Predict page or the scheduled job.
This downloaded report is saved in Inbox/Outbox explorer in a Zip file format.
The report includes tabs like Summary Report.
Detailed report Summary tab looks as below:Report tab looks as below:
Report tab looks as below:
Here are some tips to Analyse prediction data:
- We can always copy-paste the dimension member from Report worksheet to other worksheets for further analysis if required.
- We can sort and analyse data in the Report sheet in multiple ways. For example, sort the Accuracy column to review low accuracy.
- We can use Excel chart features to visualise prediction results in different ways.
I hope you have found this blog helpful. If you would like to receive additional free-training to maximize the use of this functionality, please get in touch here.