Reference the configuration options for setting up data, defining model parameters, and configuring debug settings when creating a Scoring Framework template configuration for Personalized Promotions.
Required Editions
| Available in: Lightning Experience |
| Available in: Enterprise, Performance, Unlimited, and Developer Editions with Loyalty Management and Data Cloud |
Set Up Your Data
| Option | Details |
|---|---|
| Select Data Space | Required. Select the Data 360 data space that the template configuration reads from and writes data to. |
| Select Grain Data Model Object |
Required. Select the data model object that contains the data that the model trains on. The batch data transform uses this raw data to derive the fields required for training. Only the Loyalty Transaction Journal object is supported. |
| Select the Promotion | Required. Select the promotion that the template configuration generates personalized targets for. Then, enter a default prediction value. The system uses this value whenever the model predicts zero or a number below this default. |
| Select Loyalty Program and Non-Qualifying Currency | Select the loyalty program and non-qualifying currency associated with the promotion. If the promotion isn't associated with a loyalty program, skip this option. These values are only for information purposes and aren't used for training or prediction. |
| Select Products or Categories | Select the products or categories to predict targets for. |
| Select Customer Segments | Restrict training and prediction to members of up to five segments. Salesforce generates predictions only for members in the selected segments. |
| Set Training Duration | Required. Select how many months of historical transaction data the model must train on. We recommend that you use a training period long enough to capture seasonal patterns relevant to the promotion. |
| Set Aggregation Duration and Seasonality | Define a specific timeframe to capture seasonal behavior for predicting upcoming customer activity. For example, if you want to create a promotion for the month of December and want to capture seasonal behavior for predictions, select Monthly and December. |
| Filter Training and Scoring Data Model Objects | Apply filter conditions to the training and scoring datasets. You can apply the same filters to both datasets or define separate filters for each. The filtered training dataset requires at least 400 records for the model to train. |
Define Parameters
| Option | Details |
|---|---|
| Define Target Variable | Select whether the model must predict the purchase amount or quantity. If you don't select a value, the model predicts the purchase amount. |
| Select Features | Select up to 15 additional fields the model uses as prediction inputs. By default, the Scoring Framework trains the model on 5 default features: average transaction value, purchase transaction count, and the member's last three interval purchase measures. |
| Set Conditional Scores | Set the percentages of a member's predicted purchase amount or quantity that become their Easy, Moderate, and Hard targets. For example, if a member's predicted purchase amount is $100 and the percentages are set to 60%, 75%, and 90%, the resulting targets are $60, $75, and $90. |
Configure Debug Settings
Turn on Save Assets to retain app assets even if the installation fails. These assets are useful for troubleshooting.