Configure Recipes to Map Your Data with the Consumer Goods Schema (Optional)
Salesforce provides several customizable out-of-the-box recipes.
Required Editions
| Available in: Professional, Enterprise, and Unlimited editions where Consumer Goods Cloud is enabled. |
You can update the default recipes if these scenarios occur.
- If your schema deviates from the Consumer Goods (CG) schema.
- When there’s a change from an existing entity to a custom entity.
- When the data doesn’t load properly.
- An app stopped working due to incorrect data values.
Out-of-the-box Recipes
| Recipe Name | Description | Output |
|---|---|---|
| POS Recipe | Loads data from an external connection into a dataset that is consumed to generate predictions. | POS dataset |
| Training Calendar Recipe | Generates a dataset that is related to the calendar, such as year week, week start date, and end date. These dates are within the training window range from the current date to the previous six months. |
|
| Training Input Recipe | Generates a dataset by using the output of calendar recipes to filter visits and promotions. It creates output datasets filtered for the past six months. |
|
| Training Feature Set Recipe | Performs all aggregations, computes the lead on a dataset from the previous recipe’s datasets, and creates a join of that data with the product, store, and week datasets. | De-normalized dataset for training the model |
| Scoring Calendar Recipe | Generates a dataset that is related to the calendar, such as year week, week start date, and end date. The dates are within the training window range including the past three months and the next three months. | Calendar dataset for scoring window |
| Scoring Input Recipe | Generates a dataset by using the output of calendar recipes to filter visits and promotions. It creates output datasets filtered for the past three months and the next three months. |
|
| Scoring Feature Set Recipe | Performs all aggregations, computes the lead or lag on a dataset from the previous recipe’s datasets, and creates a left-join of that data with the product, store, and week datasets. | De-normalized dataset for scoring the model |
Scenario 1: When the Data Is From a Different Object
You have an object that contains data on visits, retail stores, and product promotions. The CG schema requires that the data of each of these parameters be stored in separate objects. To map such an object to the CG schema, edit the training input recipe and then add a load node to import the data. To match the object with the CG schema-required fields, perform either the join, filter, or transform operation.
Scenario 2 - When the Data Is From the Same Object but Contains Different Column Names
The product priority column in the store object is named customer_priority. This column name is different from the default name (priority) that is provided in the CG schema. To ensure that this object complies with the CG schema, edit the scoring input recipe and add a transform node to rename customer_priority as priority.
