The Parent/Child catalog feature in Predict allows you to handle two-tier product structure. Previously, only flat product structure was supported in Emarsys, but currently, we are working on a solution that enables you to handle product catalogs containing product variants.

The Parent/Child catalog feature comes in handy when your product catalog contains items that belong to a product variant group (e.g., color variants of the same kitchen appliance). In this case, it is important to ensure that the data collection process (i.e. catalog processing, Web behavior data collection) takes the parent-child structure into account. Simultaneously, recommendations also have to take the product groups into consideration:

• Predict has to work with item data on SKU level and has to avoid displaying a variation of the same product in the recommendations (e.g., when a customer views a size S shirt, then the Predict recommendation should not show the size L version of the same shirt).
• Revenue attribution data and the Click Through Rate (CTR) have to be collected on SKU level.
• ABANDONED CART recommendation: variants have to be displayed with their proper links (instead of the item group’s link of an abandoned product).

The Parent/Child catalog feature also meets the following requirement:

• Providing maximum compatibility with Smart Insight

If you would like to use the Parent/Child catalog feature, then you need to fulfill the following requirements:

1. Currently, the feature supports only two-tier/layer product structure (e.g., size variations in retail OR color variations in electronics). 

2. Your product data file and website structure have to meet the following requirements:

• Each Child product (SKU, variant) needs to have its own individual page on the website, including individual URLs that change dynamically when visitors click a variant on the UI.
• If you are using the Emarsys CSV format: the group_id column and parent item row (per product) have to be added to the product data file.
  • This feature also works with standalone items; just add the item as its own parent (that is basically a duplication, except the parent item will have the g/id and the child a regular one).

• If you are using the Google Product Feed (GPF) format: the item_group_id field has to be populated.

3. Please contact your Client Success Manager to make sure that you are aware of the capabilities and limitations of the Parent/Child catalog feature.

Predict handles parent-child products in recommendations by using a filtering mechanism to keep the recommender modeling logic intact, robust and configurable. The recommended item list may contain SKU-level items (children) and groups (parents). The filtering mechanism filters out serially all siblings of each item from the list to avoid displaying items from the same group in the recommendation which would lead to a less desirable shopping experience.

You can see how the recommendations work in the following table:

When enabling the Parent/Child catalog feature, Predict is configured to use a grouped catalog and will work as expected (for more information on how post-filtering is applied to recommendations, see Recommendations) except for a few limitations. For more information on limitations, please contact your Client Success Manager.

Known issues:

• It is possible that instead of the expected three column recommendation format, you will see only a single recommended product.
  This can be caused by:
  • A recently activated parent-child feature
  • Singificant changes in the product catalog or its reload

If the product catalog is uploaded in the correct format, Predict will resolve the issue in 1-2 days on its own.