This page describes how to add, delete, and list placements, as well as review statistics about where recommendations appear. You add placements to use an existing recommendation model in a different location in your site.
If you want to create a new recommendation model, see Creating recommendation models.
Creating a placement
You add a new placement by using the Placements page in the Google Cloud console.
To create a new placement:
Go to the Recommendations AI Placements page in the Google Cloud console.
Go to the Recommendations AI Placements pageClick add_boxCreate placement.
Choose the recommendation model you want this placement to provide recommendations for.
Provide a name for your placement.
The name must be 1024 characters or less, and can contain only alphanumeric characters, underscores, hyphens, and spaces.
(Optional) If needed, update the ID.
The ID is generated from the name you provide, and must be unique across your project. It must be 50 characters or less, and cannot contain spaces.
Click Create to create the new placement.
It might take a few minutes for the placement to be available for use in prediction requests.
When the placement is available, it appears in your Placements page.
(Optional) On the Placements page, click on a placement name and provide a visitor ID to preview the products that would be predicted for that user.
Prediction preview can help you to confirm that your model is performing as you expect. Learn more.
Listing recommendation placements
You can see all of your recommendation placements on the Placements page.
Deleting a recommendation placement
To delete a recommendation placement:
Go to the Recommendations AI Placements page in the Google Cloud console.
Go to the Recommendations AI Placements pageClick the placement you want to delete to open its details page.
Click deleteDelete in the button bar at the top of the page.
If the placement is considered active, you must retype its ID and click Confirm to complete the deletion.
Reviewing recommendation placement statistics
You can see how many recommendations have appeared at each placement in the Placements table on the Placements page.
You can choose a time period of one day, one week, and one month, or you can enter a custom time range.
Usage count shows how many predictions were made for this placement, and Error count shows how many prediction errors happened for this placement. The % of unjoined events indicates how often the predict request contained a user event with a product ID that is not in the current product catalog.
Next steps
- Request predictions from your new placement.