Evaluation and Testing Recommendations
Test tab in the user interface
Once the model was trained and it's active, you can switch from the Design tab to the Test tab to see product recommendations generated by the engine. In the image below you can see the test tab in the Bloomreach Engagement platform.
In the top left corner of the test tab view you can switch between visual and list mode to see either product visual or product catalog attributes (in the product catalog column there must be column with the name image and it is necessary to label catalog columns during the catalog import in order to see images, product titles, URLs - these labels are image, title, and URL).
The right panel of the test tab is used for testing user-based personalized recommendations. You can select a preview of the product recommendations for an arbitrary customer in the Bloomreach Engagement platform. For easier retrieval of customers matching some conditions, there's also a standard customer filter available. In case you want to generate product recommendations in the context of item/product from a product catalog with the option “Based on item”. See the image below for an example.
In the case of Metric-based category you have to also fill in category names/ids that correspond to the category identifier specified in template settings. For more information see product discovery paper.
The last information in the right panel is brief summary and ratio of personalized/non-personalized products from the number of requested items. In the test tab it is always 10, e.g. 4/10 means 4 personalized products from 10 requested. You should use this information to check if the recommendations engine is running and personalizing in the majority of the cases.
Single Customer View
Data & Assets →
Profile of a selected customer
The suggested way to test product recommendation engines in the Bloomreach Engagement platform is to use the Single customer view because it is possible to see recommendations in the context of customer’s last seen products or last purchased products. In the Overview tab in the bottom right part of the display, you will see “Recommendations for ”. You can use this selector to see product recommendations generated from any engine previously created and trained.
Single customer view display
To correctly display the Single customer view, it is necessary to label the product catalog columns during import.
The preview searches for the properties
picture. If neither is found as an exact match, it searches for any property, including an image or picture in its name:
The same applies to
namefor a name, and
urlfor a link.
Updated 6 months ago