Predictive Sort

Predictive Sort enables you to deliver more tailored and relevant search results to shoppers via the predictive intelligence powered by Commerce Cloud Einstein. This capability is available for the following types of search:

Predictive Sort combines the existing Commerce Cloud B2C Commerce Sorting Rule approach with Einstein to calculate the affinity of an individual customer to your products via the following three data sources:

How Predictive Sort Works

The B2C Commerce Activity Tracking tracks shoppers. Based on a shopper's behavior and affinity to products, a cookie is saved on their device that includes their assignment to specific products. This assignment can change within a session and works for registered as well as for anonymous users. As data is collected, Predictive Sort learns about each shopper and increases the personalized sorting experience for them.

See Configuring Predictive Sort.

Performance Considerations

Personalization services can impact your page performance, though it shouldn't affect the performance of the search-show pipeline. Predictive Sort, however, has caching considerations for the rendering templates and search requests.

Important: B2C Commerce strongly recommends that you follow the Performance Best Practices.

Rendering Templates

Because every shopper sees a personalized sort of their search result set (search hits), caching the position of the search hits on the product grid isn't possible. B2C Commerce automatically overrides the related caching tag defined in the rendering template. There is no need for you to modify the caching setup in the rendering temples.

The caching of search hit positions in the grid is disabled by default only if requests include a sorting rule that includes the Predictive Sort sorting attribute. This means that non-personalized sorting still uses the default caching configuration.

For example, if you assign the Predictive Sorting Rule as the default sorting rule to only the Sales Category, only caching for the Sales Category Product Listing Page is disabled. If you configure an A/B test with 5% traffic to the Predictive Sorting Rule, caching is only disabled for these 5% of requests.

Search Requests

To use Predictive Sort, you must modify your ISML templates to use the iscache if attribute. By changing the template, search requests that using a sorting rule with Predictive Sort are no longer cached.