Product Recommendations uses both behavioral and catalog data:
Behavioral - Data from a shopper’s engagement on your site, such as product views, items added to a cart, and purchases. Adobe Commerce and Adobe Sensei do not collect personally identifiable information.
Catalog - Product metadata, such as name, price, and availability.
When you install the
magento/product-recommendations module, Adobe Sensei aggregates the behavioral and catalog data and creates Product Recommendations for each recommendation type. The Product Recommendations service then deploys those recommendations to your storefront. To help you implement product recommendations on your storefront, use the following workflow:
Deploy data collection to production
Deploying Product Recommendations requires two main data sources: catalog and behavioral. Because production is the only environment where your shoppers’ actions are captured and analyzed, it is in your best interest to start data collection on production as early as possible. Learn how Adobe Sensei trains machine learning models that results in higher-quality recommendations. As an added benefit, when you start collecting behavioral data on production, you can fetch recommendations based on this production data while operating in non-production environments. You can then test and experiment with different recommendations that are computed based on real shopper data collected in production.
Deploying data collection does not change your storefront’s appearance or your shoppers’ experience. Only creating and deploying recommendation units alters the customer experience on your storefront. Make sure you test on your non-production environment before deploying to production. Also, do not create recommendation units until you customize your template. See the next step.
Customize the template to match your style
Your storefront represents your brand, so make sure you modify the product recommendations template to match your site theme.
By customizing the template, you can specify your stylesheet, overwrite where a recommendation unit appears on a page, and so on.
See Customize in the developer documentation to learn how to complete this step.
Test recommendations on your non-production environment
It is always a best practice to test a new technology on your non-production environment before you deploy to production. Testing recommendations on your non-production environment allows you to play with different recommendation unit types, positioning, and pages. You can pull recommendations based on behavioral data already gathered on production while testing in your non-production environment, so that recommendation results are based on the shopping behavior of actual customers.
Ensure your non-production environment catalog is largely the same as the one that you have in production. Using similar catalogs ensures that the products returned in the recommendation units closely mimic the products on production.
See Fetch behavioral data from your production environment to learn how to complete this step.
Create and deploy recommendations to your production storefront
Now that you have deployed the behavioral data collection in production, modified the product recommendations template, and tested recommendations using actual shopper behavior, you are ready to promote all code to production and create live product recommendations.