Product recommendations

This article explains where you can use product recommendations and how the different recommendation models work.

What are product recommendations

Product recommendations help you surface the right products to the right customers at the right moment. Bloomreach uses customer behavior and catalog data to power AI-driven suggestions across your channels—boosting engagement, conversions, and revenue.

Where to use product recommendations

You can place product recommendations across multiple touchpoints in your customer experience:

• Homepage: Showcase top products or personalize content to reactivate users.
• Category detail: Highlight the most popular items within the category a customer is viewing.
• Product detail: Suggest alternatives to the product currently being viewed.
• Basket: Recommend relevant products alongside those already in the basket.
• Email campaigns: Personalize email marketing or showcase your store's product range.
• Weblayers and push notifications: Engage users with relevant product suggestions in real time.

Recommendation types

Bloomreach recommendations are distinguished by 2 factors: how the recommendation model is created, and how the model is requested.

How models are created

• Rule based: Uses statistical heuristics to generate recommendations quickly. Models are available to use right away. Examples include top-selling products and most-viewed products.
• Loomi AI powered: Uses machine learning algorithms trained on historical data. Training takes several hours. Examples include TF-IDF and matrix factorization models.

How models are requested

• Event-based: Recommendations are generated based on a customer's past events, such as purchases or product views.
• API call: Recommendations factor in a customer's most recent actions—for example, viewing a product or adding it to the cart—even if those actions haven't been loaded as events yet. This is handled through the Recommendations for customer API call.

For example, you can recommend alternative products for the item a customer is currently viewing on the product detail page.

Predefined templates

Bloomreach gives you access to AI-powered recommendation templates you can use out of the box. These templates are customizable—you can mix and match engines to meet your specific needs. All updates occur in real time across the platform.To create a new recommendation:

1. Go to Campaigns > Recommendations and click + New recommendation.
2. Select a template that fits your use case from the predefined options.
3. Build your engine using the chosen template.

Rule-based templates

• Filter based
• New items
• Popular right now
• Customer recent interactions
• Metric-based category
• More like this
• Advanced

Loomi AI templates

• Customers who bought this item also bought
• Customers who viewed this item also viewed
• Personalized recommendations for you
• Textual similarity
• Homepage
• Product details

Basic requirements

Before your recommendations can work correctly, make sure your catalog and event tracking are properly set up.

Catalogs

A catalog is the foundation of any recommendation engine. For recommendations to function correctly, ensure you have a clear item identifier, properly configured searchable fields, and mapped fields for product display.

Item identifier

Every catalog must have a unique item identifier that matches the ID used in your events, for example, view_item or purchase_item.

• Non-Data hub integrations: During the initial import, drag the item_id label into the column containing the unique item ID.
• Data hub integrations: The item identifier is configured in the Item collections Schema.

Searchable fields

Any field you want to use as a filter in your recommendations—for example, active or category— must be marked as searchable.

• Non-Data hub integrations: Set searchable fields during the initial import. You can't change them after the catalog has been imported. If you need to change searchable fields, you must replace the catalog.
• Data hub integrations: Set or update searchable fields in the Engagement destination Data & Assets > Catalogs > Schema tab at any time. There's no need to replace the catalog.

Field mapping

To display recommendation results in the single customer view and the test tab of each recommendation engine, map the fields: image, URL, title, and price.

How you map these depends on your integration:

• Non-Data hub integrations: Set mapping fields during the initial import.
• Data hub integrations: Map these fields using the system attributes in the Item collections Schema tab.Review the full guidance on data mapping.

Event tracking

Most recommendation engines rely on event tracking to understand customer behavior. Make sure the following events are tracked correctly:

• view_item — tracks which items customers view.
• cart_update — tracks which items are added or removed from the cart.
• purchase_item — tracks which items were purchased.For more information, see the Data manager article.

Data mapping

For all recommendations to work correctly, go to Data & Assets > Data manager > Mapping and confirm the following events are mapped:

• Purchase item
• Add to cart
• View item
• View category