Documentation Commerce Product Recommendations

Recommendation Types

Last update: September 3, 2026

Topics:
Storefront

CREATED FOR:

Admin

Adobe Commerce provides a large set of recommendations that you can deploy to various pages on your site. All recommendation types are data-driven. Behavioral data, product attribute data, and metrics power them. For easy reference, recommendation types are grouped as follows:

Personalized
Cross-sells and up-sells
Popularity
High-performing

As a best practice, Adobe recommends the following guidelines when using recommendations:

Diversify your recommendation types. Customers start ignoring recommendations if they suggest the same products over and over again.
Do not deploy the same recommendations to your cart page and order confirmation page. Consider using Most Added to Cart for the cart page and Bought This, Bought That for the order confirmation page.
Maintain your site configuration. Do not deploy more than three recommendation units on the same page.
If your store sells clothing, the More like this recommendation can suggest gender-specific products that do not match the gender of the product being viewed. Consider using this recommendation type only for non-clothing categories.

NOTE

For more information about the events described in this article, see storefront events in the developer documentation.

Data requirements and behavior

Product Recommendations is a data-driven system that relies on behavioral data collected from your storefront. The quality and quantity of recommendations depend on the amount of event data available.

IMPORTANT

Most recommendation types require sufficient behavioral data (such as product views, add-to-cart actions, and purchases) to generate meaningful results. The system typically needs several days of active shopper activity to build accurate recommendations. To learn how site traffic helps to populate the various recommendation types, see readiness indicators.

What happens with insufficient data

When there is not enough event data to generate recommendations, the system can:

Return empty results for the recommendation unit.
Trigger backup recommendations, such as showing Most viewed products when personalized recommendations are not yet available.
Display fewer products than configured in the recommendation unit.

Personalized personalized

These recommendation types recommend products based on the specific shopper’s behavioral history on your site. For example, if a shopper previously browsed or purchased a jacket on your site, these recommendations continue from their previous activity and recommend other jackets or similar products.

NOTE

Personalized recommendations require shoppers to have an established behavioral history. New visitors or shoppers without sufficient interaction history see backup recommendations, such as Most viewed products until they generate enough behavioral signals on your site.

Type

Description

Recommended for you

Recommends products based on each shopper’s current and previous onsite behavior. Displays highly relevant recommendations based on the shopper’s browsing and purchase history. This recommendation type is effective on the home page where most shoppers begin their journey on a site. For first-time shoppers on your site who have not generated any signal to personalize their experience, Adobe Commerce shows products based on the Most viewed recommendation type. When the shopper starts to interact with the products on the site, however, recommended products adjust in real time to their behavior.

Where used:
- Home page
- Category
- Page Builder

Suggested labels:
- Just for you
- Recommended for you
- Inspired by your shopping trends

Cross-sells and up-sells crossup

These recommendation types are social-proof driven to help shoppers find what others liked. They are also product-driven to help them find other similar products. The recommended products often complement the selected product.

Page context for recommendations

Cross-sell and up-sell recommendation types use the context of the page where the recommendation unit appears. That context tells the system which product or products to use when finding related items:

Product detail page — Uses the SKU of the product the shopper is viewing.
Cart page — Uses the SKUs of products in the shopper’s cart.
Order confirmation page — Uses the SKUs from the purchase the shopper completed.

The same recommendation type can produce different results depending on where you deploy it. For example, Viewed this, viewed that on a product detail page uses the viewed product as context. On the cart page, it uses cart contents. On the order confirmation page, it uses the completed order.

NOTE

The “viewed this, viewed that,” “viewed this, bought that,” and “bought this, bought that” recommendation types use a sophisticated collaborative-filtering algorithm to identify interesting similarities without overemphasizing popular products. The algorithm uses shoppers’ aggregate behavior across multiple sessions on your site rather than behavior from a single in-session interaction. These recommendation types help shoppers discover related products that may not be obvious pairings with the product they’re viewing.

These recommendation types require substantial cross-product interaction data to identify meaningful correlations. Stores with limited product catalog diversity or low traffic see fewer recommendations until sufficient behavioral patterns emerge.

Type

Description

Viewed this, viewed that

Recommends products that shoppers view disproportionately more often with the currently viewed product.

Where used:
- Product detail
- Cart
- Confirmation

Suggested labels:
- Customers who viewed this product also viewed (PDP)

Viewed this, bought that

Recommends products that shoppers tend to buy disproportionately more often after viewing the current product. This type helps guide shoppers to discover products that they might not have otherwise noticed.

Where used:
- Product detail
- Cart
- Confirmation

Suggested labels:
- Customers who viewed this ultimately bought
- Customers ultimately purchased
- What do others buy after viewing this product?

Bought this, bought that

Recommends products that shoppers buy disproportionately more often with the currently viewed product. This type displays highly relevant products shoppers can add to their cart by aggregating what other shoppers have bought with those products.

Where used:
- Product detail
- Cart
- Confirmation

Suggested labels:
- Get everything that you need
- Don’t forget these
- Frequently bought together

Popularity popularity

These recommendation types recommend products that are the most popular or trending within the last seven days.

NOTE

Popularity-based recommendations require sufficient event data from your storefront. If your store is new or has low traffic, these recommendation types return limited results or no results until adequate behavioral data has been collected. To ensure optimal performance, monitor your data readiness indicator.

Type

Description

High performing highperf

These recommendation types recommend top performing products based on success criteria like add-to-cart or conversion rates.

NOTE

High-performing recommendation types rely on conversion data (purchases and add-to-cart actions). New stores or stores with low conversion volumes need to collect data over 7-14 days before these recommendations become effective.

Type

Description

View to purchase conversion

Recommends products with the highest view-to-purchase conversion rate. Of all the shopper sessions that registered a product view, what is the proportion that eventually registered a purchase by the shopper.

Where used:
- Home page
- Category
- Product detail
- Cart
- Confirmation

Suggested labels:
-Top sellers
- Popular products
- You might be interested in

View to cart conversion

Recommends products with the highest view-to-cart conversion rate. Of all the shopper sessions that registered a product view, what is the proportion that eventually registered an add to cart by the shopper.

Where used:
- Home page
- Category
- Product detail
- Cart
- Confirmation

Suggested labels:
- Top sellers
- Popular products
- You might be interested in

Visual similarity visualsim

The Visual similarity recommendation type recommends similar looking products to the product being viewed. This recommendation type is most useful where images and visual aspects of the products are important parts of the shopping experience.

NOTE

Because this recommendation type is not applicable to most catalogs, the system does not enable it by default. Explicitly enable this recommendation type.

How it works

The Visual similarity recommendation type offers recommendations for other products in your catalog that have a visual similarity to the imagery currently being viewed. Visual similarity includes aspects such as:

Color
Shape
Size
Texture
Material
Style

Adobe AI uses AI to process and analyze the imagery in your catalog and build attributes used to determine visual similarities.

Special considerations

If you are testing this recommendation type in a non-production environment, make sure your image URLs are publicly accessible.
Currently, product images must be 10 MB or less in size.
The optional Fastly Image Optimization module can apply Fastly Image Optimization parameters to Product Recommendations image URLs. See Add Fastly Image Optimization support for details.

Enable Visual similarity recommendation type

NOTE

The Visual similarity recommendation type is available when you install it as an optional module.

On the Admin sidebar, go to Marketing > Promotions > Product Recommendations to display the Product Recommendations dashboard.
Click Settings (gear icon) to display the Settings page.
In the Visual Recommendations section, select to Enable Visual Recommendations.
Click Save changes when you are finished.

The Create New Recommendation page now displays Visual similarity as a selectable recommendation type when the page type is Product Detail.