Recommendations Experience Configuration Guide

Recommendations are a common way to engage shoppers and drive them deeper through the conversion funnel. However, knowing what strategies to deploy and where can be difficult.

These examples demonstrate how you can deploy recommendations throughout the customer journey. In some instances there are multiple ways to configure a strategy, which means you have an opportunity to test these configurations against one another or even leverage machine learning to automatically optimize decisions.

As you expand and optimize your recommendations program, check the Recommendations Overview dashboard in the platform to see how the recommendations are impacting your site's performance.

Home Page/Landing Page

On high-traffic landing pages, such as your site's home page, you want to quickly engage customers with relevant products so that you can reduce bounce rates and drive them deeper into the conversion funnel.

Popular Products

This strategy is useful with new visitors or for websites with top-trending products that frequently change. It leverages the wisdom of the crowd to show products that are appealing to other visitors.

Recommendation Algorithm

  • Top Selling by Purchase Count
  • Top Selling by Gross Revenue
  • Most Viewed (Product Detail Page)
  • Trending Items by Purchase Count (premium option)

Historical Data

Use behavioral data—such as products browsed, items placed in the cart but abandoned, or products purchased—to recommend products unique to a customer's interests.

Recommendation Algorithm

  • Viewed and Also Viewed
  • Viewed and Later Purchased
  • Purchased and Also Purchased
  • Engagement Optimized

The Engagement Optimized recommendation algorithm doesn't require a lookback period.

Base Recommendation on

  • Last item viewed
  • Last carted item(s)
  • Last item purchased

The Last item viewed recommendation basis is the optimal choice because product view events occur more frequently than a cart event or purchase event. Therefore, using this basis allows the strategy to reach more customers.

Machine Learning

Curate a list of products that are relevant for your top customer segments, and then plug those recommendations into an Automated Personalization experience, thus allowing the machine to determine which ones are most relevant to each individual.

Recommendation Algorithm

Onboarded Recommendation Dataset

Base Recommendation on

No Lookup Key — Show All Items in Dataset

Dataset Configuration

Create a list of the top 3 to 10 relevant products for each customer segment, and then upload it via DATASETS in the top navigation bar

Alternately, you can select a different recommendation algorithm and then add specific products to the Pinned Products field that are the top relevant products for each customer segment within the recommendations action. This method can be an easier way to manage this list.

Experience Configuration

Select an Automated Personalization experience, and then create a variant for each customer segment. Add the recommendations action to each variant, and then choose a segment-specific strategy for each variant. Here are some example segments:

  • Gender — Men, Women, and Kids, which you could replace with Boys and Girls segments or have a Families variant
  • Spend — Affluent Shoppers (high-spending customers), Deal Seekers (low-spenders customers)
  • Activity Type — Activity-based personas, such as Climbers, Cyclists, and Hikers for an outdoors equipment retailer

Index Page

On an index page, also called a product listing page, customers indicate an interest in a particular group of products. Although they're already browsing a list of products, recommendations on an index page can be an opportunity to offer additional suggestions or to cross-sell related categories.

Page Context

You can infer the attributes about the products displayed on an index page and thus recommend other products matching those attributes

Recommendation Algorithm

  • Viewed and Also Viewed
  • Viewed and Later Purchased
  • Purchased and Also Purchased
  • Items Frequently Bought Together (premium option)

Base Recommendation on

Viewed first item on the page, which serves as context to better understand the type of page the visitor is on

Recommendation Filters

Using dynamic values with certain recommendation filter attributes better ensures the recommendations action only shows products with similar attributes to what's on the index page, such as the following:

  • Gender = Gender of Product Currently Being Viewed — Pairing this attribute and dynamic value creates a filter that returns more broadly related results
  • Product Type = Product Type of Product Currently Being Viewed — Pairing this attribute and dynamic value creates a filter that only returns other products in the same category

Popular Products

Highlight top trending products for the category of the index page.

Recommendation Algorithm

  • Most Viewed (Product Detail Page)
  • Top Selling by Purchase Count
  • Top Selling by Gross Revenue
  • Trending Items by Purchase Count (premium option)

Recommendation Filters

Add a filter that uses a static value for the category of the page on which you're implementing this strategy.

Search Page

On a search page, customers indicate that they're looking for some specific product or product type. You can augment the search results with other products to give the visitor more options.

Page Context

You can infer what products the customer is looking for on a search page and recommend other products matching those attributes.

Recommendation Algorithm

  • Viewed and Also Viewed
  • Viewed and Later Purchased
  • Purchased and Also Purchased

Base Recommendation on

Viewed first item on the page, which serves as context to better understand the type of page the user is on

Recommendation Filters

Using dynamic values with certain recommendation filter attributes better ensures the recommendation action only shows products with similar attributes to what's on the index page, such as the following:

  • Gender = Gender of Product Currently Being Viewed — Pairing this attribute and dynamic value creates a filter that returns more broadly related results
  • Product Type = Product Type of Product Currently Being Viewed — Pairing this attribute and dynamic value creates a filter that only returns other products in the same category

To cross-sell products, apply a recommendation filter that uses a dynamic filter that excludes (does not equal) products that match the filter attribute. For example, if you exclude products from the same category, you then recommend complementary products.

Recently Viewed

Customers searching may want to look back at other products they were browsing. Give them easy access to their browsing history with the Recently Viewed recommendation algorithm.

'No Search Results' Page

When a customer's search yields no results, you can present recommendations to engage them with other products they may be interested in or guide them back to products they've already shown interest in.

Trending Products

When a visitor's product search doesn't return results, you can leverage the wisdom of the crowd to show products that are appealing to your site's other visitors.

Recommendation Algorithm

  • Viewed and Also Viewed
  • Viewed and Later Purchased
  • Purchased and Also Purchased

Base Recommendation on

  • Last item viewed
  • Last item purchased
  • Last carted item(s)

The Last item viewed recommendation basis is the optimal choice because product view events occur more frequently than a cart event or purchase event. Therefore, using this basis allows the strategy to reach more customers.

Historical Data

If the visitor's search doesn't return any results, use their past behavior to show products that they may still be interested in.

Recommendation Algorithm

  • Recently Viewed
  • Engagement Optimized

Recently Viewed

Customers who don't get search results may want to look back at other products they were browsing. Give them easy access to their browsing history with the Recently Viewed recommendation algorithm.

Product Detail Page

A visitor has shown clear intent by viewing a specific product detail page. This event presents an opportunity to help them find other similar products they might like or to even cross-sell products from related categories.

Similar Products

Use recommendation algorithms and filters with dynamic values to display products that are similar to what the customer is currently viewing.

Recommendation Algorithm

  • Viewed and Also Viewed
  • Viewed and Later Purchased
  • Purchased and Also Purchased

Base Recommendation on

First item on the page

Recommendation Filters

Using dynamic values with certain recommendation filter attributes better ensures the recommendations action only shows products with similar attributes to the product the customer is currently viewing. Here are some examples of filters:

  • Gender = Gender of Product Currently Being Viewed — Pairing this attribute and dynamic value creates a filter that returns more broadly related results
  • Product Type = Product Type of Product Currently Being Viewed — Pairing this attribute and dynamic value creates a filter that only returns other products in the same category

Cross-Sell Products

Use recommendation algorithms and filters with dynamic values to display products from different categories that are related to what the customer is currently viewing.

Recommendation Algorithm

  • Viewed and Also Viewed
  • Viewed and Later Purchased
  • Purchased and Also Purchased
  • Items Frequently Bought Together (premium option)

Base Recommendation on

First item on the page

Recommendation Filters

Apply a recommendation filter that uses the Product Type attribute that excludes (does not equal) the Product Type of Product Currently Being Viewed dynamic value. This filter shows other products associated with the product the customer is viewing but not ones from the same category as the product.

Recommendations Dataset

Your company's merchandising team may have curated recommendations for products they believe go together, and its data science team may have developed proprietary algorithms. Fortunately, the platform's recommendations component lets you harness this expertise via onboarded recommendations datasets.

Recommendation Algorithm

Onboarded Recommendation Dataset

Base Recommendation on

First item on the page

Cart Page

On the cart page customers typically review and complete their purchases. This page can be a good opportunity to recommend other products to fill out the purchase or even remind customers of other products they've been browsing.

Products Also Purchased

When a customer is on the cart page, show them other products frequently bought with the product(s) they are about to purchase. These recommendations are often presented under the banner "People Also Bought" or some similar text.

Recommendation Algorithm

  • Viewed and Later Purchased
  • Purchased and Also Purchased
  • Items Frequently Bought Together (premium option)

Base Recommendation on

First item on the page

Recommendation Filters

Apply a recommendation filter that uses the Item Group ID attribute that excludes (does not equal) the Item Group ID of Cart Items dynamic value.

Upsell to Free Shipping Threshold

If your site has a free shipping threshold, consider recommending products with a price point below that threshold so that customers can better find options to add to their cart to reach the threshold.

Recommendation Algorithm

  • Top Selling by Purchase Count
  • Top Selling by Gross Revenue
  • Most Viewed (Product Detail Page)
  • Viewed and Also Viewed
  • Viewed and Later Purchased
  • Purchased and Also Purchased
  • Items Frequently Bought Together (premium option)
  • Newest (premium option)
  • Similar Items (premium option)

Base Recommendation on

First item on the page

Recommendation Filters

Apply a recommendation filter that uses the Price attribute that is less than your site's free shipping threshold, which you input as a static value.

Alternate Approach

Customers often need to add just a little more to their cart to qualify for free shipping. Finding those inexpensive but useful add-on items—a pair of socks, some batteries, some cheap sunglasses—can be frustrating for shoppers aiming to finish their purchases. Consider onboarding a list of these small add-ons as a recommendations dataset.

Recently Viewed

When a customer is on the cart page, showing what they've browsed may remind them of other products they intended to purchase. Give them easy access to their browsing history with the Recently Viewed recommendation algorithm.

Purchase Confirmation Page

After a customer completes a purchase, often they then end their visit to your site. However, you still have the opportunity to engage them with personalized recommendations that can encourage an additional purchase or future return visit.

Recommend Products Also Purchased

Show customers other products frequently bought with the product(s) they just purchased. These recommendations are often presented under the banner "People Also Bought" or some similar text.

Recommendation Algorithm

  • Viewed and Also Purchased
  • Purchased and Also Purchased
  • Items Frequently Bought Together (premium option)

Base Recommendation on

First item on the page

Recommendation Filters

Apply a recommendation filter that uses the Item Group ID attribute that excludes (does not equal) the Item Group ID of Cart Items dynamic value.