If you’ve ever used an eCommerce site, you’ve probably interacted with a product recommendation engine. It’s a powerful AI-driven tool that helps analyze your purchase and interest patterns, as well as those of others of a similar profile, to identify products that may be of interest.

What data does the engine analyze?

Each shopper generates data “related to their search behavior and product preferences”. The tool gathers data while the user browses the online store, and the rest of the internet as well. In addition, the engine analyzes the products in cart or wishlist, also-viewed items and recent purchases. Using all this data, the engine crafts individual customer and product profiles with a custom algorithm.

With these customer profiles, the product recommendation engine can present purchase suggestions and contextual offers, on-site and via other marketing communication channels.

Read more about product recommendation engines here.

Why use a product recommendation engine in eCommerce?

AI-powered product recommendation results in a 26% higher average order value. Personalization of product recommendations results in twice the click-through rate and quadruple the conversion rate. AI-powered product recommendation drives 24% of all orders.

The product recommendation engine is probably the most important eCommerce tool. Customization and personalization of product listings can result in great improvement in engagement, interaction, conversion and order value.

How do you choose the best product recommendation engine?

There are three possible recommendation strategies: Collaborative filtering, Content-based filtering, and Hybrid recommendation. Collaborative filtering predicts the product to be recommended based on large volumes of data gathered from other shoppers. Content-based filtering creates a unique profile for each user based on their previous shopping and search history. Hybrid recommendation combines these two methods. Since the first two methods have their pros and cons, many online stores use a hybrid model.

“The advantage of content-based filtering is that even if you keep expanding your product inventory, the lack of historical data about the products does not affect the model’s performance. … Collaborative filtering gives you the advantage of understanding individual user patterns better,” says Aryng.

Since the right model is a hybrid one, the recommendation engine algorithm may need to be modified from time to time. Your team should be able to analyze, maintain and improve the algorithm as required based on the quality of the recommendations being delivered by the engine. Without the right updates, the model can become obsolete.

Ensure that your online store generates enough data to drive recommendations, including transaction value. With enough data, your data scientists can create, maintain, monitor and experiment with various models and algorithms to get the right results.

Which product recommendation engine should you select?

There are several parameters based on which your choice could be made. The eCommerce platform that you are using will restrict the options that are on offer, at least to some extent. While many product recommendation engines offer algorithms which learn and evolve, based on a hybrid model that can be modified to suit your needs, it’s important to understand the pros and cons of each before making the decision.

Our pick: Top 3 product recommendation engines

Adobe Commerce Product Recommendations (powered by Adobe Sensei)

Adobe Commerce Product Recommendations is deployed as a SaaS-based solution, powered by Adobe Sensei intelligence services. The storefront contains “the event collector and recommendations layout template”, while the backend includes “Data Services, SaaS Export module, and the Admin UI”.

It is the recommended tool to be used with Adobe Commerce (formerly Magento Commerce). Adobe Commerce Product Recommendations uses AI/ML algorithms to “perform a deep analysis of aggregated shopper data”, which, in combination with the product catalog listed on your Adobe Commerce store, can deliver a personalized experience.

This product can be integrated seamlessly with Adobe Commerce storefronts implemented using PWA Studios or custom tech like React or Vue JS. Behavioral data (such as product views, items added to cart, purchases) is collected on the online store, as soon as the module is installed and configured. This is correlated with catalog data such as product name, price and availability, to “calculate product associations” and deliver suggestions for each recommendation type.


  • Part of Adobe Suite; Easy implementation on Adobe Commerce platform
  • Completely customizable algorithm
  • Can integrate “existing data” with Adobe Analytics for better reporting
  • Works for complex audiences and across channels



Ideal for existing Adobe suite customers and for enterprise-grade customers. However, not suitable for those on a limited budget or for those who are not on Adobe Commerce already.


Optimizely makes recommendations based on “AI-generated insights”. The content delivered to the user is updated in real-time to increase conversions and revenue generation. Optimizely has strong automated audit options, which help you understand the right content topics for your online store, without needing to perform manual audits.

Their real-time big data analysis allows you to understand which materials drive conversion without delay. Using “deep insight machinery” and “industry-leading natural language processing”, your entire library can be tagged automatically and efficiently. This allows an easier understanding of content and product categories. When multiple possible recommendations are appropriate, Optimizely can prioritize which is better suited.

Optimizely supports email and web product recommendations with real-time personalization. Optimizely also allows you to customize triggers to upsell or cross-sell products, or to “win back” customers who have abandoned cart or not purchased a product previously viewed.


  • Considered the industry standard
  • Testing of multiple algorithms, with stats engine to analyze the “possibility of a variation leading to more conversions”
  • Strong security measures to work with multiple collaborators


  • Time- and effort-intensive testing
  • Despite existence of basic setup test wizard, tool needs heavy development skills for key functionality
  • Large amounts of data needed to make the tool effective


One of the market leaders, but expensive and tech-heavy. Suitable for “advanced testers at the enterprise level”. Requires large amounts of data and traffic for results.


In early 2020, Salesforce acquired Evergage, a real-time CDP (customer data platform) and personalization provider. Evergage is therefore intended to work as a part of the Salesforce suite of products, and works well with Salesforce Commerce Cloud (SFCC).

Leveraging Evergage’s machine learning algorithms, you can “deploy 1:1 personalized experiences on multiple channels”. As a CDP, Evergage tracks behavior to segment users, either manually or in real-time, based on input data and creates user profiles based on these signals. Each visitor segment can be targeted with specific multi-channel campaigns and messaging.


  • Named a leader by Forrester and Gartner
  • Integrates with Salesforce Marketing Cloud to improve relevance at all touchpoints
  • Search and navigation personalization
  • Wide range of features including A/B testing and multivariate testing
  • Multi-channel personalization for a unified experience, based on the detailed profiles


  • Less breadth of abilities as compared to competition: Forrester
  • Best suited for experienced marketers, with a steep learning curve
  • Visual editor hard to use, experience in HTML/CSS necessary
Frequently asked questions

What makes a good recommendation engine?

Choose one that can be completely customized, can be integrated well with your eCommerce platform, and with which your team has some experience.

When should you modify the algorithm?

If the product recommendation engine always recommends the same products or product categories, it probably needs modification. Recommendations should be personalized to the shopper and to the category or product being browsed.

How do you measure performance?

Measure performance of the engine through its actual results on the online store. Through A/B testing, you can compare actual financial results, between recommended products and others, on the following parameters:

  • click-through rate
  • conversion rate
  • increased transaction value or number of items per purchase

Test its impact on user behavior

  • interaction with emailers/web banners/online ads
  • recurring purchases
  • time spent on the website


Ideal for those needing a CDP as well as personalization. Works well with SFCC. Takes time to understand the nuances of the tool and is expensive. Not suitable for smaller businesses with low data maturity level.

While each product recommendation engine has its pros and cons, just as eCommerce platforms themselves do, Adobe Commerce Product Recommendations (powered by Adobe Sensei), Optimizely and Evergage are some of the leaders in the industry. Whichever you choose, you’re likely to need support from a trusted tech partner to integrate the chosen engine into your online store.

Talk to our Integrations team for help making the right choice and implementing it as part of your eCommerce site.