Personalization

AI Product Recommendations: How They Work and Where to Use Them

Learn how AI product recommendations work, where to show them, how to measure results, and what to look for in a product recommendation engine.


AI product recommendations are product suggestions chosen by machine learning for each shopper. They're based on what the shopper has viewed, added to cart, and bought, and on what similar customers went on to buy. A product recommendation engine is the software that makes those picks.

Shoppers have more choice than ever, and too much choice stalls decisions. Good recommendations cut through it. They show each person the few products most likely to fit, which helps shoppers find what they want and raises average order value.

This guide covers:

  • How AI product recommenders work, including collaborative, content-based, and hybrid filtering
  • Where to show recommendations on your store
  • How to measure whether they're working
  • What to look for in a product recommendation engine
  • The rules that keep AI recommendations relevant and on brand

 

How Do AI Product Recommenders Work?

Product recommenders analyze user data to learn what items appeal to a specific customer.

An advanced product recommendation engine can create personalized product suggestions based on real-time customer behavior and integrate with various ecommerce platforms.

They collect first-party data (both from the current customer and their entire customer pool). This data includes what the customer's likes, how they use the website, and what similar customers like. Then, they calculate what will appeal most to each customer in the future.

Recommendation engines typically achieve results using one or both types of data filtering. These include content-based and collaborative filtering.

Collaborative filtering combines data from similar users

Collaborative filtering systems analyze data about customers’ individual preferences and pair that info with data from users with similar tastes. Think “customers who bought this item also bought…” recommendations.

Spotify’s popular Discover Weekly playlist uses collaborative filtering systems.

The platform collects data on what a user listens to, the playlists they create, and how often they stream the tracks they enjoy. Then, it compares the user’s habits with data from over 500 million other users to find similarities. Based on these patterns, Spotify creates a curated mixtape of 30 songs it thinks you’ll love.

And judging by the numbers, most listeners love their curated playlists. As of 2020, Spotify users had spent over 2.3 billion hours streaming their Discover Weekly playlists. A testament to how powerful collaborative filtering can be for enhancing the customer experience.

A screenshot from Spotify's Discover Weekly playlist, which is an example of an ai-powered product recommender at work

Improves product discovery

Collaborative filtering helps users discover new items they might not have otherwise found. By analyzing a customer's preferences and purchase history, collaborative filtering can suggest products that align with their tastes. Just think of how many times you’ve discovered a new song or artist on Spotify. That’s all thanks to the power of collaborative filtering.

This makes collaborative filtering an excellent tool to use for upselling and cross-selling. You can help customers discover new products or interests. Customers don’t even need to actively search. You can recommend items popular among users with similar tastes at any point in the customer experience.

Can be set up without any previous customer data

Collaborative filtering can suggest items to shoppers even without any data on their purchase history. This capability can enhance the shopping experience and lead to increased customer loyalty over time.

For example, you can show new visitors popular or trending products on your homepage or product pages through recommendation widgets. A widget that says, “Customers who bought this item also bought…” shows products that are often bought together.

Content-based Filtering Uses Previous Interactions With Items

Content-based filtering generates “If you liked this, you might also like that” recommendations. These systems use machine learning algorithms to analyze the content of items and make personalized suggestions.

This system provides personalized recommendations to users based on what they’ve shown interest in. It looks at the item’s details to compare against other items on your site. It then suggests other products based on similarities in the copy or code.

For example, Manssion, a designer of premium men’s jewelry, shows customers a “You may also like” suggestion on the cart page. This includes related products based on items in their shopping carts. Using this widget as part of their product personalization strategy has helped boost Manssion’s AOV by 18.65%

Screenshot of a men's jewelry brand online store, Manssion, which shows their cart with ai-powered product recommendations after a bracelet was added to cart and there are options for necklaces as well as a progress bar showing free shipping that is unlocked

Creates greater transparency

Around three-quarters of customers want companies to be clear about how they use their information. And content-based filtering is more transparent than other recommendation approaches.

Collaborative filtering relies on personal data or browsing history. But content-based filtering doesn’t require this. Instead, it analyzes the content of items directly, such as the products in your shopping cart.

This transparency makes content-based filtering easier to explain to users. Greater transparency can significantly improve customer satisfaction by making users feel more comfortable with the recommendations. If you previously searched for silver chains, you won’t be surprised to see recommendations for silver bracelets.

You don't need much data

When making suggestions to new users, you have limited data for effective product suggestions. But content-based filtering doesn’t need data from other users to create recommendations.

Unlike collaborative filtering, the recommendation engine still provides relevant recommendations even with limited user data.

Ecommerce Personalization: The Complete Guide for Founders, Merchants, and Marketers

Your ecommerce personalization journey starts here.👇

 

Hybrid Recommendation Systems Combine Both Methods

Hybrid recommendation systems use both content-based and collaborative filtering. This approach can make a significant impact on a business's ecommerce strategy by providing more accurate and diverse recommendations. The result is even more accurate personalized product recommendations.

Netflix, for example, uses a hybrid approach for its recommendation system. It’s so accurate that the platform says personalized recommendations from the engine drive 80% of viewer activity.

Many hybrid systems start with collaborative filtering. Say, to identify popular items among users with similar tastes. Then, content-based filtering further refines those recommendations based on product features.

Another option is to start with content-based filtering to identify items similar to those the user has interacted with. And then collaborative filtering recommends items that similar users have enjoyed.

Let’s take a look at how Netflix does it. The platform uses collaborative filtering to analyze data on user behavior and preferences. This includes viewing history, ratings, and searches. Then they identify groups of users with similar interests. Netflix then combines that data with a user’s viewing history and preferences to make recommendations.

The main page of the Netflix App showing their AI-powered product recommenders for their various movies

 

Generates more diverse and personalized recommendations

A hybrid recommendation system combines the strengths of collaborative filtering with content-based filtering. But it also overcomes the limitations of each approach. This makes hybrid systems more accurate than single-method systems.

Hybrid systems also recommend a broader range of items than single-method systems. This can lead to higher user engagement as customers are more likely to find products that interest them. For example, a hybrid system that uses collaborative filtering can recommend popular products. While a system that uses content-based filtering can recommend similar or complementary products.

In addition, combing these two methods across your web pages creates more opportunities to upsell and cross-sell. You can recommend popular products that also complement the user’s purchase.

Offers resilience to cold-start problems

A hybrid system is more resilient to cold-start problems. This resilience can improve customer satisfaction by providing relevant recommendations even to new users. That’s because hybrid systems can use one method to make recommendations for new users or items until there is enough data to use another method. A user who has never visited your site will see best sellers. The next time they return, you’ll have more data and can make more specific recommendations.

Where to Show AI Product Recommendations

Placement matters as much as the algorithm. Each part of your store suits a different kind of suggestion.

  • Homepage. Show bestsellers and trending items to new visitors, and personalized picks or "buy it again" to returning customers.
  • Product page. Show "frequently bought together" items and similar products while the shopper is still deciding. Rebuy Dynamic Bundles add these offers right on the product page.
  • Cart. Suggest low-cost add-ons that fit what's already there. The Rebuy Smart Cart shows personalized recommendations inside a slide-out cart.
  • Checkout. Offer one relevant extra. On Shopify, offers inside checkout need Shopify Plus, which is Shopify's rule. Rebuy Checkout Extensions add them there.
  • Post-purchase. Show a one-click offer after payment, based on what the customer just bought, with Rebuy Post-Purchase Offers.
  • Email and SMS. Include the next best product in order confirmations and follow-ups. Rebuy's Klaviyo integration adds personalized recommendations to your flows and campaigns.

Keep Recommendations Consistent Across Channels

Shoppers move between your site, email, and SMS, and recommendations should follow them. Use the same customer data and the same logic in every channel, and update it in real time. That way a shopper sees suggestions on your site that match the ones in their inbox, and nobody is emailed a product they bought yesterday.

How to Measure AI Product Recommendations

Track these numbers for each placement, not only for the store as a whole:

  • Click-through rate. The share of shoppers who click a recommendation. It tells you whether the suggestions are relevant.
  • Add-to-cart rate from recommendations. How often a click turns into an added item.
  • Attach rate. The share of orders that include a recommended product.
  • Average order value and revenue per visitor. Compare shoppers who saw recommendations with those who didn't.
  • Revenue from recommended products. The part of total sales that came through a recommendation.

The cleanest test is an A/B test: the same page with and without the recommendation block. Rebuy A/B Testing lets you test recommendation widgets against each other and see which one earns more.

What to Look for in a Product Recommendation Engine

  • It fits your platform. Choose a tool built for your ecommerce platform that installs without custom development.
  • It learns from the right data. Order history, cart contents, and browsing behavior, updated in real time.
  • It handles new shoppers and new products. A good engine falls back on bestsellers and product details until it has behavior data.
  • It covers every placement. Homepage, product page, cart, checkout, post-purchase, and email, from one tool and one set of data.
  • It lets you add rules. You should be able to pin, exclude, and prioritize products alongside the AI.
  • It includes testing and reporting, so you can see which recommendations earn revenue.
  • Its pricing fits your size. Many tools charge by traffic, orders, or revenue, so check how the cost grows as you do.
  • It keeps your store fast. Check page speed before and after you install it.

Rebuy is built for Shopify and Shopify Plus stores. Rebuy Cart & Merchandising uses AI to recommend products on your homepage, product pages, and cart, and lets you set your own rules wherever you want control.

Rules and Guardrails for AI Recommendations

AI picks well when you give it boundaries. Set a few rules before you go live:

  • Exclude what shouldn't be recommended: out-of-stock products, free gifts, and items already in the cart.
  • Protect your margin. Keep very low-margin products out of upsell and cross-sell slots.
  • Keep prices in proportion. An add-on should cost less than the item it's recommended with.
  • Stay on brand. Pin or feature products for launches and campaigns, and let the AI fill the rest.
  • Be clear about data. Tell customers how you use their information, and rely on data they've shared with you directly.
  • Review what's showing. Check your recommendations regularly, the way you'd check a store display.

AI Product Recommendation FAQs

What are AI product recommendations?

AI product recommendations are product suggestions chosen by machine learning for each shopper, based on their behavior, their purchase history, and what similar customers bought.

How does a product recommendation engine work?

A product recommendation engine collects data on products and shopper behavior, finds patterns using collaborative filtering, content-based filtering, or both, and shows each shopper the products they are most likely to want.

Do AI product recommendations work for small stores?

Yes. With little order history, a recommendation engine can lean on product details and bestsellers, then personalize more as it collects data.

What is the cold start problem in product recommendations?

The cold start problem is the difficulty of making recommendations for a new shopper or a new product that has no history. Hybrid systems handle it by using product details and popular items until behavior data exists.

Where should I show product recommendations on a Shopify store?

Start with the product page and the cart, then add the homepage, post-purchase offers, and email. Offers inside checkout require Shopify Plus, which is a Shopify rule.

Join the AI Revolution With Rebuy

The ecommerce landscape is competitive. Businesses that don’t add product recommenders to their ecommerce merchandising strategy risk falling behind. Product recommenders aren’t just a smart move for your ecommerce site.

They’re necessary for staying ahead of the curve to create a more personalized and satisfying user experience.

Dive deeper and learn more about how AI is revolutionizing ecommerce.

And then, when you’re ready to join the AI revolution, use our AI-powered personalization platform for Shopify and Shopify Plus. It serves up different types of product recommendations to your customers and helps you improve customer engagement and increase sales. Additionally, AI-powered recommendation engines can enhance the effectiveness of email marketing campaigns by providing personalized product suggestions.

 

•••

 

Try Rebuy free and see why the world’s top brands use Rebuy to accelerate sales growth.

Interested in partnering with Rebuy? Let's do it.

To keep up with the latest trends, updates, and more, follow us on LinkedIn.

Similar posts

Assorted products for BFCM shopping

Get the 2026 BFCM Guide

Free, actionable strategies to maximize revenue this BFCM