The world’s largest retailer as of 2019, Amazon is practically synonymous with ecommerce, with more product searches beginning there than on Google. And product recommendations are the secret of the ecommerce giant’s success, accounting for 35% of its revenue. What’s more, it was actually Amazon that popularized collaborative filtering.

You know, “People who bought this item also bought that item.”

Product recommendations are crucial to any retailer’s personalization strategy. But as brands are still getting to know their customers, collaborative filtering is helpful as it doesn’t require nearly as much customer data as you may expect.

Also known as “wisdom of the crowd” recommendations, collaborative filtering makes predictions about one customer’s interests based on the interests of many. When an algorithm detects the particular URL someone is viewing, it leverages the retailer’s larger site history. From there, retailers make product recommendations based on other customers’ behavior.

Contextual Collaborative Filtering

Contextual collaborative filtering needs just one data set: the URL a customer is looking at. These algorithms recommend the items most often viewed or purchased by people who viewed or purchased an item after visiting that same URL. They suggest items typically bought or viewed together, or as substitutes for other items that have already been viewed or purchased.

Given how great contextual recommendations are for discovery, they’re particularly effective on product pages. Look at this example from ShoeDazzle. Using collaborative filtering, the retailer lets customers know what other shoppers who have looked at this gladiator sandal have browsed and bought.

Browsing History

Browsing history-based algorithms also use collaborative filtering, suggesting items based on what customers with similar histories have viewed. These recommendations don’t require user-specific data and can be used with customers who have generated as few as two page views. However, they leverage the knowledge of a customer’s browsing history and become more effective as page views increase.

These recommendations are frequently seen in browse abandonment emails that recommend additional items, like this one from Express:

Purchase History

Algorithms based on purchase history look to customers with similar purchase histories. Because these recommendations are based on high-quality, completed purchase data, they often outperforms algorithms based on browse history.

These naturally fit with post-purchase messaging, where retailers frequently recommend complementary items. Thrive Market includes these recommendations in shipping confirmations, based on similar orders from other members.

Collaborative filtering is just the beginning. To learn more about different product recommendation algorithms and how to use them, click here to download Recommendations to Revenue.