Table of Contents

While talking about e-commerce recommendation systems (or recommender systems), we often hear mention of collaborative filtering (CF). It’s one of the most popular and common ways to do product recommendations online at present. However, many people don’t know much about it

So, in this article, we’re going to give you a comprehensive introduction to collaborative filtering. Specifically, we’ll clarify:

  • The definition of collaborative filtering and its underlying idea
  • Famous examples of using collaborative filtering
  • Advantages and disadvantages of collaborative filtering

Let’s get started! 😉

What is collaborative filtering?

As a popular approach to e-commerce product recommendations, collaborative filtering is a technique that can identify similarities between customers on the basis of their site interactions and then recommend relevant products to customers across digital properties.

Wikipedia gave another explanation by disassembling the word 💡:


The underlying idea of collaborative filtering is that you will probably like the things people with similar purchasing or browsing habits also like. This idea was popularized after Koren, Bell, and Volinsky published their research paper in 2009 📚:

Screen Shot 2022-02-09 at 09.30.10

It's also important to note that collaborative filtering does NOT use features of the item to do recommendations. Instead, this method uses consumer behaviors to classify consumers into clusters that share similar browsing or purchasing habits, and then it recommends things to each consumer according to the preference of their cluster. Some experts also consider CF as a new wave of customer segmentation, since the core strategy is based on clustering customers of similar tastes and behaviors together.

Examples of collaborative filtering applications

Today, collaborative filtering is in widespread use across different industries, and the two most famous examples of CF applications are Amazon and Netflix.

Amazon uses CF to match products to customers based on their past purchases. While the world was focusing on user-based collaborative filtering, Amazon came up with a new algorithm called item-based (or item-to-item) collaborative filtering in 1998. With this item-based collaborative filtering algorithm, product recommendations are created not just based on similarities between customers but also on correlations between products/items.

Netflix is known for using collaborative filtering to build its movie recommendation system 🎬. Here’s an example of how Netflix applies used-based CF to make a movie recommendation (Movie C) to User A:


Advantages and disadvantages of collaborative filtering

The primary advantage of collaborative filtering is that shoppers can get broader exposure to many different products, which creates possibilities to encourage shoppers towards continual purchases of products 🛍️.

Another advantage of this method, as above-mentioned, is that while exploiting the correlations behind user-item interactions, it does not require an “in-detail” of the products and extensive cataloging of these product data.

However, there’re also several disadvantages with this approach:

Disadvantage #1: Data Sparsity and cold-start problem. Data sparsity is seen as a key disadvantage of collaborative filtering. Typically, it can cause the cold start problem that describes the difficulty of CF making recommendations when the users or the products are new. Why so? The operation of CF is based on historical data of site interactions between users and items. But new users and items simply do not have enough historical data (data sparsity) to make it work. Consequently, due to this limitation of CF, you might fail to provide new customers with a delightful, personalized shopping experience and struggle to leave them with a good impression of your brand. 😢

Disadvantage #2: Scalability. Traditional CF algorithms often suffer serious scalability problems. As the number of users increases and the amount of data expands, collaborative algorithms will begin to suffer a decrease in performances simply due to the sheer increase in data volume.

The figure below illustrates the dilemma created by the above two limitations of collaborative filtering:


Disadvantage #: Synonyms. Collaborative filtering is unable to identify synonyms. Here, "synonyms" refer to similar items labeled or named differently. Collaborative filtering is unable to discover the latent association between synonyms, so it will treat these products differently. For instance, CF can't figure out that the seemingly different items “Backpack” and “Knapsack” are actually referring to the same item.


Disadvantage #4: Diversity and the long tail. In principle, collaborative filters are expected to enhance diversity as they help consumers discover more products, but some may unintentionally do the complete opposite 😨. Because this approach recommends products based on historical ratings or sales, it will not recommend products with little or limited historical data. The fact of more users viewing and buying a popular product will make it even more popular and make new items sit in the shadow behind these best-selling items. In short, this approach can create a rich-get-richer effect for popular products, resulting in a lack of diversity.

A better solution?

You might consider onboarding a product recommendation engine to give your customers a broader exposure to different products and enhance their product discovery experience on your e-commerce site. Yet, the limitations of collaborative filtering are pretty problematic. So, is there a better solution that embraces all the advantages while overcoming the drawbacks of collaborative filtering?

Yes, there is a more advanced AI-powered solution that can do so ✅, and it is called ClerkCore! Differing from collaborative filtering, ClerkCore overcomes the cold-start problem and the lack of diversity. Moreover, ClerkCore can identify synonyms and get up and running from day one! 🚀

As an e-commerce business owner, do you want to enhance customer experience, increase sales and generate more revenue? ClerkCore can help you achieve these via an all-in-one personalization solution that provides smarter search results, more relevant product recommendations, more targeted email campaigns, and better customer segmentations. While helping grow your business, ClerkCore also provides superior efficiency and automation, which will save your time and energy 💪.

Wonder how you can leverage ClerkCore, a better personalization solution than collaborative filtering, to benefit your e-commerce business? Talk to one of our talented experts today!

email icon in red

Keep up with the latest e-commerce trends, best practices, special events, and more!

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.