What are the benefits of collaborative filtering for product recommendations? Are product discoveries based on "user preferences" a good fit for Gen Z?

Many people have experienced being recommended other products while shopping online, with messages like "Customers who bought this item also bought..." While it's become common for e-commerce sites to incorporate such recommendation algorithms, one method gaining particular attention in recent years is "collaborative filtering." How does this differ from traditional recommendation features? And will such techniques resonate with the younger generation, who will form the core of future consumption? This article explains the fundamentals of collaborative filtering while posing the question: "Can Collaborative Filtering Product Recommendations Capture the Hearts of Generation Z?" We explore the impact collaborative filtering will have on Generation Z, who hold the key to future consumption trends.

Collaborative Filtering: Creating Unexpected Product Discoveries

In the past, "content-based" algorithms were the mainstream for e-commerce sites. Content-based recommendations suggest highly similar products based on an individual's search and purchase history. For example, recommending other products from the same brand to someone who bought Brand A's product, or suggesting Clothing B with similar characteristics to someone who searched using keywords like "summer wear," "low price," "for men" and bought Clothing A.

One drawback of this approach is that it often displays only products within similar categories, which can be less engaging for users. Furthermore, relying solely on an individual's purchase history limits the variety of products suggested, potentially resulting in a list dominated by items users already know or have seen. Collaborative filtering addresses these weaknesses by utilizing the history of multiple users rather than just one individual.

Broadly speaking, collaborative filtering can be categorized into two approaches: "user-based" and "item-based."

・User-based
Recommends items likely preferred by specific users, such as "Users like you also bought..." For example, if user a bought products A, B, and C, and user b bought products A and B, their purchase information is similar, suggesting similar preferences. Therefore, product C would also be displayed as a recommended item for user b.

・Item-Based
This recommends other items favored by users who purchased or rated the same product, such as "People who bought this item also bought..." If users A, B, and C all bought products A and B, products A and B are considered highly similar. Therefore, product B is also displayed as a recommended item for user D, who purchased product A.

Using this collaborative filtering mechanism broadens the range of products that can be recommended to users. It can also create encounters with products in completely different categories from past purchases or unexpected items previously unknown to the user. The ability to deliver serendipitous recommendations is a major appeal.

Can collaborative filtering save the hassle of searching?

So how are collaborative filtering recommendations actually used, and what effects do they produce?

・E-commerce Platforms
For example, it's said that over 35% of sales on a certain e-commerce platform come from its recommendation feature. Furthermore, the accumulated know-how is leveraged across other fields, such as movie streaming services. By accumulating data, they enhance personalization for individual accounts, for instance, suggesting "Users who watched this also watched..."

・Music Streaming Services
Some music streaming services have achieved more sophisticated recommendations by combining collaborative filtering with deep learning. They train the system to learn user preferences not evident in browsing history, recommending songs from genres or artists not previously in the user's history. This creates new points of connection between users and music/artists.

・Video Streaming Services
Major video streaming services have enhanced their recommendation capabilities by combining collaborative filtering with 5-star ratings (reviews) and various machine learning techniques. As a result, it's said that three out of four users actually choose the recommended movies, and a full 80% of total viewing time comes from recommendations. Companies providing these services demonstrate that they view recommendation systems as core to their business. This is evident in actions like CEOs personally leading algorithm improvements and hosting algorithm contests with substantial prize money.

Thus, it's clear that leveraging collaborative filtering contributes to sales for many companies. For users, it offers the benefit of discovering unexpected products while also increasing the likelihood of finding preferred items without repeatedly searching using keywords or conditions. This makes it an algorithm with significant advantages for both parties.

Considering the perspective of reaching preferred products without searching, one hypothesis emerges: "Could collaborative filtering be a particularly effective approach for Generation Z consumers, who are said to use search engines less frequently?" The next chapter explores how collaborative filtering might influence Generation Z's consumption behavior.

Gen Z, who value opinions from people with similar tastes and types. What kind of product recommendations resonate with them?

So far, we've discussed the benefits of incorporating collaborative filtering. For users, one appeal of collaborative filtering is likely the reduction in search effort. If that's the case, Gen Z, often said to use search engines less frequently, might be a good fit for collaborative filtering.

But is it really true that "Generation Z uses search engines less"? It's often said that Gen Z users, when curious about something, tend to gather information using hashtags on social media rather than searching engines. However, surveys targeting Gen Z reveal that the most frequently used tools for information gathering are search engines like Google and Yahoo! (Joint survey by Japan Information Inc. and Trend Catch Project, 2021). Furthermore, data shows that when gathering information about newly discovered brands or products, while a higher percentage of women use Instagram, many men search directly by entering the product name into a search engine (SHIBUYA109lab. "Survey on Gen Z's Consumption Behavior via SNS," 2022). This suggests we cannot simply say "Gen Z uses search engines less."

What's even more interesting is how they use these tools. They discover products on Instagram, where hashtags make searching easy and visual imagery is readily accessible. They track current social trends on Twitter. They learn about popular music and dances on TikTok. And they use search engines for deeper research or supplementary information. This reveals they actively switch between multiple tools for information gathering, depending on their specific purpose. Moreover, when making purchases, they appear to seek "endorsements from trusted third parties"—such as credible influencers or users with similar tastes—before committing to a buy.

Therefore, while it's not necessarily true that "search engine usage is low," considering that they "refer to endorsements from trusted third parties before making a purchase," it's reasonable to infer that Gen Z's behavior aligns well with the principles of collaborative filtering. This is because "recommendations from trusted third parties" can be seen as closely resembling the concept of user-based collaborative filtering.

Gen Z often follows relatable influencers on social media or reads comments from people with similar tastes on review sites, then bases their purchase decisions on the information gathered there. In a sense, this means they are customizing/personalizing how they encounter products and services. In other words, they may be unconsciously practicing a values-based collaborative filtering algorithm in their daily lives.

If so, collaborative filtering not only aligns with Gen Z's consumption behavior but may already be central to it. Building incentive-based communities leveraging collaborative filtering for marketing could potentially become an effective approach to stimulate Gen Z consumption.

Furthermore, if collaborative filtering proves effective for Generation Z, observing and analyzing data from their parents' Generation X/Y could potentially enable predicting consumption trends by household attributes. Collaborative filtering thus holds the potential for new marketing approaches.

We posed the question, "Can collaborative filtering-based product recommendations capture the hearts of Generation Z?" and explained the fundamentals of collaborative filtering alongside Generation Z's consumption trends. The answer to this question seems to be "Yes." In fact, it might be more accurate to say that "Generation Z naturally practices consumption behaviors based on collaborative filtering-like logic." In other words, to get Generation Z to purchase products and services, appealing to how well they align with their values and resonate with their points of empathy could be a strong driving force. How about considering this perspective when exploring approaches to Generation Z, who will bear the future of consumption?

To learn about trending terms like "FOMO / JOMO," please also refer to this resource.