Recommendations: Novelty vs. Familiarity
Search and recommendation products have deeply changed how users discover and engage with content. Whether seeking information, products, or entertainment, users want to be surprised by novelty, while also exploring familiar & related items based on their past transactions and interactions.
Most product managers seek to drive repeat user behavior, and recommendations are an excellent way to influence repeat usage. Critically, striking the right balance between the "novelty of results" and the "familiarity of results" is crucial for the success of our products, especially for search products & recommendations products.
We delve into the significance of this equilibrium, the factors influencing it, and strategies to achieve it effectively, recognizing that users can be in one of two modes: actively seeking something new or passively browsing related items.
Recognizing User Modes: Active vs. Passive
Users often fall into one of two modes when interacting with search and recommendation products: active search mode, and passive browsing mode.
Active Mode: In this mode, users are actively seeking something new or specific. For example, a user might be searching for a unique gift for a friend's birthday. In such cases, they desire novel and distinctive results that cater to their immediate needs and preferences. In the active mode, balancing novelty is paramount to providing a delightful and effective user experience. Surfacing things that are familiar doesn’t make sense here, because they’re explicitly looking for something new.
Passive Mode: Conversely, in passive mode, users are not looking for anything specific but are rather browsing related items or content for leisure or exploration. An example could be a user scrolling through a music streaming platform like Spotify, looking for similar podcasts or songs. Here, striking a balance between novelty and familiarity is crucial to maintain user engagement and prevent fatigue. Don’t surface lots of new items that aren’t similar to what they want, or else they’ll question whether your algorithm is serving them well.
Now that we understand user goals when it comes to discovering new information, results, and/or recommendations, we can break down the kinds of recommendations that our products should serve back to users.
The Dynamics of In-Product Recommendations
Novel Results: When we introduce new content for users, it keeps the product fresh, competitive, and aligned with evolving user interests. Novelty is essential because it entices users to explore further, discover new items, and fosters a sense of excitement within the product ecosystem.
We need to make sure that when users signal that they’re in active mode, e.g. kicking off a search in a search box, we ought to provide them with very different results from what they’ve seen in the past.
Familiar Results: Conversely, presenting users with recommendations that are aligned with their existing preferences and interests is necessary. It offers a comfortable and reassuring experience by surfacing items users are likely to enjoy based on their past behaviors. Familiarity helps in retaining users and keeps them engaged over time.
Therefore, when users signal that they’re not in active mode, we need to assume that they’re in passive mode and aren’t actively looking for something new. If we surface totally new results while they’re in passive mode, they won’t feel like they’re being well-served. So, we should be providing recommendations that align tightly with their prior searches, purchases, transactions, and other user behaviors.
Factors Influencing the Balance
Several factors influence the delicate balance between novelty and familiarity in search and recommendation products:
User Behavior and Preferences: Understanding user behavior and preferences through data analytics is fundamental. Analyzing user interactions, such as click-through rates, time spent, and purchase history, provides insights into whether users are seeking novelty or familiarity at any given moment.
Personalization Algorithms: Sophisticated recommendation algorithms, such as collaborative filtering, content-based filtering, and hybrid approaches, play a pivotal role in delivering both novelty and familiarity. These algorithms adapt to user preferences and behavior, ensuring a tailored experience.
Result Diversity: The diversity of results within the platform is crucial. A well-balanced mix of new and familiar items ensures that users have options suited to their current mode. Maintaining a robust inventory helps to strike the right balance.
Strategies for Achieving the Balance
Real-Time Personalization: Implement real-time personalization techniques to assess the user's current mode and adjust recommendations accordingly. Algorithms can identify user intent and dynamically shift the balance between novelty and familiarity as needed.
User Control: Empower users with control over the level of novelty and familiarity they prefer. Offer settings or filters that allow users to customize their experience, indicating whether they want to explore new content or stick to what's familiar.
Continuous Testing and Experimentation: A/B testing and experimentation are invaluable tools for product managers. Regularly test different approaches to determine what resonates best with your user base. Monitor user feedback and iterate accordingly.
User Education: Educate users about the capabilities of the product. Provide tips and guidance on how to make the most out of the platform, helping them understand how to balance between novelty and familiarity effectively.
Closing thoughts
When making recommendations in our products, finding the right balance between the "novelty of results" and the "familiarity of results" is a continuous challenge.
The first step for us as product managers is that we need to recognize that users can be in active or passive modes of discovery. Then, based on which discovery mode they’re in, we should provide targeted users experiences in service of their preferences.
Striking this balance not only keeps users engaged but also fosters long-term loyalty, ensuring the ongoing success of the product in the dynamic digital ecosystem.
