What Is a Recommendation Engine?
Recommendation engines suggest products, content, or actions based on user behaviour. The technology behind Amazon's 'customers also bought'.
Key Takeaways
- Recommendation engines predict what a user will want next based on behaviour data
- Collaborative filtering uses the behaviour of similar users to make recommendations
- Content-based filtering uses the attributes of items the user has liked
- Amazon attributes up to 35% of revenue to its recommendation engine
What a recommendation engine is
A recommendation engine is an AI system that analyses user behaviour, preferences, and item attributes to predict and surface items a user is likely to want. Netflix recommending your next series, Amazon showing customers who bought this also bought, Spotify generating a personalised playlist — these are all recommendation engines at work.
Collaborative filtering
Collaborative filtering makes recommendations based on the behaviour of similar users. If users A, B, and C all bought products X and Y, and user D bought product X, the engine recommends product Y to user D — because people like them (users who bought X) also bought Y. This approach requires no knowledge of the product itself, only the patterns of user behaviour. It is the most common approach for eCommerce product recommendations.
Content-based filtering
Content-based filtering makes recommendations based on the attributes of items the user has previously liked. If a user has watched three action films set in space, the engine recommends other space action films based on matching attributes. This approach requires rich item metadata (genre, director, actors for films; category, brand, material for products) and does not depend on other users' behaviour.
Hybrid approaches
Most production recommendation engines use a hybrid approach that combines collaborative and content-based filtering, often augmented with additional signals: recency (what has the user done lately?), context (time of day, device, location), and popularity (what is trending?). The blending weights of these signals are themselves optimised by machine learning.
Business impact
Amazon attributes up to 35% of revenue to its recommendation engine. Netflix claims that 80% of content watched is driven by recommendations. For eCommerce businesses building their own recommendation capability, the ROI is substantial even at smaller scale — personalised product recommendations in post-purchase emails typically increase email revenue by 10-30% compared to static product selections.