Recommendation System

A recommendation system is software that ranks or suggests items for a user based on data such as past behaviour, item attributes and what similar users did. Main approaches are collaborative filtering (people like you liked this), content-based filtering (similar to items you liked) and hybrids that combine both with modern neural models.

Why it matters

Recommenders drive discovery on streaming, shopping, news and social platforms, and can strongly affect engagement and sales. They face known problems: the cold start problem for new users or items with no history, popularity bias that favours already-popular items, feedback loops that narrow what people see, and privacy concerns around the behavioural data they need.

Example

User A bought: router, ethernet cable
User B bought: router, ethernet cable, network switch
-> Suggest "network switch" to User A (collaborative filtering)

Embedding both users and items and comparing them with cosine similarity is a common modern implementation.

Common mistakes

• Optimising only for clicks, which can promote low-quality or repetitive content.
• Evaluating offline accuracy only, without live A/B tests.
• Ignoring cold start, so new products never surface.

Related

• Cosine similarity
• Embedding
• A/B testing
• Machine learning

This page is general information written by our editorial team and updated when documentation changes. Servers and providers differ, so test changes safely first. See our editorial policy and disclaimer.