Recommendation system: A recommendation system is software that predicts what a person is most likely to want next, such as a product, an article, or a video, based on past behavior and the behavior of similar people. It drives feeds and "you might also like" boxes across the web.
Three basic approaches power most recommenders. Collaborative filtering looks at behavior: people who bought what you bought also bought this. Content-based filtering looks at attributes: you read articles about freight rates, here is another one. Hybrid systems blend both, and modern ones use embeddings, numerical representations of items and users, so that "similar" can be computed at scale. A ranking model then decides what to show first.
For a business, recommenders show up in more places than retail. E-commerce cross-sells and "frequently bought together" boxes are the obvious case. CRM tools suggest a next best action for a sales rep. Knowledge bases suggest related articles to support agents. Learning platforms suggest the next course. Job boards match candidates to roles. Anywhere you have many items and many users, a recommender can be pointing the way.
Their known problems are worth naming. New users and new items have no history (the cold-start problem). Popular things get recommended more, which makes them more popular, which crowds out everything else. Feeds optimized for clicks drift toward whatever holds attention, not whatever helps. And a recommender that looks good in a dashboard may not be adding anything: the real test is an A/B comparison against no recommendations, or against a simple "most popular" list.
Example at work
An e-commerce manager at an industrial supply company adds a "customers also ordered" module built from two years of order history. It surfaces pairings the merchandising team had never thought to bundle, such as a specific gasket with a specific pump. The team runs it against a control group for six weeks before rolling it out to every product page.
Why it matters
Recommendation systems shape what your customers see, what your employees find, and what you get shown all day. Knowing how they work, and how they go wrong, lets you use them deliberately instead of being steered by them.
Related terms
- Machine learning (ML)Machine learning is the branch of AI in which software learns patterns from examples instead of following rules a programmer wrote by hand. Nearly every modern AI tool, including chat assistants, is built on it.
- EmbeddingAn embedding is a list of numbers that represents the meaning of a piece of text, an image, or other content, so that items with similar meaning end up with similar numbers. Embeddings are what let AI systems search by meaning instead of exact words.
- Predictive analyticsPredictive analytics uses historical data and machine learning models to estimate how likely a future outcome is: which customers will churn, which invoices will be paid late, what a store will sell next month. The output is a probability, not a fact.
- Unsupervised learningUnsupervised learning is machine learning on data with no correct answers attached. The model finds structure on its own, grouping similar customers, spotting unusual transactions, or clustering tickets by topic; a human then decides what the groups mean.
- AI biasAI bias is a systematic tendency for an AI system to produce outputs that are unfair or skewed toward certain groups, usually because the data it learned from reflected historical patterns, gaps, or human prejudice.