Recommendation System Architecture Diagram Template

What this template is for

A recommendation system architecture diagram shows how a recommender turns a user request into a ranked list of items. This template lays out the classic two-stage funnel: candidate generation (recall) narrows millions of items to hundreds, a ranking model scores those candidates, and a re-rank/filter stage applies business rules before returning recommendations. A feature store feeds the ranking model, and user behavior flows back into features. Use it to design a recommender, document an existing one, or explain why recommendations are computed in stages rather than one pass.

When to use this template

• Design a recommendation system's retrieval and ranking stages before building.
• Explain why recommenders use a two-stage funnel instead of ranking everything.
• Document where features come from and how they feed the ranking model.
• Show the feedback loop from user behavior back into the feature store.
• Plan where business rules and filters apply in the re-rank stage.
• Compare a collaborative-filtering recommender against a two-tower retrieval model.

How to use it

1. 1Start with the user request entering the funnel at the top.
2. 2Add candidate generation (recall) that narrows the full item pool to a shortlist.
3. 3Add the ranking model that scores each candidate, fed by a feature store.
4. 4Add a re-rank/filter stage for business rules, diversity, and freshness.
5. 5Add the recommendations output at the bottom of the funnel.
6. 6Add the user-behavior feedback loop into the feature store.

Quick example