Explaining Instagram's Recommendation Funnel: Sourcing, Ranking, Signals and Feedback

Explain how Instagram's recommendation system works. Describe how it decides which content to show a user and in what order, what information those decisions rely on, and how the system improves over time. Aim the explanation at the depth expected of a product manager. ```hint Think in stages No system can score every post on the platform for every user in the time it takes a screen to load. Ask how the pool of possible items gets narrowed before anything is carefully ranked. ``` ### Clarifying Questions - Which surface should I focus on: the main Feed, Stories, Explore, Reels or suggested accounts? Each draws on a different pool of content. - Should the answer cover only organic content, or also how ads are selected and mixed in? - Are you more interested in the technical pipeline, in the product levers a PM controls, or in both? ### What a Strong Answer Covers - How the surfaces differ in their content pool and their goal. - The flow from candidate sourcing through filtering, ranking and final adjustments, and why it is staged. - The signals about the post, the creator, the viewer and their relationship, and how predicted actions become one ordering. - How the system learns from feedback, handles new users and new posts, and is evaluated with experiments. - Product trade-offs such as engagement versus well-being, relevance versus exploration, and followed versus recommended content. ### Follow-up Questions - A new creator with no followers posts a first Reel. How does it reach anyone? - A ranking change raises time spent in Reels but also raises reports of unwanted content. Would you ship it? - How would you give users more control over their recommendations, and how would you know the controls work?