System design at Meta shows up from E5 onward and leans toward real product-scale problems — feed ranking, ad delivery, and real-time leaderboards — rather than generic templates. Interviewers expect scaling, consistency and fault-tolerance trade-offs to be raised without being prompted.
Problem 1Design a Near-Real-Time Personalized Ad Delivery System
Design a system similar to how a large social platform delivers personalized ads in near real-time — ads should be targeted, ranked, and served with low latency at very large scale.
Requirements to clarify first
- What's the hard latency budget for serving one ad request — tens of milliseconds, or a bit more slack?
- Should ranking be personalized per-user in real time, or is a coarser, periodically-refreshed personalization acceptable?
- Is slight overspend against an advertiser's budget (a pacing miss) tolerable, or must budget pacing be exact?
Only one verified system design prompt has cleared review so far for Meta — more are added here as additional candidate reports are verified.