How Meta interviews
1. The shape of a Meta loop
Meta runs the fastest top-tier loop in the industry. Four to five rounds in a single day, each one forty-five minutes, decisions in days not weeks.
A standard E5 (senior) loop looks like this:
9:00 Recruiter intro / day briefing (15 min)
9:30 Coding round 1 (45 min)
10:30 Coding round 2 (45 min)
11:30 Lunch with a host engineer (45 min, not graded)
13:00 System design round (45 min)
14:00 Behavioral / "Jedi" round (45 min)
15:00 Cross-functional round (45 min, optional)
16:00 Recruiter debrief (10 min)
By the end of the day every interviewer files their notes. The hiring manager reads them within twenty-four hours. The recruiter calls you back within a week. The whole loop is engineered for speed.
2. What Meta grades you on
Three things in roughly equal weight on a system design round.
| Signal | What it means at Meta | The thing that fails it |
|---|---|---|
| Scoping | Can you cut the question to three things and pick the hard one fast? | Spending eight minutes asking the interviewer what to design. |
| Tradeoffs at scale | Every Meta question has a "but you have a billion users" twist. Did you bring it up first? | Drawing a clean diagram for ten million users and waiting to be corrected. |
| The hardest sub-problem | Identify the one piece that's actually hard, dig into it, defend the choice. | Going wide on every component. Five shallow boxes instead of one deep dive. |
Meta interviewers are looking for the person who picked the hard sub-problem on their own. They will not push you toward it. If you spend the round on the easy 80% of the design, the rubric will read "didn't reach depth."
3. What's different here
A few patterns specific to this company:
- The hybrid round. Some loops include a single round that's coding and design in one. You scope a system, build a small piece of it, then defend the design. This round is brutal if you treat it as either pure coding or pure design — you have to actually do both.
- Scoping ambiguity on purpose. Meta interviewers under-specify questions on purpose to see if you scope them. "Design messaging." Group? One-on-one? With end-to-end encryption? At what scale? They want you to pick the version of the question worth answering and tell them what you picked.
- The billion-user twist. Almost every Meta question has a "what happens at our scale" follow-up. Have an answer ready before they ask.
- No Googley-style committee. The hiring manager makes the call, not a committee. This means the hiring manager's specific tastes matter more here than at Google. Try to learn during the loop what your hiring manager weights — sometimes you can ask the recruiter beforehand.
- The behavioral round is technical. Meta's behavioral round ("Jedi") asks about real conflicts on real projects. They want stories with specifics. "We disagreed on the data model" is a starting point, not an ending point. They want the resolution, the tradeoff, what you'd do differently now.
4. The level bar
The bar at each level:
- E3 (entry IC). Generally no system design round. Coding-heavy. Skip this whole course at this level.
- E4 (mid IC). One system design round, scoped tightly. You're not expected to dive deep on every component. Naming the right pieces correctly is enough.
- E5 (senior). This is the most-interviewed level. You need one genuine deep dive that survives three pushes, plus the scaling story.
- E6 (staff). Two deep dives. One alternative architecture you considered and rejected with reasoning. The migration story ("this is how I'd roll this out on top of an existing system").
- E7+ (senior staff and up). Now you're being graded on the organization implications. How does the system fit a roadmap, what teams own what slices, what's the launch plan.
In this chapter the default target is E5. We call out the E6 shift where it matters.
5. The Meta-flavored mistakes
Watch for these. They lose loops here specifically:
6. The seven scenarios in this chapter
| Unit | What it covers | Why Meta asks it |
|---|---|---|
| 2.1 The news feed | Fan-out write vs read, ranking, the celebrity problem. | The canonical question. Every E5 loop has a version of this. |
| 2.2 Messenger | One-on-one + group chat, read receipts, delivery semantics. | Tests real-time systems and ordering guarantees. |
| 2.3 Instagram Stories | Ephemeral content, view tracking, the TTL boundary. | Combines media pipeline with a hard product constraint. |
| 2.4 Facebook typeahead | Prefix search, ranking, personalization. | The "tiny" question that's actually about caching and ranking. |
| 2.5 The like button | Counter at extreme write throughput. The hidden hard one. | Looks trivial. Isn't. The interviewer is grading how soon you notice. |
| 2.6 Friend graph BFS | Multi-hop graph queries at billions of edges. | Tests whether you can design over a primitive Meta actually invented. |
7. The next page
We start with the question Meta has asked more than any other: design the news feed. We'll go through it the way an E5 candidate would on a good day, then break down what shifts at E6.