Meta Data Scientist Interview Experience — Three of Four Onsite Rounds Done, Technical Session Pushed Two Weeks

I finished 3 out of 4 rounds of my VO (virtual onsite) yesterday, and while it's still fresh in my memory I wanted to share the experience. The reason I only got through 3/4 is that right after my third round, the recruiter told me the final Technical session would need to be rescheduled — it ended up getting pushed two weeks out, which feels like such slow torture (I originally wanted to knock out the whole thing in one day, specifically so I could get it over with quickly one way or the other).

Since I signed an NDA, I won't describe the exact content in precise detail, but I'll share how I prepared and roughly what kinds of questions I ran into.

Prep process:

1. I'd really recommend Jeff Su's videos, and I'll also share some behavioral prep advice from a friend of mine who's currently a director at a major retail company (the PDF she sent me wouldn't upload, but the content was similar to something I'd seen online). The key point: your career story is not your career history. I actually made exactly this mistake a while back when I interviewed for a senior manager role at a large multinational HR company (which is why I failed). But that was my first-ever final round, so I'm just chalking it up as practice (sigh).
2. Every session opens with a "short self-introduction." So I'd suggest preparing a longer self-intro (around 2-3 minutes) for the "tell me about yourself" question, plus a shorter version to break the ice at the start of an interview. Either way, my structure was:
   • Who I am right now — besides talking about what I currently do, I'd highlight 2-3 things I'm good at, like leading cross-functional teams, analyzing customer insights, and being the "firefighter" for ad hoc requests from our VP and SVP whenever they have customer-oriented or purpose-driven questions.
   • Who I was, and why I became a DS — here I'd quickly connect my PhD background to data science, both to show my interdisciplinary background and to explain why I became a DS.
   • Who I want to be in the future.
3. For Behavioral, I'd strongly recommend finding someone to mock-interview you, ideally a manager or someone with more senior experience. If you can't find anyone, you can pay for an interview coach or a mock interview online — I tried two: Prepfully and IGotAnOffer. Each had its pros and cons. Prepfully was more expensive, and when I did the mock interview there I hadn't even finished my prep reading and had zero confidence, so my mock analytical execution round was a disaster. But honestly, that experience woke me up and changed the direction of my studying afterward, so it did help. IGotAnOffer was a bit cheaper, and I booked two different coaches, both from big FAANG companies. The first one told me upfront that his DS background leaned software engineering, so he didn't think he could give me good technical interview advice, but he was a big help on the behavioral side. The second one happened to be a senior DSA employee at Meta — he reviewed my phone screen answers, gave me advice on my analytical reasoning process, and finished with a 10-15 minute mock, which I found extremely helpful.
4. ChatGPT is a good friend, but it's not a magic tool: when reviewing some of my behavioral answers, I did use ChatGPT to help polish them, but the actual interview is a very different situation. ChatGPT and other AI tools are just aids — what really matters is internalizing the scenarios a behavioral interview might throw at you so they become genuinely your own. Also, don't make up stories: in the real interview, the interviewer is very likely to keep peeling back your answer layer by layer (I'll give an example of this later).
5. Stories to prepare for Behavioral (necessary if you're interviewing for L5 or above):
   • Any work challenge you've faced and how you eventually resolved it
   • Any accomplishment at work (big or small) and what impact it had on the people around you
   • Any conflict you've handled
6. Analytical Execution & Analytical Reasoning: I really recommend Emma Ding's videos — she organizes the material really well. Embarrassingly, I have an MS in Statistics (picked it up along the way during my PhD) with a 3.9 GPA, but I'd basically forgotten almost everything after graduating (seriously), and I've been bad at combinatorics since high school (I hate that chapter) — I'd flat-out give up whenever I hit a probability question (this is true). This time I spent two weeks seriously rehabbing my brain, looking up a ton of Bayes' Theorem and binomial probability problems to practice, and reviewing a lot of the required coursework from a Statistics master's program. Honestly, no matter how the interview turns out, this whole process was genuinely useful for my future, because for the first time in my life I actually understood what combinatorics is doing (lol).

Bayes' Theorem and the binomial distribution matter a lot, because a lot of social media or e-commerce design comes down to click or not click (binomial), and probability questions often finish with a combo move layering Bayes' Theorem on top. If it's a dice question, that's uniform distribution, but that's not hard either — once you've got the basics down, you won't panic on the variations.

7. The following two videos were recommended to me by the senior Meta DSA I paid as an interview coach, and I personally found them very helpful: one is a framework for framing "how to measure success" that works for both data and product-oriented roles, and the other walks through an example of Facebook Marketplace metrics.

Behavior Session: mainly focused on a few scenarios — (1) have you ever had a coworker who felt "not welcome"? How did you handle it? (2) What's something you pushed yourself to learn quickly in a short amount of time? (3) How do you handle conflict?

The first question was pretty fresh to me, but I thought it was a great one. The interviewer really dug into the details — he wanted to know how I handled both sides of the situation (one side disliked the other, and that other person felt unwelcome as a result). He asked extremely, extremely detailed follow-ups... all I can say is, good thing I'd actually dealt with something like that before (sweat).

Analytical Execution: I got the famous fake-account problem. Honestly, the write-ups on this forum are always so fragmented that I never actually knew what the question really looked like until I got it myself (sweat) — though that's partly on me too, since I'm the kind of lazy person whose motivation dies by page two of a search. But from looking at who passed and who failed the VO, basically everyone ends up with 4-5 sub-questions. So if you get the fake-account problem and manage to work through 4-5 sub-questions within the time limit, that should count as a pass (?).

My interviewer started with the scenario, asking how you'd determine whether an account is fake — we spent about five minutes on that, and I wasn't sure if he was satisfied with my answer, but eventually we moved into the actual calculations.

The calculation part is all connected: he gives you the ratio of fake to authentic accounts, then sets up a scenario. The first probability question matters a lot, because its answer gets used later in the calculation. I actually got stuck on the first question for a bit, because I overcomplicated it in my head — stuck for maybe 30-60 seconds — but once I figured it out, the rest went smoothly. The first question was really just P = a / (a+b): once you write out the algebra it cancels down and the probability p just falls out, nothing to do with binomial or Bayes' at all.

The second question involved the binomial distribution. I calculated it a bit fast at first, so the interviewer asked how I got that number, and I spent another 30 seconds expanding the formula and explaining my reasoning.

If I remember right, only the last question used Bayes' Theorem. Honestly, the numbers I got this time were all suspiciously clean — for a couple of questions I didn't even need a calculator. On the last one, to double-check I hadn't misread the problem, I proactively used "false positive" and "false negative" to confirm a couple of the numbers (mainly because the numbers were so clean I wanted to make sure I hadn't misread anything). The interviewer's reaction seemed pretty positive, so I guess in some way that also confirmed to him that I understood Type I and Type II errors.

Same as the Behavioral round, we spent the last few minutes chatting, and I'm hoping that one minute I got stuck on doesn't turn into a death sentence, QAQ.

Analytical Reasoning: I don't know why, but I had this feeling (a mysterious sixth sense?) that this section is what actually decides whether you pass.

The interviewer for this round didn't look at the whiteboard at all — he was actively listening, discussing, and typing notes the whole time, but told me, "you don't need to use the whiteboard at all, because I'm not going to look at it anyway," lol. So this whole section was just me looking at the screen and talking with him directly.

I got the (in)famous feature-recommendation question. The prompt was something like this (all very high-level, but I could tell the interviewer wanted me to discuss it from a very top-down perspective, covering everything): if you wanted to design a feature — basically a new product — today, how would you convince your boss?

• First was the mission: what is our mission — growing users, growing revenue, or both?
• Once the mission is settled, discuss whether it's even worth building this new feature/product on top of the existing one. If so, make the case using external + internal reasons.
• Externally: is there actually a market opportunity here overall? Are there similar products? Who are the competitors, and why should we be the ones carving out a piece of this market?
• Internally: what advantages do we get from adding this feature? What can it add?

That's roughly the angle I took. If your boss is convinced, what data would you use as the basis for your analysis?

• If it's a new feature on FB or Instagram itself, what factors matter — things like posts, what kinds of newsfeed clicks, lifestyle data (places traveled, etc.)
• If it's built on FB, besides FB's own data, you could also think about pulling data from "sibling" products under the same parent company, like Instagram or Threads — certain data can usually be shared between subsidiaries under the same parent company.
• Is there any other third-party data you could use? Some companies pay for third-party data, and I'd factor that in too.

If you're going to measure it, what do you measure?

• You need to clearly define your North Star Metric (NSM) and your Counter Metrics (CM).
• For example, your NSM might be traffic, and the CM might be conversion rate, because higher traffic doesn't necessarily mean higher conversion rate.
• Conversion rate can also be broken into multiple layers by engagement type — some can be compared test vs. control, some only exist in the test group — and you need to clearly explain why you'd monitor each of these.
• Probably because I talk too much, I brought up something that maybe fewer people mention: potential revenue. I said this isn't an NSM or a CM, but it's still a metric worth tracking. If the goal is monetizing traffic or CTR, estimating potential revenue matters — if a single click generates $0.5 in revenue, then based on the experiment you can eventually estimate total potential revenue.

Honestly, I only brought this up because of my own personal experience. The interviewer's face lit up a bit at that point, but I genuinely couldn't tell if that meant (a) "I like your answer" or (b) "this rookie is overthinking it, we don't need this." If it's (a), great. If it's (b)... learn from my mistake, everyone, QAQ.

If you needed to run an experiment today, how would you do it? (Since this feature has never launched before, it's the standard A/B test question.)

First is picking the experiment site. Whenever I answered, I made sure to frame it as "assume we are going to XXXXX" — I think that assumption matters a lot, since the right decision changes depending on the situation.

Based on my own experience, I'd start with a specific city or metro area (if there's enough budget, sure, pick several locations), and ideally that city or metro area should be (1) large enough in population, and (2) have a demographic distribution that, in every dimension, represents the median of the company's product users, or is close to the population distribution of the company's core user base. If the city or metro area you pick instead represents "the population distribution of that whole country," that's also fine — it just produces a different kind of result, closer to "how people in this country react to our new product" rather than "how our company's current users react to it." (3) Test/control stratification — the interviewer got very detailed here, so I ended up walking through how to build the stratification system, how to score it, how to bucket into deciles, and then how to split into test/control from there.

Judging from his expression, I think he was fairly satisfied with my answer on this part? (Of course, if he's the poker-face type, who knows, QAQ.)

One more thing I added: after I finished, for some reason I can't quite recall, I threw in a note about the final measurement and decision. I (probably talking too much again) told the interviewer: even if the final experiment results are great, that doesn't necessarily mean we should launch. In my report to my boss, I'd also add cost considerations — technical feasibility, human resources, etc. — because launching a whole new product globally might require a lot of investment, not just hardware but software staffing too: do we have enough engineers to build it, do we need to hire more, and how much ongoing headcount does maintenance need afterward. Whether it pays back short-term or only long-term, I'd write all of that into the report as well (and honestly, if I were actually reporting this to a VP or SVP, I would include these considerations).

After I said that, the interviewer looked pleased, but again... I have no idea if that was a good sign or just me talking too much.

Then came the usual 3-5 minutes of chatting at the end, except we actually ran seven minutes over — hopefully that's a good sign.

Technical Session: because the original interviewer wasn't available, it got pushed back two weeks, QAQ. This is supposed to be the part I feel most confident about, but now it just feels like slow torture waiting for it.

I hope this post helps everyone else out there prepping for their VO, and here's hoping I get the offer soon — I really want to switch jobs.