Meta DS Prep

Sample Ratio Mismatch — diagnose and decide

Design a metric for a Marketplace listing-expiry notification

SQL — 7-day retention by signup cohort and country

Estimate medium-term revenue impact of an ads ranking change from a 2-week A/B test

Scheduled posts: estimate failure rate and define feature success

SQL — unconnected-post counts and per-relationship reaction averages

Define success for an unconnected-content feed launch

Experiment design for an unconnected-content feed launch, with risk focus

SQL — heavy callers and per-DAU call involvement over 7 days

Infer group-call demand from 1:1 call data

Success metrics for a group-call launch, with cannibalization

SQL — posts with more than 10 views in the past 7 days

SQL — view prevalence of Spam/Scam content in the last 30 days

Measure the severity of harmful content and evaluate a detection model

A/B design for launching group call on WhatsApp: metric and cluster-randomization traps

Validating feature demand without an A/B test: survey, historical data, and the C-level case

SQL — sellers with more than three products that had multiple interactions

SQL — vehicle category's share of interactions on newly created US listings

Is a new-listing notification for Marketplace buyers worth building?

SQL — number of users with 2, 3, and 4+ accounts

SQL — percentage of multi-account users who have any unread notifications

Validate an account-switcher reorder from historical data before running the A/B

A/B design for reordering the account switcher by unread notifications

SQL — SHOP vs WEBSITE revenue share over 30 days and reading model health from it

Evaluating a new ad recommendation algorithm when 50/50 isn't safe and when no test is possible

Should we uprank SHOP ads? Evaluating the idea and designing the ranking change

SQL — prior-year $1k+ advertisers and their this-year spend share

Creation source growth — real lift or cannibalization, and the exclusion SQL

Evaluating an AI-assisted ads creation tool — metrics, test setup, and broader impact

Evaluating an emoji-reaction feature with a 5-second long-press trigger — measurement and the C-level report

SQL — relationship between pixel signal quality and ads performance

Evaluating a pixel-health notification feature for advertisers — is it a win?

Measure fake-account prevalence on Meta and evaluate a detection model

Evaluate whether a driving simulator's output is realistic compared to real-world data

Predict whether each car at an intersection will turn left, go straight, or turn right

Detect whether any of n cars will collide given their kinematics, from 1D to 2D

Ads A/B read: CTR up, revenue down — diagnose and decide

SQL — ads impression→conversion, peak vs off-peak hours

Evaluate a new ads ranking algorithm balancing revenue, advertisers, and users — and pitch it to leadership

Chatbot LLM: probability of a good response, two in a row, and updating after three

Define engagement metrics for a video platform, and read the comment-per-user distribution

SQL — authors whose posts received at least two replies in the last 7 days

SQL — percentage of users whose posts received replies from two distinct US users

Stolen post detection — diagnose harm, design a new algorithm, and evaluate it

SQL — good-condition unreturned copies and their renewed-more-than-twice share

Python — max points from up to 3 books, each in a different category

Clustering users for group call: traditional clustering vs. social-network community detection

Why are Instagram Stories used more than Facebook Stories — and how would you close the gap?

Bayes theorem on a fake-account test, and the at-least-once probability

SQL — sort each user's unread notifications by sender affinity from a daily-snapshot table

Active Advertisers per Country — Bottom-N and YoY Threshold SQL

Detecting Fake Accounts at Scale — Labels, Features, Evaluation

Tell Me About Reporting to Someone Brand-New to the Team

How Do You Make a New Team Member Feel Welcome?

Determining group size and diagnosing bad cluster-randomization clusters for a group-call launch

Friend recommendation: how do you improve the success rate of People You May Know

Friend recommendations: filtering commercial and ad-driven accounts out of PYMK

Testing a friend-recommendation change — unit of randomization, spillover, and the metric horizon

Friend-request acceptance rate — the simplest SQL question and the three traps in it

VR hand-waving gesture — define accuracy and decide whether to launch

Fake-account detection from friend-request signals only — no model, no nothing

Phone-screen SQL — 30-day user metric, top per category, compare two products

Tell me about a time you did something inclusive

SQL — unconnected posts with views longer than 60 seconds in the past 7 days

Measure whether a post type promotes friendship between friends

AR round: extracting insight from a 'flat' chart when an A/B test isn't applicable

AE round: binomial probability then comparing two models via a two-proportion Z-test (pooled vs unpooled)

SQL: daily cumulative sum with a date-join gotcha, then build a predictive model in SQL

SQL — invalid-event percentage per pixel for yesterday

Design a metric to measure platform-wide pixel signal health

Oculus SQL round — most used app, time-spent share per category, social-vs-game engagement, unhealthy users

Designing a success metric for 'high-quality' Facebook notifications

Measuring a nearby-event notification feature with network effects in the A/B test

Survey response rate by country — the easy SQL with two quiet traps

Feature selection for a SHOP-ads promoter — with a model, and without one

SQL — SHOP spend share per advertiser, and whether SHOP ads outperform WEB ads

Multi-account switcher: how to interpret a user who glances for a few seconds and logs off

Brand ads spend appears to be declining — diagnose it, triangulate it, and prove brand ads are worth buying

Launching video as a new brand ad format — A/B design, optimal length, metric choice, and ecosystem effects

SQL — last-week CTR overall, then CTR by brand vs direct campaign type

Pre-launch retailer messaging platform — pilot metric, expansion targeting, and pre/post cost analysis

If ad CTR rises 5% after a ranking change, will advertisers spend more?

AR chart: treatment CTR is already higher in the pre-launch period — what's wrong and how would you fix the visualization?

Proxy metrics for whether fake-account activity on Facebook changed over the past month — when fakes aren't yet labeled

Detecting a fake account at signup — before any behavior signal exists

AR round: +5% overall but +100% in one demographic subgroup — is the new ad ranker a launch?

SQL — top-10 video-call initiators in 7 days and share of France DAU on a video call yesterday

FB Lite retention is low a year after launch — how would you find out why?

Python — smallest non-negative number from the odd digits of a given number

Python — most common customer comment across locations, deduped per location

Python — area of the largest island of 1s in a 0/1 matrix

SQL — total sales and unique paying customers by payment type

SQL — top 5 customers by average payment per book of people they invited

SQL — author count, percent with .com URL, and percent who never made a sale

SQL — ads revenue in USD across a three-table FX-conversion schema

Python — cheapest round-trip ticket from departure and return price arrays

Python — generate all k-digit strobogrammatic numbers (same after 180-degree rotation)

DOT — convincing the PM, sizing the need, and picking a model for duplicate-content detection

SQL — top-N ads by conversions over the last M days, three-table ad funnel

AR round: reporting a new ad ranker's A/B result to a CFO vs a DS, where to put CI on the chart, and how to communicate uncertainty

Marketplace ads A/B: DAU down, ad revenue up — diagnose and decide

AE: ratio of average comments (real vs fake users) — sampling distribution, 95% CI, and what sample size changes

AR: Instagram shopping launch — opportunity sizing, revenue estimation, A/B design, and post-launch troubleshooting

Reels launching inside the Instagram feed — questions to ask, metrics to pick, how to visualize

Data modeling — tables and attributes for a product scenario, plus grain choice for a multi-step process

SQL — four-part video watch stats and reaction-without-comment share

Python — streaming-event buffer: flush-on-full, drop-earliest, skip test users

Python — do overlapping meetings ever exceed a maximum total participant count?

SQL — heavy message receivers per day and average read-rate for spammers

Measuring spam without a labeled table, interpreting report-rate drops, and A/B design when spammers are rare

Likes are down 10% today on a social product — diagnose it

Fake-news surge: brief the VP in 24 hours — scope before impact

Identify business travelers on a social platform — signal design and validation

AE: short-video recommender — top engagement metrics, A/B sample size, views distribution, and top-10 overlap between users

Bayesian probability a received friend request is from a fake, given a 10× send-rate disparity

Circle launch — success metrics, experiment design, eng-resource tradeoff, and comments-per-post interpretation