Analyze User Transfer Distribution in Initial Launch Period
Quick Overview
Evaluates distribution analysis for transfer counts after a new P2P payments launch. Strong answers specify the denominator, describe zero inflation and right skew, locate mean, median, mode, and p95, and explain how adoption, cohorts, and exposure change the distribution over time.
Analyze User Transfer Distribution in Initial Launch Period
Company: Meta
Role: Data Scientist
Category: Statistics & Math
Difficulty: medium
Interview Round: Onsite
##### Scenario
For the newly launched P2P payments feature, analyze user transfer counts.
##### Question
What distribution do you expect for number of transfers per user during the first 30 days post-launch? Estimate where the mean, median, mode, and 95th percentile would lie in that period. After two months, how and why would this distribution change?
##### Hints
Discuss right-skew, long tail, possible log-normal/Pareto behavior, and effects of adoption and churn on summary statistics.
Quick Answer: Evaluates distribution analysis for transfer counts after a new P2P payments launch. Strong answers specify the denominator, describe zero inflation and right skew, locate mean, median, mode, and p95, and explain how adoption, cohorts, and exposure change the distribution over time.
Analyze User Transfer Distribution in Initial Launch Period
Meta
Jul 12, 2025, 6:59 PM
mediumData ScientistOnsiteStatistics & Math
26
0
Analyze User Transfer Distribution in an Initial Launch Period
A new peer-to-peer payments feature has launched. You are asked to analyze the number of transfers per user during the first 30 days after launch.
Constraints & Assumptions
State the denominator: all eligible users or only activated P2P users who made at least one transfer.
Treat transfer counts as nonnegative integer data.
Early launch data may be zero-inflated, right-skewed, and affected by partial exposure.
Compare the first month with the distribution after two months.
Clarifying Questions to Ask Guidance
Are we analyzing all eligible users or only users who used the feature?
Is the window calendar days after launch or each user's first 30 days after eligibility?
Are failed, reversed, or fraudulent transfers included?
Is adoption driven by invitations, markets, or payment-method availability?
Part 1 - Expected Distribution
What distribution do you expect for number of transfers per user during the first 30 days post-launch?
What This Part Should Cover Guidance
Zero inflation among eligible users.
Right skew and long tail among active users.
Candidate models such as zero-inflated negative binomial, negative binomial, Poisson baseline, lognormal-like heavy tail, or hurdle models.
Why simple normal assumptions are inappropriate for raw counts.
Part 2 - Mean, Median, Mode, and P95
Estimate where the mean, median, mode, and 95th percentile would lie in that period.
What This Part Should Cover Guidance
For all eligible users, mode likely zero and median possibly zero early in launch.
Mean pulled above the median by heavy users.
P95 far to the right relative to typical users.
Differences when conditioning only on activated users.
Part 3 - Distribution After Two Months
After two months, how and why would this distribution change?
What This Part Should Cover Guidance
Adoption may increase, zero mass may decline, repeat users may grow, and tails may become more visible.
Network effects, seasonality, marketing, and product changes may alter shape.
Cohort-based analysis to avoid confusing older users' longer exposure with behavior change.
What a Strong Answer Covers Guidance
A strong answer specifies the denominator, describes zero inflation and right skew, places summary statistics correctly, and explains how launch dynamics and cohort exposure affect distribution changes over time.
Follow-up Questions Guidance
How would you model transfer count for forecasting server load?
How would fraud or spam users affect the tail?
How would you compare two markets with different launch dates?