Catalog
8 branches of analytics, product and AI interview questions
Explore eight skill tracks, called branches, filterable by subtopic, company, tool and difficulty. SQL questions run against sample tables, choice and numeric answers are checked instantly, and written cases are scored against evaluation criteria.
Filters
157 questions · page 2 of 7
Cost per resolved ticket for an AI agent
Compute unit economics for an AI support agent.
Calculation · Product metrics · 5 min · Easy
AI Agents & Automation · Original Uplift question
Metrics to judge a support agent rollout
Choose the scoreboard for an AI support agent.
Written case · Product metrics · 15 min · Medium
AI Agents & Automation · Original Uplift question
Demo-day agent, churning users
ProDiagnose why a high-scoring agent loses real users.
Written case · Evaluation · Python · 20 min · Medium
AI Agents & Automation · Original Uplift question
Why user-level randomization breaks in a marketplace
ProYou want to test a new driver incentive at Uber.
Written case · Interference marketplaces · 20 min · Hard
Experimentation & A/B Testing · Company-context scenario
They want to ship despite a null result
Your A/B test came back flat (no significant effect).
Written case · Stakeholder management · 15 min · Medium
Analytics Case & Communication · Company-context scenario
Explain a p-value to a non-technical exec
A VP asks 'what does p = 0.03 actually mean?' Explain it correctly in plain language, and state clearly what it does NOT mean.
Written case · Communicating stats · 8 min · Medium
Statistics & Inference · Company-context scenario
Which attribution model over-credits early touchpoints
A customer sees a Facebook ad (first touch), later clicks a Google search ad, then converts after an email.
Quick check · Attribution · 6 min · Medium
Digital & Marketing Analytics · Company-context scenario
Running total of daily revenue
From the orders table, return each order_date with that day's revenue and a cumulative running total of revenue through that date, ordered by date.
Live SQL · Windows · PostgreSQL · 10 min · Medium
SQL & Data Modeling · Original Uplift question
Offline vs online evaluation
You're shipping an AI support agent.
Quick check · Evaluation · 6 min · Medium
AI Agents & Automation · Company-context scenario
Reddit traffic is down 5%
Reddit's traffic went down 5%.
Written case · Diagnosis · 18 min · Medium
Product Analytics & Metrics · Reported interview question
Estimate daily rideshare trips in Chicago
Estimate the number of rideshare trips (Uber + Lyft) taken in the city of Chicago on a typical weekday.
Written case · Estimation · 15 min · Medium
Product Management · Company-context scenario
What the Central Limit Theorem lets you do
Why does the Central Limit Theorem matter for A/B testing, where the underlying metric (e.g.
Quick check · Foundations · 6 min · Easy
Statistics & Inference · Company-context scenario
Big week-1 lift that fades to zero
A redesigned button shows +8% engagement in week 1, but by week 4 treatment and control are indistinguishable.
Quick check · Readout · 5 min · Easy
Experimentation & A/B Testing · Company-context scenario
Trailing 7-day active users
ProFrom an events table, for each calendar date return the number of DISTINCT users active in the trailing 7 days (that date and the 6 days before it).
Live SQL · Windows · PostgreSQL · 15 min · Hard
SQL & Data Modeling · Original Uplift question
Facebook third-party login down 2% WoW
ProFacebook third-party login (using your FB account to sign into other apps) is declining 2% week over week.
Written case · Diagnosis · 18 min · Medium
Product Analytics & Metrics · Reported interview question
Risks of using an LLM to grade your agent
ProYour team uses a strong LLM to automatically grade your agent's outputs ('LLM-as-judge') at scale.
Written case · Evaluation · 15 min · Medium
AI Agents & Automation · Company-context scenario
How would you improve Google Maps?
ProHow would you improve Google Maps?
Written case · Product sense · 20 min · Medium
Product Management · Reported interview question
Prove paid social is actually incremental
ProYour paid social campaigns report a 4x ROAS from the ad platform's attribution.
Written case · Attribution · 20 min · Hard
Digital & Marketing Analytics · Company-context scenario
Two dashboards, two different numbers
ProFinance's dashboard says last month's revenue was $2.1M; the product team's dashboard says $2.4M for the same metric and period.
Written case · Data trust · 15 min · Medium
Analytics Case & Communication · Company-context scenario
ROAS looks great, is the campaign profitable
A campaign spent $10,000 and drove $40,000 in revenue (ROAS = 4).
Calculation · Campaigns · 6 min · Medium
Digital & Marketing Analytics · Company-context scenario
Success metrics for Facebook Marketplace
You just launched a feature in Facebook Marketplace.
Written case · Metrics · 15 min · Medium
Product Management · Company-context scenario
Month-over-month revenue growth
Return each month, its revenue, and the month-over-month growth as a percentage rounded to one decimal.
Live SQL · Windows · PostgreSQL · 10 min · Medium
SQL & Data Modeling · Original Uplift question
'Can you pull engagement for me?'
A PM Slacks you: 'Can you pull engagement numbers for me?' with no other context.
Written case · Structure · 10 min · Easy
Analytics Case & Communication · Original Uplift question
Measure the success of YouTube Shorts
How would you measure the success of YouTube Shorts?
Written case · Measuring impact · 15 min · Medium
Product Analytics & Metrics · Reported interview question