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 (1)
Branches
Subtopics
Access
Free onlyQuestion type
Type to find a company, then pick it from the list below.
Tool
18 questions
Sample size for a 2% lift
Compute the per-arm sample size for a conversion test.
Calculation · Design · 8 min · Medium
Experimentation & A/B Testing · Original Uplift question
Why peeking inflates false positives
Explain the statistical cost of checking results daily.
Quick check · Readout · 5 min · Easy
Experimentation & A/B Testing · Original Uplift question
Design metrics for a checkout redesign
ProChoose primary, secondary and guardrail metrics and defend them.
Written case · Guardrails · 20 min · Medium
Experimentation & A/B Testing · 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
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
Sample size for a 1-point conversion lift
Baseline conversion is 10%.
Calculation · Design · 8 min · Medium
Experimentation & A/B Testing · Original Uplift question
False-positive risk with four variants
ProYou run one control against four independent treatment variants, testing each at α = 0.05.
Calculation · Multiple testing · 6 min · Medium
Experimentation & A/B Testing · Reported interview question
Guardrails for a checkout change
ProYou're testing a one-click checkout that you expect to raise conversion.
Written case · Guardrails · 15 min · Medium
Experimentation & A/B Testing · Company-context scenario
Wins in every segment, loses overall
ProYour treatment beats control in BOTH the mobile segment and the desktop segment, yet loses when you pool all users together.
Quick check · Readout · 6 min · Medium
Experimentation & A/B Testing · Company-context scenario
A 50/50 split that came out 52/48
You designed a 50/50 experiment.
Quick check · Data quality · 6 min · Medium
Experimentation & A/B Testing · Company-context scenario
Choose the OEC for a homepage change
ProYou're testing a new homepage row on Netflix.
Written case · Guardrails · 15 min · Medium
Experimentation & A/B Testing · Company-context scenario
What CUPED buys you
ProYour experiments are underpowered and you can't get more users.
Quick check · Design · 7 min · Hard
Experimentation & A/B Testing · Company-context scenario
A PM wants to ship the moment p < 0.05
A PM watches the experiment dashboard daily and wants to ship as soon as it crosses p < 0.05.
Written case · Readout · 12 min · Medium
Experimentation & A/B Testing · Company-context scenario
Measure surge pricing's effect on driver retention
ProUber is considering a surge-pricing policy that changes marketplace conditions for every driver and rider in a zone.
Written case · Design · 25 min · Hard
Experimentation & A/B Testing · Reported at Uber
Triage a sample-ratio mismatch
ProA nominal 50/50 Meta messaging experiment reports 520,000 control users and 480,000 treatment users.
Written case · Readout · 20 min · Hard
Experimentation & A/B Testing · Reported at Meta
Test a new shoppable-ad ranking model
Meta has a new ranking model for shoppable-feed ads.
Written case · Design · 20 min · Medium
Experimentation & A/B Testing · Reported at Meta
Spot a novelty effect in a VR interface test
A new VR interface lifts daily use by 18% in days 1–3, 7% in days 4–7, and 0% in days 8–21.
Quick check · Readout · 7 min · Easy
Experimentation & A/B Testing · Reported at Meta
False-positive risk across four variants
You compare four independent treatment variants with one control.
Calculation · Design · 10 min · Medium
Experimentation & A/B Testing · Reported interview question