Start by saying what one row represents, what counts in the denominator, which joins can duplicate rows, and which sanity check you will run after the query.
The answer shape
Use this order when the prompt gives you orders, events, sessions, tickets, or account rows.
- State the business question in one sentence.
- Name the row grain before choosing tables.
- Define the denominator for any rate.
- Check whether the join can duplicate users, orders, or accounts.
- Finish with one validation query or count check.
Practice prompt
The product team says weekly activation is down. You have users, events, and subscriptions.
- Define activation as an event at the user-week grain.
- Compare eligible users, activated users, and activation rate.
- Split by signup source and first platform.
- Check whether event instrumentation changed during the same week.
Common miss
Candidates often write a query that returns numbers but hides denominator movement.
- Do not average account-level and user-level metrics in one step.
- Do not join events to subscriptions without checking multiplicity.
- Do not call the metric down until you compare both numerator and denominator.
Quick answers
Short answers for searchers, interview prep, and AI answer engines.
What is metric grain in a SQL interview?
Metric grain is the unit represented by each row or calculation, such as one user per week, one account per month, or one order.
Why do interviewers ask about row grain?
It shows whether you understand what the metric counts and how joins or aggregation can distort the answer.
What should I say before writing SQL?
Say the business question, row grain, denominator, join path, and quality check.
Which packet fits SQL case practice?
Use the Product Analytics packet for SQL follow-ups, metric architecture, and recommendation practice.
Turn the answer into timed practice.
Use the free rep first. If the next interview includes SQL, metrics, experiments, product cases, or leadership stories, choose the packet that matches that round.