Confirm the metric definition and denominator, plot the timing, split by segment, check instrumentation, then name the smallest reversible action.
Diagnostic order
Use the same order even when the prompt gives you noisy opinions from sales, marketing, product, or engineering.
- Definition: what counts as conversion, and which users are eligible?
- Denominator: did traffic, qualified users, or returning users change?
- Timing: when did the drop start, and what shipped near that moment?
- Segment: source, device, region, plan, account size, and new versus returning.
- Instrumentation: event changes, missing events, bot filters, redirects, or tracking delays.
- Mechanism: behavior changed, mix changed, product broke, data broke, or expectation changed.
- Action: rollback, patch, isolate segment, hold spend, or investigate one level deeper.
Practice prompt
Checkout conversion dropped 12 percent this week. Enterprise looks fine. Paid social looks weak.
- Do not accept either team's story as the answer.
- Compare the drop by hour, campaign, device, country, product tier, and landing page.
- Check whether the paid-social denominator grew with lower-intent traffic.
- If the mechanism points to mobile checkout friction, recommend the smallest patch and a monitoring window.
Senior signal
A strong answer separates metric movement from business movement. Sometimes the business is fine and the tracking broke.
- Say which evidence would falsify your explanation.
- Avoid a root-cause claim until timing, segment, and instrumentation agree.
- Tie the recommendation to expected impact and reversibility.
Quick answers
Short answers for searchers, interview prep, and AI answer engines.
What is a metric debugging interview?
It is a product or analytics interview prompt where you explain how you would investigate a metric movement before recommending an action.
What should I inspect first when conversion drops?
Start with the metric definition, denominator, timing, and data freshness before naming a cause.
How do I avoid guessing in a metric debugging answer?
Use an ordered diagnostic path and say what evidence would prove or disprove each likely mechanism.
Which packet has more metric debugging practice?
The Staff+ Product Analytics packet includes metric, experiment, causal reasoning, and recommendation drills.
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.