Tell Me About a Metric You Built That Changed a Real Decision

Think of a time when an existing way of measuring something at work was missing the point, and you introduced a different metric that changed what people actually did. The interviewer wants a specific, personal story, not a general philosophy of metrics.

Structure your answer around a real situation:

• What decision was being made (or avoided) using the old measurement, and what exactly was that old measurement blind to?
• How you defined the new metric: precisely what it counts, over what population or time window, and how you checked that it actually tracked the thing you cared about before anyone trusted it.
• Who you had to convince, and what happened when the new number and an existing, trusted number disagreed.
• What specifically changed afterward — a decision made differently, a resource reallocated, a project stopped or started — and how confident you can honestly be that the metric caused it rather than merely coincided with it.

What a Strong Answer Covers

• A concrete blind spot in the prior measurement, stated precisely enough that a stranger would understand why it mattered, not just "the old metric was bad."
• The new metric's exact construction — what's in the numerator, what's in the denominator, what population it's computed over — and at least one concrete step taken to validate it before relying on it.
• Your own role versus the team's: did you define it, build the pipeline, sell it to stakeholders, or some mix, and how adoption actually happened.
• Honest acknowledgment of what you can and cannot claim about causing the downstream outcome, including any attempt to check the metric wasn't just capturing noise or a confound.

Constraints and Clarifications

• One real example, told with enough specificity that it couldn't apply to any other project.
• Be ready to go two or three levels deep if asked to define the metric precisely, or to explain a validation step in more detail.
• If the business impact is still unfolding or was partial, say so rather than overstating it — a credible, modest claim lands better than an inflated one that doesn't survive follow-up questions.

Follow-up Questions

• Walk through a moment where the number you built looked healthy even as the real-world thing it stood in for was quietly getting worse — how did you notice the gap, and what fixed it?
• Describe a specific clash between your new number and a KPI the organization already leaned on — how was it resolved, and which side turned out to be closer to the truth?