How do you determine whether an analytical finding is actionable?
Many correct findings are useless. Actionable means a specific person can do a specific thing differently as a result, and the change is worth making — which is a much higher bar than statistically interesting.
There is a lever and an owner — a named team can actually change the thing the finding points to; "customers in rural areas churn more" is only actionable if someone can alter pricing, service or targeting.
The effect is materially large — the expected impact clearly exceeds the cost and disruption of acting, which is business significance rather than statistical significance.
The evidence survives challenge — the pattern holds after checking segments, confounders and data quality, because acting on an artefact is worse than not acting at all.
The timing still allows action — an insight arriving after the budget is committed or the season has ended is history, not a recommendation.
It names the next step — the finding converts into a concrete proposal with an expected effect and a way to measure whether it worked, not merely an observation.
Key terms: actionability, business significance, effect size, confounder, lever, ownership, recommendation, decision window
The clearest way to answer is with the test question "who would do what differently tomorrow?" — it is concrete, memorable, and immediately shows you evaluate findings by their consequences. Draw the explicit contrast between statistical and business significance, since that is what the interviewer is usually probing and many candidates conflate them.
Have an example of a finding you deliberately did not escalate because the effect was too small or nobody owned the lever; that demonstrates judgement rather than eagerness. Expect the follow-up "what do you do with an interesting but non-actionable finding?" — document it, flag it for the roadmap or propose the instrumentation that would make it actionable later.