Explain a p-value to a non-technical stakeholder.

"If the new checkout made no difference whatsoever, how often would random variation alone produce a gap this big? That is the p-value. Ours is 0.04, so about 4 times in 100."

Then the important part: "It does not mean there is a 96% chance the change works. It only tells us this result would be unusual if the change did nothing."

And then the part stakeholders actually need: "The more useful number is the confidence interval. Our best estimate is a 0.6 point lift, and the plausible range is roughly 0.05 to 1.2 points. So the change probably helps, and we cannot yet tell whether it helps a little or a lot."

This question tests communication more than statistics. Interviewers are listening for three things:

Do you avoid the inversion? A p-value is P(data | no effect), not P(no effect | data). Getting this backwards is the single most common statistical error in industry.

Do you move to effect size? Stakeholders make decisions on magnitude, not significance. An analyst who stops at "it's significant" has not finished the job.

Do you avoid jargon without being condescending? "Null hypothesis" adds nothing here. Neither does over-explaining.