A coefficient of 0.7 means a one unit increase in that predictor multiplies the odds of churn by e to the 0.7, about 2.0, holding the other predictors fixed. It is a change in log odds, not in probability. The effect on probability depends on where you start, since the logistic curve is steepest near 0.5 and nearly flat at the extremes.
Why interviewers ask this
Almost everyone says odds ratio and stops. The interviewer wants the second half: that the probability impact is not constant, and that holding everything else fixed is a real assumption which breaks when predictors are correlated. They also want to know you would check whether the feature was standardized, because a one unit change means nothing until you know what the unit is.
How to structure your answer
- Convert the coefficient to an odds ratio out loud.
- Say clearly that this is odds, not probability.
- Explain why the probability effect depends on the baseline.
- Flag the units and the holding others fixed assumption.
Example answer
Exponentiate it. E to the 0.7 is about two, so a one unit increase in that predictor roughly doubles the odds of churn, with everything else held constant. The part I always say next is that doubling the odds is not doubling the probability. If a customer sits at a 2 percent churn probability, doubling the odds moves them to about 4 percent. If they are already at 50 percent, it moves them to 67 percent. Same coefficient, very different real world meaning, because the logistic curve is not linear in probability. Then two caveats. One unit of what? If I standardized the feature, that is one standard deviation, so I would restate it in the original units before telling any stakeholder anything. And holding everything else fixed is doing a lot of work: if this feature is tangled up with tenure, no customer actually moves along this axis alone, so I would rather show predicted probabilities at realistic feature combinations than quote the raw coefficient at all.
Walking into this interview soon? GhostPilot listens to your live call, spots the question the moment it is asked, and puts a structured answer on your screen in real time. Try it on your next mock, or grab a $29 Session Pass, no subscription, for the real thing.
See how it worksFollow-up questions to expect
- How would you turn that into a partial dependence style explanation?
- What happens to the coefficient if you add a highly correlated feature?
- How do you check whether the logistic model is well calibrated?
Related data scientist questions
Your interviewer will ask their own version of this. Paste your actual job description into the free Question Predictor and get the 20 questions that role is most likely to ask, with what each one is really probing.
Predict my questions