Agree on the decision and the success metric first, then propose the rule as a baseline the model has to beat. Build the heuristic in a day or two, measure it, and show the number. If a model beats it by enough to justify the maintenance, build the model; if it does not, you have shipped value in a week and kept the option open.
Why interviewers ask this
This probes whether you are a scientist or a model vendor. Interviewers want to see you challenge the request without being difficult about it, and specifically want the baseline framing, because that turns an argument about opinions into a measurement. Mentioning ongoing maintenance cost shows you think about the two year total rather than just the launch.
How to structure your answer
- Redirect from the method to the decision being made.
- Propose the simple rule as an explicit baseline, not an alternative.
- Commit to building and measuring it quickly.
- Name the maintenance cost a model carries.
- Leave the door open if the model wins on the number.
Example answer
I would not argue about whether it needs machine learning, because that is an argument about taste. I would ask what decision this drives and how we would know it worked. Then I would say let us make the rule the baseline. Give me two days, I will build the heuristic, we will measure it on the same metric, and if a model beats it by enough to be worth it I will build the model happily. That reframes it from me saying no into us running a test. I did exactly this on a lead scoring request. The rule was three conditions on company size, page visits, and demo requests. It captured most of the achievable lift, and the gradient boosted version I built afterward added a few points of precision. The team weighed that against the cost of a retraining pipeline, monitoring, and someone owning it, and chose the rule for a year. It shipped in a week instead of a quarter. I also point out that a model is a permanent commitment: it drifts, it needs monitoring, someone gets paged. A rule does not.
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 much lift would justify the model in that example?
- What if the stakeholder wants the model for external credibility rather than accuracy?
- How do you keep a heuristic from rotting over time?
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