The pattern you are describing is real, and it is worth naming plainly: hiring managers often treat technical depth and organizational fluency as opposing forces, not complementary skills. One interview rewards your time in messy, political environments. The next treats your formal math background as a signal that you will be rigid, theoretical, or slow to ship. Neither perspective is grounded in evidence about your actual capability, but both shape the outcome of your search. That is not a failure on your part. It is a reflection of how loosely many teams define the data scientist role in the first place.
What this means for you in practical terms is that you are not imagining a contradiction between interviews. Different managers are optimizing for different problems. A leader at a company with heavy stakeholder friction may have been burned by someone who could build a model but could not navigate a roadmap debate. Another manager, one building a more research-forward team, may have watched a polished communicator struggle to move past A/B testing into causal inference. Neither is right or wrong in the abstract. They are hiring for their own context, and they are projecting a false trade-off onto you. Your job is not to fit a uniform mold, because none exists. Your job is to diagnose the context before you decide how much of yourself to reveal.
That means the interview is not just a test of your skills. It is a two-way calibration of fit. Before you walk in, ask yourself what kind of environment you actually want. Are you looking for a place where cross-functional influence is the core of the job? Then lead with the political experience. Are you looking for a role where deep modeling rigor is the priority? Then your formal math background is an asset, and you should not hide it. But do not assume that one strength cancels out the other in the eyes of a thoughtful interviewer. A thoughtful interviewer will see both and ask how they combine. The ones who do not are telling you something important about their team.
The practical takeaway is this: stop trying to be everything to every interviewer. Instead, treat the interview as a discovery process where you are both evaluating whether the role matches the real shape of your work. If a hiring manager profiles you as one-dimensional, that is not a signal about your value. It is a signal about their limited view of the role. You cannot control which manager sits across the table, but you can control whether you walk away having learned what they are optimizing for. Ask direct questions about the last time a data scientist had to navigate a disagreement with a non-technical stakeholder. Ask about the types of problems the team takes on and whether the expectation is to build new methods or apply existing ones. Their answers will tell you more than any generic pitch about culture. Then decide if their version of depth matches yours. That is the only filter that matters.