Red Flags to Look Out for When Joining a Data Team

Eugene Yan's latest post reads less like career advice and more like a pre-flight checklist for anyone about to sign an offer letter at a data team. His core argument is simple: the interview isn't just the company evaluating you, it's your one real chance to evaluate them, and most candidates waste it asking softball questions instead of digging into the stuff that actually determines whether the job will be miserable.

The data itself is the first tell. If a company has no data, or has data scattered across five schemas nobody bothered to reconcile, you're not doing data science, you're doing plumbing. Yan points to startups without enough customers yet, and B2B outfits in healthcare or fintech that lean on partners for data, as classic setups where a new hire spends year one just wrangling pipelines instead of building anything. He suggests asking bluntly what objects the systems actually generate, how many rows show up per day, and how you'd even get access as a new joiner.

He's just as skeptical of teams without a real roadmap. Yan recalls interviewing a director of data science who, when pressed on his 12-month plan, talked about headcount growth and tech-stack upgrades but couldn't say a single concrete thing about customer or business impact. That vagueness isn't harmless: engineering and product teams can point to shipped features, but data science can drift into open-ended research that never lands anywhere, and those are the teams that get gutted first when budgets tighten.

Titles lie too, according to Yan, and he's got receipts. Facebook's data scientists were doing analyst work as far back as 2017, and Lyft literally renamed its analysts to data scientists in 2018 to win talent wars. So a