Good Data Scientist, Bad Data Scientist

Summary

Whitestone adapts the “Good PM / Bad PM” format to data science, producing a reference-quality list of professional behaviors that separate effective data practitioners from mediocre ones. Core mental models:

1. Problem Obsession Over Tool Obsession. Good DS is obsessed with solving business problems, then selects the right tool. Bad DS picks a technology and goes looking for problems to fit it. Start with the business, not the method.

2. Organizational Glue. Good DS bridges disciplines — connecting business and tech organizations — and becomes a central node in information exchange. Bad DS is a side node, easily ignorable. The best practitioners are connectors, not specialists in isolation.

3. Ship Early, Iterate. Good DS starts simple, ships, then iterates. Bad DS starts with the most advanced technique. Good DS understands marginal return on effort — whether that 1% accuracy lift will actually change any decisions.

4. Push Over Pull. Good DS actively proposes work, thinks beyond what stakeholders ask, and generates new ideas for how data can add value. Bad DS operates purely as a question-and-answer service. The difference between being a cost center and a value driver.

5. Product Knowledge Is Non-Negotiable. Good DS has deep understanding of the product, the business model, and how their team contributes to the P&L. They use the product, test it, talk to customers. This context is what turns numbers into insights and insights into recommendations.

6. Audience Calibration. Good DS adjusts messaging based on audience. They know when to provide context and how deep to go. Bad DS delivers the same message regardless. Communication is the last mile of analysis.

Relevance

• 06-reference/2026-04-03-analytics-engineering-everywhere — The analytics engineer role embodies several of these traits: bridging, shipping, iterating. But Whitestone’s framework applies regardless of title.
• 06-reference/2026-04-03-embrace-the-grind — “Good DS gets their hands dirty” is the grind principle applied to data work. No task is below you if it generates insight.
• 06-reference/2026-04-03-reforge-why-analytics-efforts-fail — Most analytics failures map to “Bad DS” behaviors at the organizational level: tool obsession, pull-only service model, disconnection from the business.

Open Questions

• How do you interview for “problem obsession” vs. “tool obsession” in a 45-minute conversation?
• Is the “push over pull” model sustainable without executive sponsorship? At what org maturity level does proactive analysis get rewarded vs. punished?