How Market Research Tricks Leaders into Making The Wrong Decisions

Jon Cohen spent decades calling elections before founding an AI company built to fix his own industry.

I sat down with Jon Cohen, founder and CEO of TrueDot.ai, who spent his career running polling at ABC News, the Washington Post, and Pew before becoming chief research officer at SurveyMonkey. We talk about why he never let himself get promoted out of the actual work, what it felt like calling presidential elections on deadline, and why he thinks executive fit depends on leaders getting real data instead of numbers designed to look good.

Jon walks through the Michigan Senate primary that upended pre-election polling, and the idea he coined to explain it: plausibility is not validity. We get into why so much market research is built for number generation instead of truth seeking, how leadership coaching programs and executive coaching for leaders could benefit from the same kind of rigor pollsters apply to their own data, and why checking numbers by hand has quietly been the way researchers train judgment for generations. Jon also talks about what happens to that training pipeline as AI takes over the checking work itself, and where the next generation of deep expertise actually comes from.

This conversation also gets into culture fit inside research teams, the discipline of asking a genuinely useful question instead of an easy one, and why leadership fit so often comes down to whether a leader is willing to hear the truth instead of the number that makes the trend line go up.

Key takeaways:

Jon built a system to ingest historical polling data because scarcity forced him to do more with less at both the Washington Post and ABC NewsThe Michigan Senate primary showed how pre-election polls and prediction markets can all be wrong in the same direction at oncePlausibility is not validity: a number can sound reasonable and still be wrong, and leaders need a system for testing which one they are looking atMost market research is built to keep a trend line moving up and to the right, not to generate genuine insightChecking numbers by hand has traditionally been how researchers develop judgment, and AI tools that automate that step raise real questions about how the next generation learnsJon's advice against choosing a career too early shaped a path that took him from a history degree to programming to running polling operationsExecutive coaching for leaders and leadership coaching programs benefit from the same rigor Jon applies to survey design: know what you wi