After analyzing survey results for a decade, I've noticed that most people trust numbers way more than they should. I once saw a well-designed poll with a margin of error of 3% that was completely wrong because the sample was drawn from an online panel that underrepresented older voters. But does that mean we should disregard polls altogether? I don't think so—they're useful if you understand their limitations, yet we keep treating them as gospel in public discourse. Are we underestimating the value of a good statistic just because we see too many bad ones?