On Sunday Night Football, the Los Angeles Chargers trailed 17-14 in the third quarter when they faced 4th and 1 at their own 16 yard line.
In the history of Professional Football, it is unlikely that anybody in the 3rd quarter has EVER gone for it in this situation.
The Los Angeles Chargers went for it.
They failed to convert.
The Las Vegas Raiders kicked a field goal 4 plays later ... three free points for the Raiders in a game that ultimately went to overtime.
Twitter, predictably, exploded ... 'HOW STUPID IS THE COACH OF THE CHARGERS??'
There's a bot on Twitter that calculates the odds of a team winning by going for it vs. punting in that situation. The bot said that the odds of winning were GREATER by going for it vs. not going for it (in a situation where not one coach likely ever went for it in the history of the NFL).
Twitter, predictably, would respond 'I DON'T CARE, THAT'S A STUPID DECISION'.
When you have data that goes against mainstream decision-making, you're going to have a hard time convincing anybody to make a decision that goes against mainstream decision-making.
Now think about the thesis that customer acquisition programs are more important than loyalty programs. All the data in the world tells you the latter is more important. Simulation data tells you the opposite. The information available to us changes, and we likely change (over time) as a consequence.