The Number Isn't the Argument
A statistic can be completely accurate and still be doing no real work in the argument it's attached to.
There's a specific kind of confidence that shows up whenever a number enters a conversation. Engagement is up fourteen percent. Retention improved after the redesign. The number sounds like it settles something, and often the conversation moves on as if it has.
But a number is only as strong as the argument connecting it to the conclusion being drawn from it. Engagement can rise for a dozen reasons that have nothing to do with the change being credited. Retention can improve for a cohort that isn't representative of the customers who matter most. The statistic can be entirely accurate and still be doing none of the work it's being asked to do.
This is why the useful question is rarely 'is the number correct.' It usually is. The useful question is 'what else would have produced this exact number,' and whether the argument has actually ruled those alternatives out.
People who are good at this aren't more skeptical of data in general — they're more specific about what a given piece of evidence can and can't support. They ask what was measured, what wasn't, and who was excluded from the sample before they ask what the number says.
Treat a number as the start of a question, not the end of one, and most misleading arguments lose their footing before they finish the sentence.