How do you test evolutionary hypotheses without overinterpreting cross-cultural data?

I work on behavioural variation across societies and keep running into the same tension: cross-cultural datasets are rich but noisy, and it's easy to read adaptation into patterns that could be sampling artefacts.

I'd like to hear how others decide when a cross-cultural result is strong enough to support an evolutionary claim.

What checks or pre-registration steps do you use before treating a pattern as evidence for selection?