The weaker the theory, the more robust the results. The stronger the theory, the more meaningful the results.

When you use only a weak theory (weak assumptions), the problem in the analysis is: am I measuring the right thing? When analyzing an experiment/difference-in-difference/synthetic controls/etc, one of the main questions is: am I looking at the right outcome variable?

When you use a strong theory (an explicit model, strong assumptions), the problem in the analysis is: are these assumptions true? But the problem of measurement is less difficult because the theory itself proscribes the right metric.