I see this as a common problem in #evaluation too - premature focus on statistical significance of differences in means without checking distributions (bi-modal? skewed?) or bivariate or multivariate patterns. Something I used to emphasise when teaching statistics to students.