What pitfalls would you watch for when running experiments on healthcare workflows or clinician-facing products like those at Truveta?
