I'm increasingly of the opinion, after having to deal with biologist data format hackery for many years (they have no shame!), that every biologist should be taught relational database normalisation. That would make them actually stop and think about the structure of the data they're recording and how to make it computationally tractable. Even more, they should not even start their experiment until they've cleared their data recording structure with a bioinformatician.

Then we ban them from using flat files. Related tables or nothing. I've wasted too much time, and they're not competent enough to work with JSON/XML/Excel.