Yeah, we have perhaps 4 quite different customer types [student, researcher, research group/lab and student classes/lab]. Currently we have a single lab tier which was designed for research groups but it seems a number of universities want to use that tier for their classes. We didn't expect this use-case and the "per seat" pricing in this tier doesn't work for this case.
We've run the numbers and generally the two lab based use-cases look like so:
Accounts used for research contain: - a low number of experiments - a high number of participants/results - long(er) "active" time per experiment
Accounts used for teaching: - a high number of experiments - a low number of participants/results - short(er) "active" time per experiment
The conclusion we've reached is that we need tiers based upon n experiments and n participants/results (with options to increase both). This would allow lab licensees to add/remove/transfer accounts at will.
I'd still be interested to hear from others who've moved from "per seat" to other metrics.