I worked on a reporting/dataviz project a few years back, showing survey results from states/districts/schools. All the data was included in the totals, but if a school had 5 or fewer responses, we didn't allow viewing of the school data. It was specifically because being able to see answers to things like "do you support your principal?" when there were only, say, 3 teachers at a school, and seeing 2 for "no"... it was way too easy to determine who the respondents were. Even with '5' as a cutoff, it still felt a bit dicey for some schools.