It is difficult, if not inserted in the theme. I will try to be explicit enough.
You must consider which variables modify the response of the experimental unit (in this case the teacher), for example age, or years of experience. therefore assume that there are sampling units that belong to a single group, and within the group the response is homogeneous, so at least you should consider sampling in multiple stages, stratum by age, stratum by experience, simple random within the final group.
this leads to think that it has variances by group, and you should consider stratified sampling