I'm look for some generalized guidance. Let's say I have 200 stores. I test 10 of these for a month doing some sort of specialized promotion. I want to see if it worked. Let's say my metric I'm testing is a lift in sales from the previous year.

My question comes in as to the resolution I should be looking at. I can look at all of the test stores as a group over the month and come up with 1 value. I can also look at each of the stores individually over the month and come up with 10 values. I can also look at each store each week and come up with 40 values. I can look at each store each day and come up with 300 values.

If I'm trying to compare the difference of the means between my control and test and find significance, which distribution makes sense? What should I be looking at ? A distribution of one provides no S.D. but 300 doesn't seem to makes sense either. Thanks for any generalized guidance on what I should be considering.