What day count would be best for experimenting ad placements after one test and another?? 3-5 days? 5-7 days?
yup, 7 days would give you the best. also, keep in mind that weekends possibly have higher counts if its leisure.
Of the two choices above, the 5-7, but that isn't really going to give results that are very statistically significant. I try to do 30-day tests.
I would say seven should be the minimum, so you cover all the days of the week. But I'd agree with Adpubster, the longer you run the test the better you can get a read on it. If a month is too long for you, then perhaps 2 weeks.
The reason I choose 30 days is that is usually the minimum N of a sample set to derive statistically meaningful results. The other reason is that it catches the entire month (okay, sometimes use 31 lol) so you can catch the advertisers at the full extremes of the pocketbook if you are doing pC tests.