…I don’t think granularity has to always be at the expense of testing validity. If some of your headlines, ad descriptions and display URLs are the same across multiple ad groups, why not download an ad text report and pool data across ad groups. Yes, it might not be as insightful as looking at each ad group’s keywords and ads in isolation, but surely it’s no worse than having multiple different keywords in the same ad group?

I think being granular is slightly more important, so I make that my first priority. Then I test as best I can within the smaller groups. I’ll A/B test ads or LP’s common among the smaller adgroups to see if there is any significance as a whole or within the sub groups.