> I'm far from a statistician, but the difficulty in the study seems to be that you have products which are radically different from each other in demand structures.

>How do you tell whether the sales delta is driven by DRM or a much, much higher quality artist?

I'm not going to weigh in on DRM vs. non DRM, but you'd get around that problem by explicit randomization. If you have enough observations then differences in quality, etc., literally average out. If you have more information up front you can randomize conditional on the observed characteristics, which reduces the number of observations you need.