[michaelmior]

I would tend to agree. Although for something like testing ads it seems like it would be relatively straightforward to produce an A/B test that compares the performance of two ads relative t TinyTroupe's predictions.