say I have a banner rotation tool, and i want to test A-Z banners.
Would it make more sense to test A/B for a couple of days, then take the winner (e.g. A) and test A/C, and so on down the line, or, should I run A/B/C/D.../Z concurrently for a longer period of time and slowly weed the non performing banners (in terms of clicks) out? I would imagine that A/B ->A/C would yield more of an incremental return (since we always keep the higher performing banner in play throughout the whole testing phase) vs A/B/C/D..Z at the same time which would take longer to find the clearer performer? Can someone validate or dis-validate that theory?