I would assumg Google is smart enough to set an intelligent minimum level of conversion data before actually employing the optimization. Does anyone have an educated guess on what the minimum number of clicks and/or conversions an ad must first get before the rotations starts being optimized and does the optimization effect become stronger over time as long as conversion rate does not decline on the ad receiving preference?
I would guess the "strength" of preference is dependent on the difference in the conversion rate between the ads being rotated. I'm just basically wondering, once you fee you've experimented enough on ad variations, is this a reliable "set it and forget it" feature or is it still wiser to rotate envenly and run reports to determine which ads to pause.
An interesting take on this is that some ad variations show pricing. Depending on the keyword their may be competitors advertising lower pricing in their ads. For those keywords we would want the ads to focus on other differentiating factors. Competitor ad positions can change and come and go. So it seems to me, one major benefit of automatic ad display optimization based on conversion rate is that as the competetive landscape changes the system could adjust the ad rotation to that. If the advertisers with the lower price drops out, the ad with the price in it may start gaining ground. But I would expect in a low-volume situation it could take a LONG time for the system to recognize the change - but so would be the case if you were manually comparing the conversion stats. In either case you need a statistically significant amount of data to base the decision on.