I noticed that I was accidentally running the same KW in a few different ad groups at once. While I was consolidating, I noticed that they all had different stats. Now, this could have simply been an effect of the different ad copy, bids, etc., but I wanted to test. I set up 8 different ad groups in the same campaign. Each had the same, single keyword, as well as same settings, and same bids. The only difference between all the groups was ad copy. The destination url was the same throughout. I had some pretty wacky stats after a few days. Not only did the same keyword have different quality scores throughout the different ad groups, but the group with the WORST QS actually ended up bringing in the most impressions. I also would have thought that number of impressions the original KW was getting would have decreased, but this wasn't the case. There was no discernible difference in the original KW's performance, even though I was outbidding it with my new ad groups. Among the new ad groups, some completely suck and haven't gotten any impressions. Of those that have, their performance varies from day to day. See graph: http://i.imgur.com/5PHmn.png Has anyone else tried this or seen similar results? I always thought any given KW could only have one quality score within a campaign. Looks like this isn't the case... Also, why would ad groups with exactly the same settings have such wildly different results? The first two groups I made were control groups, ie EXACTLY the same, and their performance was different as well.
The keywords are affecting each other's performance, in a way that's impossible to analyze the overall performance (of the same keyword in different ad groups). Google's quality score is far from perfection. For example I found that it's actually influenced by the bid, even though theoretically it shouldn't be. Ad Rank = Bid * QualityScore I doubt that keyword-coning can be useful, since it creates complexity and hurts the optimization. Still, if you only care about your short-term performance and this method increases your overall traffic/profitability then it's another story
What you did was no different than having eight ads in the same group. The difference is that now you saw the true QS of each ad. If you ran in the same group, the QS of the keyword would be an average of all the ads. QS is shown at the keyword level but it's really the quality of the ad-keyword combo. Each ad and keyword combo has their own QS calculation. As for the one with the worst QS bringing in more impressions, how many more are you talking about? How many impressions overall? You got to give the system some time to gather enough info to make it statistically relevant. Since you're diluting over eight different ads, and you mention only a couple of days, that may not be statistically valid yet. Keep in mind too that QS being shown to you as an integer, the margin of error is great. The system calculates to many decimal points which you don't see. That QS of 6 may be 6.44 and the 7 actually 6.55 (assuming it rounds up), not all that much of a difference when you look at it that way. As was pointed, ranking is calculated by multiplying bid and QS so the real QS numbers affect that too. So even with a higher bid, your new groups may not be outranking your original keyword. Report back when you have at least 30 clicks for each ad. Personally, I would not advise continuing, at least not having 8 active groups with the same keyword. It's the same as having 8 active ads in the same group. Not something that helps you optimizing in any way. > Google's quality score is far from perfection. For example I found that it's actually influenced by the bid How did you come up with that conclusion? There is no way you could have noticed a significant change of QS when changing bids.
Sometimes (I saw it happening) a significant increase in the bid results in a big shift of average position and higher CTR. This in turn puts your keyword-ad in a "full potential" situation that wasn't there before, with the lower bid. So - yes, sometimes you can gain higher CTR and QS simply by bidding higher. But I wouldn't recommend doing it regularly Obviously when Google calculated QS it normalizes CTR against average position and competitors' CTR. Nevertheless the calculation can not be "perfect" since the equation contains many endogenous variables (i.e. variables which are dependent of each other). "Ad Rank = Bid * QualityScore " is a Reduced Form Equation, so it's only an approximation. Hence there are exceptions to the rule "Bid doesn't influence QS". I think of it as lighting a match - you need to invest an initial energy in order to get it burn. Lucid Web Marketing - perhaps you too launch a new campaign/ad group/keyword with relatively high bids, for the same reason.