This is something that seems to work very well for me, but not sure how many of you also do this. Now I'll assume that most of you correctly cluster keywords that a) are closely themed and b) don't get into your top say, 30% most popular keywords for that campaign (as these are generally placed in their own Adgroup - also known as 1-2-1) Now: Let's say for arguments sake I have 30 keywords in an Adgroup which I'll call 'Widgets Clustered 1' (they are very closely widget related and related closely to eachother), however - after a couple of weeks I find that 20 of the keywords have a Quality Score of 10/10 whilst the other 10 have a Quality Score of 7/10, even though their CTR is reasonable. Google's method of giving a keyword it's Quality Score is more complexed than most of us realise, and sometimes I'm not even convinced they know themselves what QS to give the keyword, due to lack of history. When working with misspelled keywords for example, Google will sometimes fancy one misspelling over another, even if the difference is only 1 letter and it's blatantly obvious that they are both misspellings for the same keyword. Example: widgetss = quality score of 10/10 widgetsd = quality score of 7/10 So what I do is move these 10 keywords which have a QS of 7/10 into a new adgroup (perhaps 'Widgets Clustered 2') and write 2 different ads for those 10 keywords. What sometimes happens is these 10 keywords, will then get given a Quality Score higher than 7/10, but it's all down to testing the Ads. I tend to choose a different display URL, landing page or website. Now, I'm not saying Quality Score is the most important thing known to man when working with Google, but a higher QS means you don't need to bid as high to end up in the same position. Remember Rank = QS * Max Bid. I mean if your ROI is good for that keyword, that's all that matters, but sometimes doing this can improve your Quality Score and lower your spending. It's just something to think about when working with clustered groups. One thing to make sure of and that's only move the keywords to another Adgroup if collectively, they add up to a decent amount of impressions and clicks. Otherwise this new Adgroup will take forever to split test the Ads.
Muchacho79, I like your thinking very much regarding keyword clustering! Too few impressions and clicks to a keyword is definitely a problem for Google to figure out the quality score that the keyword deserves, and at the same time, it will take forever for you to get results from your nicely-prepared split test.