We are running campaigns which total approx 90,000 keywords and we are trying to test the optimal structure for the campigns for quality score, ctr, conversion rate, relevancy etc. We do all the usual a/b testing on landing page and ad text etc etc etc. Currently we have tightly focused adgroups with a few keywords per group BUT we put all the keyword variations relating to those keywords in the same adgroup (i.e. plurals, singular, mispellings etc.) We have been reading much about the 1 keyword per adgroup theory and the resulting improvment in CTR, and conversions etc - so we decided to test this on a sub set of our keywords (i.e. literally 1 keyword per adgroup and a different adgroup for each variation - mispellings, plurals, singular etc). So the results are in after 60 days trial - CTR is down on this test section from an average of 4.3% (on original campaigns) to 2.2%. Converions are lower and quality score is non changed (i.e. still 'great' and still same minimum bid). Outcome - we can't see value in the 1 keyword per adgroup approach. Although we do not wnat to dismiss it. Questions - are we doing it wrong - have we just plain missed something? For all of you that support the one keyword per adgroup theory - why? and what results have you had? How do you manage the volume of adgroups that you end up with on large scale campaigns? do you hit any limitations from adwords? appreciate all of your input, especially those who are used to large scale campaigns and have tested similar situations. Thanks
golles, a couple of general comments followed by specific answers to your questions. I'm an advocate of the one keyword per ad group in most cases. There are times it makes sense and times that it doesn't. You have to look at everything on a case by case basis and make a decision that's best for the specific situation. The only way to really provide specific recommendations would be to look at your account. That's not an option (you're not offering and I'm not asking ) so I will offer some general feedback and things you may want to consider. The CTR drop could be for a number of reasons; - ad groups now competing with other ad groups (ie singular competing with plural) - lack of ad testing. I've found that "catch all" ads that work for a group of keyword can be less effective when used on individual keywords. I've found similar results when moving between match types. - you mention "average" which means in some cases CTR could have increased and in others decreased? - variances in the average position. - use (or not) of keyword insertion. - match types and ad group structure. ie if you had 10 broad matched keywords in a single ad group with 100 negatives and split each into it's own group and simply copied the original 100 negatives the resulting 10 groups may be less targeted. There are lots of other factors that could come into play as well...the above is just meant to kick start the troubleshooting process. Using one keyword per ad group is a completely different approach and needs to be built out in a completely different fashion. It's not applicable to every campaign. It may be it's just not right for what you're trying to do. I manage campaigns where this approach doesn't make sense or would be overkill and in those cases I use a more traditional ad group structure. Maximum control in relation to bidding, budget allocation, data analysis, ad targeting, etc. From a management point of view it is often more work but to me it's been worth it in terms of ROI benefits. Typically with the AdWords Editor. In some cases yes, I have hit limitations. In those cases I end up slitting the ad groups across multiple AdWords accounts. I guess overall my advice would be to run with what works. If you're hitting homeruns with 20 - 30 keywords per ad group, by all means, keep hitting them I would encourage you though to continue testing this approach where and if you feel it could be beneficial. I've personally seen dramatic improvements in my ability to successfully manage and improve campaign performance using this method in a variety of campaign types.
wow- thanks for such a detailed response - that is really great stuff to start working on. lack of ad testing - we are guilty of this one to a certain extent. We do split test but usually ad variations. I will pick up and run with the suggestions, test, and let you know how we progress. Thanks again - this is a great help.
No problem. I struggled with this approach (and sometimes still do) for well over a year before I felt like I had it where I wanted it. The most important part in my opinion is knowing when it's the right approach to take. That knowledge comes with experience and testing and I have not found a way to shorten that learning curve. Anyway, I hope some of my questions/comments helped spur some additional thoughts and ideas. Keep us posted if you can and best of luck.