Comments on: Search Queries & Quality Score – The Truth (Amended)

By: Craig Danuloff

Craig Danuloff — Sat, 30 Apr 2011 02:10:00 +0000

I’m very sure/comfortable that I’m getting honest answers. There is endless complexity, and they have declined to answer a few things, but there are a number of issues – i really should count them – where to my knowledge things were cloudy but they have agreed to bring some sunshine. I have no doubt the book is imperfect, but I also know it moves the ball quite a ways down field, and the Google help was/is a big part of that. Anxious to hear what you think when you get to read it. – Craig

By: Craig Danuloff

Craig Danuloff — Sat, 30 Apr 2011 02:07:00 +0000

Jeremy – I see what you’re saying. Actually, since eligibility is driven by quality score, the % of the time that you’ll get the impression – from the broad match or the exact match – depends on how the keyword does against that query. It’s likely that the exact match will do better, but not assured. If ‘cheap dog food’ get a good enough CTR for the ‘dog food’ keyword, the quality score would likely earn eligibility quite frequently. nnThe key is that we don’t/can’t know. Broadly I agree with your point that any good query should be ‘promoted’ into it’s own keyword.

By: Jeremy Brown

Jeremy Brown — Fri, 29 Apr 2011 21:07:00 +0000

I’ll likely buy the book, but I wonder what percentage of what they told you was correct. You are reliant on them saying ‘this is the truth’ when they have been quite cloudy on many of these issues.

By: Jeremy Brown

Jeremy Brown — Fri, 29 Apr 2011 20:44:00 +0000

That’s based on a questionable assumption: that Google is showing you in the auction for a large percentage of all the searches.nnIn fact, if you don’t have a search query delineated as a keyword, it’s likely that Google shows you in the auction less often. You may see a number like 2 clicks out of 100 impressions on the search query ‘cheap dog food’ using the keyword ‘dog food’. However, maybe ‘cheap dog food’ was searched 400 times and Google only showed your ad 25% of the time.nnYour CTR and QS assumptions may be right for those 100 times you were displayed, but those numbers may change when you break that query out into a keyword.

By: Kalin Dudley

Kalin Dudley — Wed, 20 Apr 2011 16:16:00 +0000

Craig,nnCame across this article on Twitter from Dave Szetela, great post and definitely a must for Adwords users to understand. The frustration that I have with how Google is calculating Quality Scores is that it creates massive accounts to manage. nnI have found that because search queries do play into the keyword quality score, you have to create ad groups specific to each unique query that makes sense (relevant to what you offer/enough impressions available for effort). You also have to be sure to have negatives in other ad groups to block the wrong ads from showing per search query. In the example you give, which is great, best practice would be to create an ORGANIC DOG FOOD ad group and a CHEAP DOG FOOD ad group so that your copy/landing pages could be more specific thus hopefully improving CTR and user-experience. But you then must be sure to add the terms CHEAP and ORGANIC as negatives to all other ad groups to prevent the wrong ads from showing. But this also adds another level of complexity, what if I am looking for CHEAP ORGANIC DOG FOOD? Again understanding these dilemmas can help you to use negative match-types to your advantage but you see the frustration with the effort involved…nnUltimately, I am all for QS as I believe it keeps a strong environment that prevents advertisers from raising costs on terms that they aren’t relevant for..nnGreat post!nnThanks again, nnKalin

By: Craig Danuloff

Craig Danuloff — Wed, 20 Apr 2011 03:04:00 +0000

Depends how you mean it, but I don’t think so. Visible quality score is based on the total data set for a keyword (even in mutiple match types and/or geo targeted versions). I’m not exactly sure how they different KW-Ad pairs that exist across the different ad groups play in, but let’s say it’s a weighted average of that.nnActual quality scores (such as the ones used for Ad Rank and CPC, start with that same base of data, but are then tuned/modified based on the real-time query, geo, and other factors. nnIs that helpful?

By: Alan Mitchell

Alan Mitchell — Wed, 20 Apr 2011 02:55:00 +0000

Excellent article Craig. So in it’s simplest sense, are you saying Quality Score for a keyword is the weighted average of the Quality Scores of all search queries which match to that keyword? I guess that would make it extremely beneficial to create a highly granular ad group structure with loads of long-tail keywords.

By: Craig Danuloff

Craig Danuloff — Tue, 19 Apr 2011 20:35:00 +0000

For better or worse, it’s their game, their ball, their field. They make the rules. It can be frustrating – and I think you’re right that transparency would go a long way. When i talk to them about these issues, they make sense and seem reasonable. But with no data or dialog it can be very frustrating.

By: Brian

Brian — Tue, 19 Apr 2011 19:09:00 +0000

The real issue that frustrates me about Google is not the lack of transparency on QS but the lack of transparency on things like pausing a keyword for things like “low search query volume.” I know that Google will treat search queries like “Plano Plumber” and “Plumber Plano Texas” quite differently but my “Plumber Plano Texas” kw has a good chance of the dreaded low search query volume trigger. Your book may expose new strategies, but Google may place additional restrictions (e.g. increasing the threshold for a kw to be live) in order to prevent taxing on the Adwords system that prevent me from executing those strategies. It would be nice if Google gave me the option of buying my way out of low search query volumes (e.g. I will pay a toll if I want to bid on low volume kws) so I could fully utilize my kw knowledge to maximize my overall spend. I hate executing a long-tail kw strategy (built around themes) where 40% of my kws are dead out of the water.

By: Craig Danuloff

Craig Danuloff — Tue, 19 Apr 2011 16:37:00 +0000

Good questions. There is a great range of data that Google may or may not have available when they calculate quality score – the account may be old or new, the keyword may be old or new, the ad copy may be old or new, etc. So I think of it as a ‘best available data’ system that hopes for the most detailed but relies on less specific if needed. So a new keyword may have never had a click in your account, but they do know how other advertisers have done with it, how you do as an account, how the display URL has done, and maybe how the ad copy has done – these plus other clues let them set an initial score even for a ‘new’ keyword. nnWhat you call ‘instant qs’ is really first quality score for ad rank and then quality score for cpc. Neither has much to do visible quality score, they’re fresh real-time calculations which use the whole stack of available data in whatever mix they have to or deem appropriate. I think any % this and % that interpretations are just ideals or guidelines because per the above they have to factor the quality of what they have in any given situation.

By: Anonymous

Anonymous — Tue, 19 Apr 2011 03:21:00 +0000

Great post Craig. I had always assumed there were other relevancy factors and account factors at play when the search query didn’t match the keyword. Your conclusion, though, that you should add the search query as a KW is a smart one. The reality is that if your KW has a low exact match impression share (I wish Google would provide this at the KW level), it’s likely that your ad text isn’t as relevant as it should be and therefore your CTR is lower than it should be. Adding search queries as keywords and refining your ad text to be relevant will help CTR and your QS.nnA couple questions for you:n1) How is visible quality score calculated for KWs that have never had an exact match to the search query? I’m assuming account QS is at play. My guess is that Google applies a QS leveraging their historical knowledge of the performance of that KW on other accounts normalized by some factor that accounts for the difference between your account QS and the account QS of other advertisers. In the case that no data exists, I’d guess that Google weighs your account QS heavily.n2) How is instant QS (QS at the time of search) actually calculated? How much of it depends on visible QS, how much depends on some relevancy factor, and how much depends on account QS?

By: Craig Danuloff

Craig Danuloff — Tue, 19 Apr 2011 01:52:00 +0000

Cool. Need to spend more time back in AdWords and not these fancy 3rd party platforms…nnGiven this, the benefit of splitting them out would be getting to see qualtity score – nor real way to infer that from CTR (although you would get a good idea by comparing to the visible of the current keyword, but that’s a lot of work on an ongoing basis.) Thanks for the new info.

By: Chad Summerhill

Chad Summerhill — Mon, 18 Apr 2011 20:36:00 +0000

You get impression counts by search term in the AdWords Search Term Report, and more recently they’ve been reporting more zero click search terms:nnhttp://www.adwordshelpexperts.com/2011/04/adwords-search-query-report-becomes-mor-transparent/nnUnless I’m missing something (which is possible) you can determine a CTR for a search query. Promoting a search query to a keyword just makes the data more visible in the Keyword Report.

By: Craig Danuloff

Craig Danuloff — Mon, 18 Apr 2011 20:09:00 +0000

I’ve wrestled a lot with the ‘morality’ and ‘business issues’ of what they disclose or don’t. And I tackle the issue in the book head-on. I don’t think that it’s primarily a malicious effort (although they certainly make decisions that benefit themselves) but rather the fact that most of their users only want the simple explanation – a flood of details in the system and long complicated descriptions in the help file, would only confuse the vast majority. In fact their willingness to help on this project was prefaced by their saying they were happy to have a vehicle to put the ‘hard core’ info out to those who would want it. Of course they could have found a way beside this, but given that they haven’t roadblocked very much info with me does suggest it isn’t that info is hiding, it’s that it hasn’t had a forum. I do agree they owe advertisers better, but am glad this project can at least partially serve that function. nnThanks for the comment.

By: Craig Danuloff

Craig Danuloff — Mon, 18 Apr 2011 20:05:00 +0000

Yes, good summary.