The definition of Quality Score, as provided by Google themselves, seems to almost revolve around this mysterious ineffable quality called relevancy:
“The AdWords system calculates a ‘Quality Score’ for each of your keywords. It looks at a variety of factors to measure how relevant your keyword is to your ad text and to a user’s search query.”
There are relevancy factors in your account beyond the ones used to calculate the on-the-fly Quality Score that comes about each time a search triggers an ad auction. In fact, the conspiracy theorist side of my brain
easily lets me imagine Google using whatever they can to try can calculate relevancy – because ultimately, an Adwords system filled with ads that better match queries produces more satisfied searchers (clickety click!), and earns Google better money.
So tinfoil hats on, let us simply accept that evil future Googles will use any aspect of your account or website that it can get its fuzzy illogical little head around to better match your ads and their queries – but rest assured, your fuzzy illogical wee head is still better at understanding relevancy than Google*.
Long story short, if you apply the perspective of relevancy, in whatever depth you can manage to fathom on a rainy Wednesday afternoon, <really important bit>for the relationships between each of the elements of your Adwords campaign</really important bit>, you’ll be doing more than lip service to your Quality Score, you’ll probably find things to improve. So let’s look at some relationships:
1 – CTR & Creative
CTR is at the core of relevancy because it is the ultimate indicator of it – a clicked ad is a relevant ad. Google keeps track of CTR for both keywords and creatives (your ads). How well the ads you’re serving have been doing click-through-rate wise is vitally important in determining your Quality Score. The ad auction that’s going on really really wants to serve something that has a high likelihood of getting clicked, and what better indicator than that ad’s history?
2 – Creative & Search Query
The more of the user search-query (what was actually put in the Google search box) that exists in your ad-copy or URL, the more likely you are to get clicked, and so this is a realistic relationship Google might look at.
Considering the nature of Google’s traditional search algorithm it only makes sense to see them parse ad-copy on-the-fly, matching keywords to the text within a creative, and rewarding it somewhere, either in the Quality Score calculation or some other mysterious place.
3 – CTR & Keyword
This is another hot tamale – it has been hinted in dark corners that the CTR variable is the grand-daddy of the QS calculation, and in combination with its relationship to the Ad itself, CTR is a function of the Keyword (separated with sticky stuff by match-types). History matters. That’s why PPC black-hats know the value of erasing history.
4 – Creative & Keyword
This is basically up to you – you’ve heard by now to maintain only small, tightly themed ad-groups yes? Well this is one of the reasons why that technique works so well – the less keyword spread you have in the ad-group, the better shot you have of matching directly your creative copy and your keywords. If you can get direct matches here (and keep in mind, you can write a lot of creatives, Mr. Lazy), you’re heaps ahead of the crappy Quality Score ignorant guy in the cubical that smells.
5 – Keyword & Search Query
What are you kidding me? This is the beating heart – the closer you get to real search queries with your keyword selections, the better you’ll do. Okay so admittedly, match-types account for huge amounts of variance here, but what do you think intuitively? Would an exact-match-type keyword, that matches a query exactly, be likely to get clicked? Yeah, that means good QS. Would a broad-match keyword that barely matches a search query be likely to be clic… oh wow people do I really have to finish this paragraph? Didn’t think so. Thanks. I was getting bored.
6 – CTR & Landing Page
This is getting a little more out there – I know Google is keeping track of load time of landing pages, and so logically they can keep independent stats on them, but I would say they would be more likely to keep track of abandonment rate (back-clicks or other measures at their disposal) when assessing landing pages. Too many things mitigate the relationship between landing pages and any boiled down CTR average they might come up with for it, but it is possible that Google could calculate something that resembles it, if just for shits and giggles.
7 – Landing Page & Keyword
This is the most important real deal relationship for landing pages and you know it. I often battle with how much actual text may make it on to a landing page, because honestly, sexy graphics are sexy, and sex sells better than text sells, though text still sells pretty wells. That said (don’t ask me why it was said), if you haven’t learned by now that including the keyword on your landing pages is vital, go back to PPC 101, collect 200 dollars, send it to me, and then… well actually I’m not so interested in what you do after I get that cheque.
What become interesting are the extensions – does it matter if the keyword is in the landing page meta title? How about the meta-description? Is it good if it links to other information rich pages? How text-hungry is Google when I’m just trying to sell a freakin’ waffle maker anyway? (I know, I know, Sue’s All-User-Generated Waffle Maker Reviews Social News Voting Communal Community Secret Society Website is going to whip my butt with all the Quality Score pouring out it, it’s inevitable).
The landing page also has a relevancy relationship with the Ad-copy, mitigated by the keyword, and also likely by the search query at hand.
Some even go so far as to suggest that a well linked-to page (a traditional indicator of relevance of pages in Search Engine Optimization, something Google, you know, understands), has a better chance of being ‘relevant’ which could bleed over to quality score. The thousands of people who host orphaned PPC landing pages may beg to differ, however.
8 – Ad-Group & Search Query
Adwords has to scour through your campaign to find the ‘best’ Ad-Group for a search query, and then scour that Ad-Group for a Keyword that matches well enough, so long as it’s triggering an ad creative that stands a decent chance of being clicked.
The smaller, tighter and more concentrated (in meaning, or searcher-intent) your Ad-Group is, the easier it is to write a comprehensive set of ads to cover all of the keywords in that group. The better the match between search query and a high performing ad-group (CTR anyone?), the better chance of profit for all.
9 – Keyword & Ad-Group
Most of the Quality Score problems I’ve encountered have to do with bad ad-group construction. If you thought you had enough Ad-Groups because the account is manageable, re-evaluate your criteria. You want lots of Ad-Groups that are all small and highly internally themed. In each Ad-Group there is a limited set of Ads, and so the keywords all have to imply the same action (or searcher-intent) – that is, the action that your ad creative tries to initiate.
If you blend keywords that imply different actions in one Ad-Group, you can’t match them to specific ads. Adwords has to assign an ad when a search query happens – and you’re by default lessening the chance that this ad will match the implied action (or searcher-intent) of your keyword. Quality score goes boooo.
And Now For Something Completely Different
Other potential relevancy relationships that I didn’t bother to draw lines for because it would have been too hard and my drawing hand got tired (and may not really exist):
CTR & Ad-GroupDon’t put aggregates past Google’s calculatron gods. One more reason to keep ‘em small and specific.
Creative & Landing PageThe keyword and search query may mitigate this relationship, but there could be a pure calculation done ahead of time looking for a relationship between the words in your ads and the words on your landing pages.
Creative & Ad-GroupWell, creatives (or Ads, sorry for using them interchangeably like a nomenclatural eejit, my bad) are part of an Ad-Group, tied to it – the relationship is mitigated by the keywords in your Ad-Group and how often or how consistently they appear in the creative.
CTR & Search QueryHow good a CTR can Google get for competing Ads on this search query? Google can calculate a CTR for your Ads and your Keywords, but it also knows what averages exist out there in the entire Adwords system for that Search Query that just came in – even within the cloud of match-types, competing landing pages, competing bids etc.
Search Query & Landing PageThink Google could parse all the landing pages in the small set of Ads that get returned for a Search Query? I do. This may not actually be happening, I’m not sure, but I don’t see any reason why it wouldn’t in the future, it’s likely just a processing power issue.
This begs for more content-rich landing pages, which I don’t always believe are good for conversions, depending on your business type, but Google is all gaga over, so whadda ya gonna do? Prepare for it – get a half decent paragraph or three of content professionally written for your landing pages. For now it doesn’t matter if there is content duplication (so long as you’re not linking to these landing pages from your site), but knowing how wonky Google is about duplicate content, I’m going to recommend you just write unique text for every page ever created by your entire organization.
K Disclaimer guys:
I’m just mapping points between elements in the Adwords accounts – I do not at all claim to know the secret sauce formula for Quality Score, but let me explain why I bothered to write all that stuff, in case people think I’m trying to define things.
I think examining the potential relationships between the different objects in and associated with your account, and trying to understand how there may or may not be elements of relevance that binds them, will inevitably lead to you improving some aspects of your account, or gaining a better global sense of what’s going on – Does Google actually look at all of these things? Geeeeze I hope not – but as an SEO, I know that simply trying to think about relevance in a logical way helps me to ensure I’m building sites and links that are going to be respected for their relationships in the future – I get the feeling the same kind of fuzzy logic is easily applicable to the Adwords system through things like Quality Score, so I just want to encourage people to try and think about the relevance of EVERY piece of their account, and one way to do that is to present a whole bunch of them, even if quite a few don’t have any impact or affect on the world as it currently is. Disclaimer over.
* (completely different disclaimer: claims made in this document have a statute of limitations extending no further than 2016, or if Google buys absolutely everything in the next two years, make that 2013)
About Naoise Osborne
Phew Naoise, where do I begin? A very comprehensive post. And I think a useful opportunity to begin/continue a debate. The next really good opportunity to be updated on this, IMHO, will be at SES San Jose in seven weeks. Nick Fox will be on a panel and his insights are always golden. You get more than just boilerplate information out of him.
Anyway, a couple of background points about how I come at this.
A. Tinfoil hat perspectives can be fun, but the problem with them is if the tinfoil is tuned wrong, you latch onto the *wrong* “hidden stuff” and go chasing down blind alleys.
B. To a certain degree, Google is transparent in some of what they share with us about QS. I’ve directly asked Google key detail questions about QS on various occasions, and unless you take the stance that those comments right out of Google’s mouth are simply made up, or are trivial, then I take certain Google statements as pretty much facts. So the question is, how much do we start inventing algorithms with all kinds of factors out of our heads, when the real algorithmic facts are somewhat known to us, both through a combination of studying key account types in the real world, and the facts of what Google directly discloses?
So while agreeing with the general sentiment about relevancy, I’d say that I’d frame it more cleanly if it were me, to focus on the real meaning behind Ad Quality – a focus on increased search user satisfaction. In this quest, all signals are not equal, and importantly, *most quality advertisers are not penalized* by minor factors (such as those to do with landing page and website quality), and in essence do not need to worry about them because that is not how the algorithm works. I guess that would be a long discussion, but for now I’ll restrict myself to a few quibbles or comments on some of your specific points. And thank you for keeping the ideas in the forefront.
0. The purposes of the fixes done following Aug. 2005 are interesting in themselves. Users dislike certain advertiser “types” like thin affiliates and arbitrageurs. Google dislikes those types for more reasons than just user dissatisfaction: those kinds of advertisers have tended to “jam up” or “gum up” the Google AdWords system. They’ve taken advantage of the (former) goodwill of the system by creating massive campaigns using DKI and a lot of generic automated techniques, and protecting their downside through “lowball bidding”. Google doesn’t want to serve such ads; in fact for the most part they do not wish to serve the needs of such advertisers. Thus the multifaceted quest towards Ad Quality was born. A secondary objective IMHO was to replace more and more editorial judgments (when is a pop-up not a pop-up) with algorithmic, black box calculations, to reduce the amount of discussion and back-and-forth on editorial policy and user experience issues.
1. Google will often note that many of the other signals they look at (other than CTR) are just proxies for CTR.
2. To focus on “ad” CTR is a factor is a bit odd to me. Ad testing has always looked at CTR, obviously. Ads always drive keyword CTR, so day to day, you improve ad CTR, you also improve keyword CTR. Why then would ad CTR be a separate factor worth discussing? People who’ve managed accounts for years understand that ad testing is a marketing exercise and must be pursued with multiple objectives in mind. No secret recipe in QS changes that at all: it just makes the underlying principles of testing all that more necessary and all the more powerful. (Self-reference: http://searchengineland.com/paid-search-ad-testing-manage-to-roi-or-ctr-20309)
3. You write: “It has been hinted in dark corners” that CTR is the “grand-daddy of the QS equation.” No hints, no dark corners required: it is. It always has been, and still is. That’s the secret. Along with bids, ad position is pretty much driven by CTR. With other things thrown in for exceptions and for predictive scores on new accounts.
4. “History matters.” Completely agree. That is a straightforward issue of statistical confidence. Old accounts are full of keywords and signals that have proven themselves, and as such they are often full of “Great” quality scores (if all goes well). New accounts must create a history.
5. “…That’s why PPC black hats know the value of erasing history.” That sounds like voodoo to me. You don’t need to bring black hat into it. If you have a bad history you either fix the account or start a new one. If consulting our Google rep and hearing a recommendation that we set the “reset button” on the whole account is “black hat”, so be it. But overall, I find that focusing on marketing works better than imagining you can outsmart the system somehow. Real companies tend to get a lot of keyword quality scores between 7 and 10, and if they do not, the situation is usually fixable through methods that may be complex but certainly not black hat.
6. Regarding point #5 – maybe I’m reading quickly but I think you forgot to mention that QS is match-type agnostic and you won’t be penalized (except in the pocketbook) for using broad match types. I tend to instinctively agree with the fear of broad matching to a point, but I don’t think advertisers should fear broad-matching across the board, any more than I think they should use it lazily.
I certainly agree with you that I “hope” Google really isn’t looking at all of this. So while your post is a lot speculative with regard to the specific formula, it reinforces a valuable point. Be aware of the user experience and the relationships at every level of the navigation process. The AdWords system, and many (not all) users, prefer “tight targeting.”
Sometimes I like to reinforce this concept by referencing Spool et al. (and Eisenberg)’s focus on Information Scent (my preferred PowerPoint slide photo is a beagle or bloodhound). It’s interesting to see how we can generally arrive at similar principles or conclusions from a variety of different vocabularies that talk about the user experience online.
Hi Andrew, thanks for stopping by. I really appreciate you taking the time to read my ramblings, and also the insightful comments.
I honestly didn’t intend to (or consider myself capable of) writing anything that could be considered comprehensive, I just like the idea of examining things from new perspectives to gain insight, or a fresh take, on what I know I should be concentrating on. That said, I’m not a professional PPC manager (I manage a couple of grand here or there for my personal sites, but not for clients), I’m an SEO, and I understand the distance between the two disciplines, and how my paranoid (and admittedly very sceptical) SEO brain often fails to frame PPC challenges in a particularly appropriate way – you called me on it with a very pleasant tone I have to say, which is also appreciated
(anybody who may be reading this but doesn’t know Andrew’s work, look it up, he’s really an expert on PPC).
—”To a certain degree, Google is transparent in some of what they share with us about QS… I take certain Google statements as pretty much facts. “—
I don’t think it’s particularly appropriate to dwell on the fact that, as an SEO, I quite flatly do not believe much that comes out of Google’s mouth – but I also know from a few run-ins with employees, the Adwords side and Search Quality side of the company are a left and right hand that barely know of the others’ existence, so I’ll only defend myself slightly and say, if someone at Google other than their spam team said, it’s likely quite true.
—”So the question is, how much do we start inventing algorithms with all kinds of factors out of our heads, when the real algorithmic facts are somewhat known to us”—
Not too much, certainly – but I do want to encourage people to use common sense about the relationships between the parts of their account, and how those things directly or indirectly will affect the performance of it – it may simply have been too much for me to try and frame it all in terms of Quality Score itself.
—”To focus on “ad” CTR is a factor is a bit odd to me. Ad testing has always looked at CTR, obviously. Ads always drive keyword CTR, so day to day, you improve ad CTR, you also improve keyword CTR. Why then would ad CTR be a separate factor worth discussing?”—
Yeah I really didn’t intend to ‘focus’ on it by placing it first, it wasn’t an ordered list, I kinda just drew a line in photoshop, but yes your description of managing ads as a standard marketing day-to-day job, and it taking care of the keywords, as you’re hinting at overall, is more on target with real world tasks.
Are you saying that ad CTR doesn’t factor into QS at all, and further to that doesn’t affect account performance in other ways? Or are you saying that because it’s only related, any effect would be indirect?
I wanted people to think about the fact that their ad, obviously via the searcher-intent of the keywords which trigger them, is going to be the thing that draws the click or not, so its CTR is pretty important (hey if I’m drawing that many lines, some are gonna be duh!’s) – not really directed at account managers with years of experience, I’ll admit.
—”You write: “It has been hinted in dark corners” that CTR is the “grand-daddy of the QS equation.” No hints, no dark corners required: it is. It always has been, and still is. That’s the secret. Along with bids, ad position is pretty much driven by CTR. With other things thrown in for exceptions and for predictive scores on new accounts.”—
When people have problems with things like QS (or even SEO, as it has the same sort of, uh-oh here’s an algorithm, what’s happening here?) they want to know how to fix it – framing things in terms of ‘knowledge they do not yet know’ will get people to pay attention to it – even if you’re just trying to make them pay some attention to the fundamentals, which is likely what they’ve been ignoring. Works great, till I talk about it in comments… bye, bye placebo effect. Seriously though, point taken. It’s not a secret people, I just want you to think of it that way, for your own good.
—”5. “…That’s why PPC black hats know the value of erasing history.” That sounds like voodoo to me. You don’t need to bring black hat into it. If you have a bad history you either fix the account or start a new one. “—
Well I don’t really equate black-hat to voodoo, I consider ‘erasing history’ to be erasing keywords, adgroups, campaigns and accounts – and to be honest this just came out because of the black hat PPC companies (and they are companies, millions of dollars a month in ad-spend) that I’ve known practise with ‘erase and retry’ perhaps more than is humanly conceivable, to a point where the rolling of it becomes the processes. They don’t do it to solidify QS up in an account of course, they do it skirt the issue of QS by force.
In the past I’ve been asked by a few people with Quality Score problems which appeared pretty inexplicable for some help, and I tell them to talk to their account rep – it happened a couple of times (granted this was about … two years ago) that when I couldn’t identify the QS problem relatively quickly (because, as you’re saying, in most cases it’s not a complex thing), they were given advice to kill things off and start fresh.
From those couple of experiences I’ve done a lot of ‘retries’ with ad-groups, restructuring keyword groupings and ad combinations, but what I seem to find is that I shouldn’t try to reshuffle or alter my current ad-groups and campaigns, I should reshuffle and restructure, and then re-launch with different names. How old (or originally erroneous) is this methodology, and should I go back to not deleting and starting fresh?
—”Regarding point #5 – maybe I’m reading quickly but I think you forgot to mention that QS is match-type agnostic and you won’t be penalized (except in the pocketbook) for using broad match types. I tend to instinctively agree with the fear of broad matching to a point, but I don’t think advertisers should fear broad-matching across the board, any more than I think they should use it lazily.”—
I was just trying to encourage people to understand that more specific matches (exact over phrase over broad) are more likely to improve CTR on that keyword, and so via this will affect QS. I’ll flat out say that my fear of broad matching is instinctual, it just doesn’t make much sense from the entire perspective of a better user experience – but when you say advertisers shouldn’t fear broad-matching ‘across the board’, you mean .. umm, ‘entirely’? As opposed to ‘across their account’? Right?
What kind of a set-up do you prefer, because I’ve heard basically opposite answers on this question from PPC managers recently:
-Every keyword in every match type
-A whole bunch of trying to determine the searcher intent of how each match-type might apply and only bidding on ones where you can address that in the ad
-Every keyword in broad every time
Strangely enough I had someone tell me the third option was not only quickest and easiest, but for a large part of the tech audience (least expected audience imho) they managed, it performed the best. Counter-intuitive to me and I’d love to hear your take on it.
Anyway Andrew, I hope you’ll forgive my meandering and not very well focused post – if you feel there is any specific wording I should change, or an emphasis or de-emphasis that should be added, to reduce the chance that people will take it in as bad information, please don’t hesitate to say so.