Quality score is more complex than you think it is, but probably less useful. It's an integral part of the ad auction, yet the metrics Google provides to advertisers can be misleading. Don't get suckered into making the wrong decisions.
Why Does Quality Score Even Exist?
Imagine you're Google and you want to maximize your revenues. You can break that down into two main goals: revenue per search, number of searches.
Number of searches performed is pretty outside the control of the AdWords team. Google can do lots of work to break into new markets, to take share from other search engines, and to be so helpful in so many areas that they become the first port of call for answering problems. That's still pretty limited and very difficult to make an impact.
Revenue per search can be influenced by the design of AdWords. Let's break it down further: revenue per search is equal to the ad clicks per search, multiplied by the revenue per click. Revenue per click is based on advertisers' bids. The better the ROI of AdWords, the higher bids get. Great. This is the side of things that drives Google to keep improving AdWords for the advertisers.
Ad clicks per search? Now that one is trickier. It's generally assumed that the top of the page is more likely to attract a click. Anybody who has ever run an AdWords campaign can see how their CTR varies as they show higher up the page, particularly the jump when moving from the right hand side into the banner. So those top slots are valuable. If the ads in the top slots are the most valuable to Google (because they attract more ad clicks per search) then the ads in those slots need to be ads people want to click, otherwise they're wasted.
What Does It Mean?
Google's revenue gets maximized when the ads in the most valuable slots (the top of the page) are the ones with the best product (multiple) of two things: their bid and their likelihood of getting clicked.
Introducing Quality Score: essentially a measure of your predicted likelihood of getting clicked. By using that as a multiplier that works alongside your bid, Google pushes the revenue-driving ads into the best positions.
But Quality Score Isn't the Same as Predicted Click Through Rate...
No, but it kind of is. It's pretty close. There is a reason that the biggest part of the quality score calculation is your historical clickthrough rate. There are other factors, but they're all just gravy in comparison. The meat and potatoes are your CTR.
The Other Factors:
Let's look at those a moment, because we can't simply dismiss them. If we're working on the assumption that Google is trying to predict your CTR, they need more than just your historical CTR. Things change, right?
- The presence of the search query in your ad text. This isn't a good predictor on any individual level. If you perform any search now that shows ads, the ads closest to your search aren't necessarily the best ones. But across a million advertisers? The ones that aren't related to your search are likely to be worse. So when there isn't that much data about your historical CTR, or it might be a poor predictor... Suddenly this factor can help.
- The presence of the search query on your landing page. Basically the same as above, this predictor has a thousand reasons to be wrong. But across all advertisers it can give Google a bit of a guide towards the poorer quality AdWords accounts.
- Your ad extensions. These have always indirectly affected your QS by their effect on your clickthrough rate, but now it's explicit. Google will show the extensions they predict will maximise your click through rate (and therefore Google's total ad clicks) and use your history for those extensions to help predict your click likelihood now.
But What About CTR?
I've left this till last because it's so crucial. Click through rate depends on your ad text, it's that simple. There are many other factors out of your immediate control (like brand recognition) but the quality of your ad text is the distinguishing factor in how well people respond after performing a search.
Historical CTR applies only to previous searches, where the same ad showed. Immediately, you can see how much more complicated quality score can be than a simple score per keyword. Every time the search is performed, that adds on to your historical stats. The calculation needs to be performed fresh every time. What was your historical CTR when you showed the same ad you're showing now?
What Else is Important?
Good question. We have every reason to believe that the predicted click through rate of search query A is different from that of search query B, even if they trigger the same keyword.
That's an important distinction. For quality score calculations to accurately determine your predicted click through rate, they must be calculated based on your historical click through rate for the same search query, when you showed the same ad. That's not the same as a keyword.
But I Have a Keyword-Level Quality Score in the Interface!
You do. And it's worthless.
Now I appreciate that's a bold claim, and I mostly just wanted to rile you up. I'll calm down in a minute. Bear with me here.
- Your keyword represents an indeterminate percentage of your searches. The actual searches that triggered that keyword for which the current search is relevant? Only 100 percent for exact match, and for everything else a keyword level score is not representative of what you're entering the auction with. Especially when you consider that the current ad only showed some of that time.
- Your quality score is not really a score out of ten. Google are trying to predict your click through rate. A score out of ten does that poorly. The score out of ten simply represents how well you meet the criteria: Did you typically break above a 1 percent CTR? Or 2 percent? Are search queries represented in your ad text? Those kinds of questions let you reach benchmarks that give you the metric you see. But it's not representative of what's actually going on in the auction.
- The only stats that are used for calculating that metric are those for when the search query exactly matched to the keyword. That's unlikely to be all of them. So that mark out of ten? Not representative of the auction dynamics. By failing to take into account search query and ad combinations, they become unhelpful. If they're unhelpful, don't trust them.
So What Can I Do?
First off, focus on your commercial metrics instead. If your ROI is within target, do what it takes to push more sales. You've heard all this before, I don't need to repeat it.
As far as quality score goes: assume it can always be improved. Once you do that, it ceases to matter whether at the moment you're good or bad. In either case, you can make it better.
Ads. Your ad text is your weapon. Test, test and test again. Make sure your best ads are awesome, then make them better. Steal ruthlessly from your competitors, use every extension, apply your creativity, and judge by the numbers. Just make them better.
The better your historical clickthrough rate, the better your predicted click through rate, the better your ranking for your nicely calculated, profitable bids. Awesome.
Twitter Canada MD Kirstine Stewart to Keynote Toronto
ClickZ Live Toronto (May 14-16) is a new event addressing the rapidly changing landscape that digital marketers face. The agenda focuses on customer engagement and attaining maximum ROI through online marketing efforts across paid, owned & earned media. Register now and save!