Over the Easter weekend one of our PPC managers managed to catch a little glimpse of an AdWords ‘glitch’ in the Keywords tab. As you can see from the image below, in the AdWords UI, a little code was exposed in a new untitled row on the far right.

We didn’t manage to grab the source code of the page, but we did get a copy of some of the strings that appeared on the page. From initial viewing, it only looks like a comma separated string of data.

Check them out below, and let us know what you make of it.


Is this an AdWords "glitch"?
What we do know is that we can pull out some of those attributes and look into them a little further. From any of the below, can we work towards improving our knowledge of what contributes to a keyword’s Quality Score?

The string variables

As we didn’t get a value here, we’re not too sure what this might usually contain. The keyword for the row that the data was output wasn’t a new keyword so perhaps we should have expected to see some historical data here.

This sounds almost self explanatory. Are we closely matching keywords with relevant ads in granular Ad Groups? This value looks like a boolean in the code above, but we wonder whether there’s a sliding scale happening in the background that we weren’t exposed to.

Again, a boolean value, and similar principles to the above. Are we matching closely keywords and landing pages? We all know that landing page relevancy is important. Do we think this could also be on a hidden sliding scale rather than just true/false?

In our code, we didn’t get a value for this value. Perhaps this is new? Or perhaps this hasn’t yet been decided for this keyword. Either way, we can now confirm that we’ve seen an example of this being factored into the algorithm “in the wild”. We think it’s interesting that landing page latency appears to be assessed on a keyword level.

It’s likely that the following three values are just used to build the UI. It might actually (unfortunately) be the case that the above variables are only used for this too, but it shouldn’t make this blog post any less useful for those trying to understand quality score and how it’s calculated, as you can see that the AdWords developers have labelled them slightly differently, which is mode indicative of what they are looking for. It’s a shame we didn’t get an insight into the algorithms behind these values as that would have been really valuable.

You can see where this data is factored into the UI below;

adwords keyword performance


How well are your ads paired with the keywords in your account? Read more about: Ad relevance.


This is looking at the experience Google expects users to have on your website once they land. Are they likely to bounce? Is your page structured well? Read more about: Landing page experience.


This is looking at the expected click through rate of your advert when shown to users. Read more about: Expected clickthrough rate.


And this is the output… Unfortunately there’s no clear way to translate the values above into a 0-10 metric for Quality Score, as this value is clearly just provided to the UI in this comma separated sting.


We’ve had a chat internally and would really like to open the conversation a little wider to the PPC and wider community.

  • Have we seen anything here that’s new to anyone?
  • Has this behaviour been experienced by anyone else before?
  • Does anyone have any page latency vs. Quality Score data? (Should we create some??)

All thoughts would be appreciated in the comments below!

Aaron Dicks

Performance Director

Performance Director of Impression. Search engine optimisation, paid media and web analytics consultant. Also programmer and digital all-rounder. @aarondicks

Aaron has specialist knowledge in SEO, PPC and Analytics Consultancy.

14 thoughts on “Does this AdWords ‘glitch’ leave any Quality Score clues?

  1. Matt Antonino says:

    This makes me want to go TRY to break things. 🙂 I’ll have to take a look later (almost 1am here) but really interesting find.

    1. Aaron Dicks says:

      Thanks Matt. I get the same urge! It’s a shame we didn’t get the source code… There’s always next time. Keep us posted if you manage to find anything useful!

  2. lauralhampton says:

    Really interesting, thanks for sharing Aaron.

    I’m intrigued by the use of the word ‘acceptable’ and the label ‘true’ or ‘false’. Does this suggest that your ad either is or is not relevant? You mention a potential sliding scale that we’re not exposed to but at the same time, it’s possible that Google is using a set of criteria to decide on a ‘yes’ or ‘no’ – so if you meet 7 out of the 10 criteria, let’s say, your ad is deemed ‘acceptable’. I’m interested to know from others reading this post – what do you think?

  3. Jamil Kassam says:

    True/False values make it all seem too simplistic if you ask me.

    1. Aaron Dicks says:

      Exactly Jamil. This is something we immediately picked up on. Each of the metrics clearly have their own underlying algo’s themselves, so unfortunately this isn’t the golden ticket answer, however it is good to see that some metrics, such as the landing page latency are coming through into the front end as it will push people into stepping up their hosting spec.

    2. Chris says:

      Why should it be complicated?

  4. Larry Kim says:

    wow great find. thanks for posting. We’ve done tons of research that show that stuff like that landing pae and keyword relevance is a binary, and that the QS calculation is essentially just CTR normalized by ad spot. Your little code snippet seems to confirm that is indeed the case. http://www.wordstream.com/blog/ws/2014/01/29/hacking-adwords

    1. Aaron Dicks says:

      Great, thanks for sharing Larry! Glad we could confirm that stuff. We wonder internally to what extent these metrics have their own algos hidden behind them

  5. Chris says:

    It’s quite amazing how we want things to be more completed than the obvious. Trying to look beyond what is there is refusing to see what is actually right there for all to see. I second Larry on this, always have.

    The only discovery for me is the existence of a postClickQualityScore value that seems unrelated to landing page. I’ve argued for long that bounce rate had no impact on quality score, and this reminds me of a conversation I had with Brad Geddes where he suspected that although bounce rate as we know it didn’t matter, that Bounce Back Rate (people clicking the ad and bouncing directly back to search results to click another ad) could be a factor. This could be it… or not. What it does tell me is that google really does take into account, in some form, the behaviour of the visitor when they land on the page.

    1. Aaron Dicks says:

      Thanks Chris. Absolutely agree on the post-click behaviour playing a part in Google’s algo. It makes sense for an advertiser to work on this regardless of QS, but it’s nice if a better experience on your site means you’ll end up paying slightly less for advertising it.

    2. Frederick Vallaeys says:

      Great find Aaron!

      Chris, I can tell you that bounce rate does matter to Google when it comes to QS but as always, and as you and Brad are alluding to, it’s more nuanced than a simple bounce rate metric like the one we see in Analytics for example. Google really wants users to like ads so when they see a user clicking an ad and then quickly leaving the page to do something else on the SERP, that’s an indication the user wasn’t happy with the result, and QS may decline as a result.

      The way I’ve always put it is that we should all care about bounce rate a lot because as advertisers, we paid for a click that didn’t convert. And in addition, it hurts our QS and increases our future costs of PPC ads on Google.

      1. Chris says:

        Thanks for confirming this Fred. This could be an exciting subject matter if we can find a way to test it and provide empirical evidence.

  6. PPCPROZ says: