• Ian Leong

    I’m not sure I fully agree. For example, on the PPC side, positions 1-3 will get your ad additional impressions on search partner sites like Ask and AOL and subsequently increase your traffic and revenue (although there is not necessarily any correlation with quality of incremental traffic). On the organic side, there is no question top three positions are the holy grail – at least for most SERPs the way there are currently displayed.

  • http://www.rimmkaufman.com/ George Michie

    It’s a complex topic that deserves more research and explanation than I’ve provided here. The point I didn’t hit hard enough is that your avg position doesn’t necessarily correlate with traffic volume. You can have average position 2 and only qualify for 10% of the auctions, while someone with average position 5 who qualifies for all the auctions gets far more traffic. Moreover, Google is already preferentially serving ads based on how relevant your ad is to the specific person, so pushing Google to serve the ad in higher positions to people Google thinks won’t like you or your ads for some reason (past behavior, perhaps) may be a really bad idea.

  • Martin Röttgerding

    Very cool post, George! Love the story about Dora :)

    I believe that personalized quality score (anticipated CTR or click-through probablilty) plays a big role in this and makes it impossible for us to fully assess where we really stand compared to the competition.

    Two years ago I did a not-at-all scientific experiment and searched for a the same product every day, always clicking on some advertiser on the lower right rail. After a few days I began seeing the ad climbing higher, eventually taking the top spot from Amazon. Another browser (without my personal search history) continued to show the ad on a lower position.

    Another aspect in this is that users often search for the same thing twice or more, sometimes much more often. If there’s an ad in the top spot that the user constantly ignores it makes sense for Google to assume a lower click-through probablilty for that ad in the context of the current user. For the ads at the right hand side this isn’t exactly a vote of confidence either, but while it’s clear that the user has ignored the ad on top, the ads on the right rail may have been overlooked. The assumed click-through probablilty for all ads might go down, but the top ads would take a bigger hit than the others. This could lead to another order for the ads, thus contributing to the phenomenon discussed here.

    By the way – these ever changing ad positions make it impossible to assess an ad’s CTR from the outside. Ad A may have a higher CTR than ad B, but without the context of ad position AND everything else this is meaningless. This renders the impression-based evaluation of A/B tests pointless (even though statistical significance will always be a great thing to impress people).

  • http://www.rimmkaufman.com/ George Michie

    Yes, and the personalization is nothing new. We used to have to explain to clients all the time that the reason they never see their ads show up is because they search all the time and never click on their own ads. Most folks understand that now, yet haven’t connected the dots with the implications for average position.

    Thanks as always for your thoughtful and excellent commentary!

  • Martin Röttgerding

    About that last point (pushing Google to serve the ad to people Google thinks won’t like you or your ads): It’s a problem that we can’t consider this in our bidding – except in rare cases. When you are already bidding to position 1.0 and bid simulator shows minimal gains for huge bid increases, I believe that is because the only thing the higher bid will do is keep the ad in front of people who rarely ever click on it. This eventually leads to a few additional clicks for a high price, but the question is whether these clicks are as valuable as clicks from users who didn’t have to be “forced” to click on your ad. This might throw a monkey wrench into strategies built around incremental CPC…

    However, I don’t see how we could ever validate this theory. Maybe if Googles gives as a click ID and some context some day…

  • Ronnie’s Mustache

    George, thanks for explaining this so well.

    It seems to get more confusing by the day w/ so many different result sets, screens and ad formats.

  • http://www.rimmkaufman.com/ George Michie

    Adriano, you can have Google append a valuetrack parameter to your urls that will pass the page, location (top, side, other), and position for each click through. Parsing the referrer string for this parameter then allows you to do the analysis. We have our own platform at RKG so we’ve built this in to be a relatively easy analysis.

    Good luck!

  • http://www.rimmkaufman.com/ George Michie

    Thanks Ronnie. It is remarkable the sticking power of KPIs; even when they are proved to be invalid and/or not actionable, people insist on watching them. People cling to them as though they are handrails in a turbulent industry. They aren’t.