Pay Attention to the Man Behind the Curtain

Posted on Wednesday, February 24th, 2010 by Print This Post Print This Post

Categories - Featured, SEM

Many searches on Google and Bing turn up only a few ads, even on terms that one would think should have many. For example, when I search for "buy" on Google, I usually only see an ad for buy.com. A search for "europe" typically turns up only one or two ads. Is there really only one company in the entire world bidding on the keyword "buy" and two on the word "europe"? Or, is it more reasonable to think that the search engines are simply intentionally restricting the supply of ad space? And, if they are restricting the availability of ad space, why are they doing this? A colleague recently asked me a very reasonable question to the effect of: "If the search engines want to make the most money, why not show 10 ads on every search?" Google offers a very simple answer to this question: relevance. They claim that landing page quality and the relatedness of the ad's text and landing page to the search query determines the ad's Quality Score (QS). Ads with a low relevance simply aren't shown, to improve the experience for the users. The real reason, though, is even simpler than Google's answer: money. It's fairly simple to show that, many times, it's simply more profitable to just leave a blank white stripe where ads could otherwise be shown. For example, imagine there's a keyword on which a dozen advertisers are bidding. (For simplicity, assume that all twelve of those ads have great text and great landing pages, so they all have a 10 for Quality Score. This example would work just as well if they had a Quality Score of 8 or 7 or 5 or whatever - I've set them all to 10 just to make the math easier to follow.) The search engine needs to decide how many of these to show. Let's say that for every 1000 impressions, if Google shows just one ad, then 100 people will click on that ad. If Google shows a second ad, the first ad will likely get fewer clicks (because now people will have two ads to choose from, rather than just one). To make the math easy, let's assume that showing an additional ad causes each position to get 80% of the clicks it would have gotten if one fewer ad were shown. Also, let's assume that each ad gets 80% of the clicks of the ad in the position above it. (These figures have been chosen to make the example as simple as possible and are not necessarily realistic for any or all keywords.) Based on these assumptions, the table below shows how many clicks each ad position would get if Google chose to show 1, 2, 3 or 4 ads. clicks-per-1000-impressions So, by showing two ads, the search engine will register 144 clicks. And by showing three, the number of clicks for the ads in positions 1 and 2 will drop, but the total number of clicks will go up to 156. However, for this specific example, showing four ads results in fewer total clicks (per 1000 impressions) than showing just three ads. That is, adding a fourth ad waters down the results, even though that ad is just as 'relevant' as the others. To maximize the number of clicks that occur for this word, Google's best option then is to show just three ads, giving a 15.6% clickthrough rate. (Coincidentally, a study by Majestic Research found that upwards of about 17% of all searches on Google result in clicks on paid links.) Since, for ads with the same Quality Score, CPC tends to drop with position, we can also calculate the total Cost (that is, the search engine's revenue) for these ads. If we assume that the CPC of the ad in position 1 is $1.00 and each position below has a CPC that 75% of the position above it, then we can find that showing one ad would bring Google $100 in revenue per 1000 impressions (also a reasonable figure, according to the Majestic Research study). rev-per-1000-impressions Showing two ads brings in more revenue than showing only one ad, but you can see that it also brings in more revenue than showing three or even four ads, since those additional ads bring in (relatively) small numbers of (relatively) low-cost clicks, but drain high-cost clicks from the higher-placed ads. Thus, at any given time, Google has enormous leverage over maximizing its revenue (or maximizing the sponsored click-traffic at its site) simply by controlling the supply of ad space on the market. Again, this is totally independent of the 'quality' or 'relevance' of the prospective ads since all the ads in this example are equally worthy of being displayed. Attentive readers might think I've pulled some sort of trick by keeping the CPCs the same in each case even though the number of ads shown changes. In the example where two ads are shown, the ad in position 2 pays $0.75 per click. But since there's no ad shown below it, some may ask, shouldn't the CPC for this ad be only $0.01? After all, there's no bidder in third position. But, in fact, there is a bidder in third position (and fourth position, and fifth position...). There are a dozen bidders in the auction, but only the top 2 will get their ads shown. Thus, the ad in the last position that gets shown typically pays much more than $0.01 per click. Google is very careful with the wording on their AdWords help pages about this. For example, on the page 'How much do I pay for a click?' under the section called "See an example", they describe two bidders (one with a Quality Score of '11' and one with a Quality Score of an astounding '20') where the CPC of the bidder with the lower AdRank is described simply as "min price for auction", not $0.01. Just as sellers on eBay are allowed to specify a minimum price (called a 'reserve price') below which they are unwilling to sell their items, Google (which acts as both the seller and the auctioneer in its ad auctions), specifies a minimum AdRank (that is, bid multiplied by Quality Score) below which it is unwilling to sell an additional slot of advertising space. For any given advertiser, this minimum AdRank essentially sets both a minimum bid which an advertiser must make to get shown and a minimum CPC (the 'min price for the auction') that bidder must pay per click if their ad in fact does get shown. Many advertisers think that they are competing with other bidders for ad positions. In fact, they are first competing with the reserve price that Google sets. Beat it and your ad gets shown. Bid less and it does not, regardless of how 'relevant' it is.

Tags | , , , ,

8 Responses to “Pay Attention to the Man Behind the Curtain”

  1. SX says:

    Really interesting…I can see how this theory is actually true in practice but never realised it up until now. Nice article :-)

  2. roy morejon says:

    great insight on Google competition for advertisers.

  3. Shei says:

    hi your analysis holds true under the assumption of 10% ctr. but i see your point especially on the reserve price…i had a campaign that was profitable, lasted a week, status shown, rarely shown due to low quality score.

  4. Roy de Souza says:

    This is very interesting and sounds like a good explanation of what they do.

    What would happen in an auction for display advertising if they set a high floor price?
    - Google doesn’t own the sites
    - The second position is equivalent to the second page viewed by each user on that site
    - Site owners don’t like no ad shown – but do tolerate a PSA (Free Public Service Announcement)

    Would the auction show lots of PSAs?

    The sites could show ads from another ad network if Google shows many PSAs. Big sites would do this. Small sites won’t bother

    • Bradd Libby says:

      It’s in the publisher’s best interest to not show an ad if it will steal more value in traffic from other ads than what it delivers.

      For top-of-the-page banner ads, this concept doesn’t really apply, since there’s typically only one. The best ad to show is simply the one that brings in the most money per impression, on average.

      But websites that use a stacked column of ads might find that this principle applies if CTRs or CPCs tend to decrease with the position on the page. In that case, it would be advantageous to limit the availability of ad space to whatever maximizes revenue. (This would likely tend to minimize the number of PSA-type ads shown.)

  5. Great article, Bradd.

    I have been reading a few of your posts about Google. Can an organization acquire too much intelligence? It’s hard not to be Evil when you know *exactly* what everyone does on your site and when you can follow all your visitors accross the web via Analytics.

    They also know how well your ads convert and at the same time they have the ability to artificially raise prices on said ads, all of this while still calling it an auction. The part I like most is GA Ecommerce – companies that willfully share their real-time online revenue information with the sleeping giant.

    What’s next? Could my Adwords habits have an influence on how well I rank in the organics? After all, if they know everything about my ROI, they certainly know that I will spend more on PPCs if I have to..

    How far will Google’s Optimizer go to optimize Google’s revenues?

    Just sitting and watching..

    • Bradd Libby says:

      Bartek,

      I think you’ve hit the nail on the head as to why many large advertisers don’t trust Google with their conversion data. Sure, “free” features like the Google Conversion Optimizer sound great in principle, but there’s probably a good reason it’s called the Conversion Optimizer and not the Conversion *Maximizer*.

      SEM firms play a very important role in ensuring that no single company has access to the full spectrum of data from Search volume – to which only Google has (and should have) access – to Conversion volume. Unfortunately, too many companies, in my opinion, are willing to give Google access to the henhouse without even a promise in return that Google will use that information to the advertiser’s benefit only.

      Best regards,
      Bradd

Trackbacks/Pingbacks

  1. Ad Auctions are Not Auctions - Web Design Greece – WEBTEC

Leave a Reply

Follow Us on Twitter

Authors