« A preview of Google's "health URL" | Main | Lost in the shitstream »

When the auctioneer bids

December 08, 2006

How many times have you searched for something on Google and found, in the shaded "Sponsored Link" box at the top of the list of results, an ad for a product offered by ... Google? It happens fairly frequently for me, and every time I see it, it strikes me as odd somehow. Can a company sponsor itself? Is it fair to let the auctioneer bid in an auction? The answers, apparently, are "yes" and "yes."

Google's AdWords auction is a pretty interesting auction, anyway. Google doesn't act simply as an honest broker, with no interest in either the seller's or the buyer's fortunes. It's the auctioneer, but it's also the seller. And sometimes, as noted above, it's a buyer, too. The way the auction works is a bit funky as well. Even if you bid the most for a keyword, you don't necessarily "win." Google adds some special secret algorithm sauce to the auction results to determine the actual order in which the ads appear. (The goal is to increase the relevance of the ads to the user. Or, to put it a different way, the goal is to maximize the likelihood that a viewer will click on an ad and hence to maximize overall ad revenues.)

So, at times, Google is the auctioneer, the seller, a bidder, and it gets to fiddle with the results without divulging to other bidders exactly how it's fiddling with them.

A couple of days ago, Isaac Garcia, the CEO of CentralDesktop, a maker of online collaboration tools for businesses and an AdWords customer, wrote an angry post about Google's advertising practices. Google recently bought JotSpot, a competitor to CentralDesktop, and that has, for good reason, made Garcia a little more sensitive than he used to be. CentralDesktop pays Google to advertise on its search pages, and now it's also competing with Google.

Garcia lists a number of common keywords related to CentralDesktop's product where he found Google holding the top ad spot on the results page. I did my own searches on the keywords, and I didn't find Google at the top on every one, but I did find Google holding the top ad spot for: intranet, spreadsheet, documents, calendar, word processor, email, video, instant messenger, blog, photo sharing, online groups, and maps. If you're in the online collaboration tools business, those are some pretty powerful keywords.

Garcia's conclusion: "Google cheats." He says, "What this tells me is if you are trying to advertise a product that is competitive to Google, then you'll never be able to receive the Top Ad Position, no matter how much money you bid and spend."

I don't see any evidence that Google "cheats," which I take to mean "consciously rigs the results for its own benefit." And the statement "you'll never be able to receive the Top Ad Position, no matter how much money you bid and spend" is demonstrably false, as Philipp Lenssen notes.

But Garcia's charge, and the subsequent discussion of it in the blogosphere, got under Google's skin, and, yesterday, the company responded (without mentioning Garcia by name). A certain "Walter H." (does Google actually try to be Kafkaesque?) from Google Marketing wrote on the AdWords blog:

Our ads are created and managed under the exact same guidelines, principles, practices and algorithms as the ads of any other advertiser. Likewise, we use the very same tools and account interface ... there are no special buttons to push or levers to pull that give our internal account managers special treatment or leverage. Quality Score is automatically evaluated in the same way for our keywords as it is for any advertiser's keywords. Likewise, the potential to show up in the top spots above the search results is the same for Google's ads as it is for any other. We rely on the AdWords system to let relevancy and usefulness to our users be the driving force behind our ad placement.

I'm sure that's all true, but it's also disingenuous. Even though Garcia overstated his case, he's right to be concerned. There's something fishy here. Whether Google bids "fairly" or "unfairly" in these auctions, the fact is that it's playing with the house's money. Whether it bids a buck or a hundred bucks on a keyword doesn't make much difference to it, because the cash is just going from its back pocket to its front pocket. And if the auctioneer can bid as much as he wants without actually incurring any cost, that has to distort the prices being paid by all the other bidders, who are paying real money. I don't see how it could be anything but inflationary - whether or not the auctioneer actually secures the top ad spot - and the party that benefits from keyword price inflation is Google. (I'm sure the effects are fairly complex and that I'm oversimplifying here. I'd like to see what an expert in the economics of auctions would say about the consequences of Google bidding for its own products in its own auctions.)

Garcia writes:

In the beginning, AdWords was hailed as the revolutionary platform that enabled small start-ups, mom and pop stores and businesses all around the world to 'compete fairly in an open market bid system.' It was written that "small businesses can now compete evenly with big business - it levels the playing field." But that’s not entirely true. Is it?

Well, no, it isn't. In the cases when the auctioneer is bidding, the auctioneer has the advantage.

Unfortunately, it's difficult to judge the real costs or other effects of that advantage because Google doesn't disclose the data needed to do the analysis. It's all locked up in the big black AdWords box. If I were competing with Google and also bidding against it for ads in its AdWords auction, I'd be a little wary of that box, too.

UPDATE: Zoli Erdos and Donna Bogatin have both raised similar questions.

UPDATE: Google gets cute, redefining ads as "tips" in order to bypass AdWords altogether.


Well put. I can offer a couple more reasons to feel uncomfortable competing with Google:

* If your startup relies on Google ads, then Google probably knows more about your site traffic than you do.

* If Google is your research tool, then you're feeding your brains to their (internal) market intelligence engine.

* They will use their market intelligence engine to beat you.

Okay, maybe I read too much Kafka. ;-) And I don't know if they really have a market intelligence engine -- but they should. Imagine how your search history helps Google build synapses in their brain. All day, all these searchers drawing connections between ideas. Who is searching for what and when? What else did they click on? Very interesting stuff.

And consider how their servers could track your progress through the web using their ads. Even if you don't use Google, or you leave Google's search page and go on a random walk, they can follow you pretty well. More food for their brain.

Kafka gives me warm fuzzies in comparison.

Posted by: Sid Steward [TypeKey Profile Page] at December 8, 2006 12:25 PM

Being a PhD student in economics, with some background on auction theory, I can tell you that, to me, Google has all the incentives to be fair and honest --- and not playing with the house's money.

All the ads for CentralDesktop (or any other company competing against Google) that people don't click on because they are not displayed first, all these mean less money for Google. That is called opportunity cost, and is essential in understanding any auction, or business decision.

I'm afraid however that Google is bound to have the top spot for several reasons:

- Google is a stronger brand than anything: people trust it for providing relevant, quick answers --- hence more click because of the face-value;

- Google products tend to be better designed (well, I think so -- Blogger being an exception), and retain more users (I don't know any data exactly on that, but have seen converging elements in that direction);

- Internet ads are a new product, and most people tend to distrust them --- or rather to approach them by the bottom: play small first, then increase the rates because you realize they make sense; Google knows exactly what they mean, and takes an immediate rational action, often bolder; the way the ad algorithm works favors such affirmative actions.

The AdWords program doesn't rely on mysterious principles, there are two elements:

- they multiply your bid by the click ratio, and rate the ads in that order, in order to favor the most relevant/profitable ad;

- they don't just consider the ratio, but use finer ("secret sauce") statistics to anticipate the click-conversion.

Regarding playing with the house's money, in spite of being "not a conventional company", Google still use internal accounting: any bid from the marketing of new product that is transferred to the AdSense program is considered a cost for launch-teams, and an improvement for the Ad-teams.

Doing this wrongly means loosing money on one or either of both side:

- Launches bid too much;

- Ads earn less than by granting the best seat to better products...

This would send a clear signal to the management.

However, if you introduce the idea that they are purposely over-bidding now to kill the competing firms, and be a monopoly later, than you can look at the essential facility doctrine, namely:
Search engines are key sources of customers, one is not allowed to control one and make it too expensive to access for the others. Then having two clear-cut system, one that bids, the other that auctions, is crucial.

Any investment in launches has to be justified by the expected return rather than by dumping. This demands a very fine inside-control, and most issues here are about what actually makes a consistent strategy.

Posted by: Bertil [TypeKey Profile Page] at December 8, 2006 01:07 PM

More fuel for the auction fire.



Paid search advertisers who bid what they're actually willing to pay end up paying more than they need to, according to research by economics scholars at Stanford Business School.

The research paper by Michael Ostrovsky, assistant professor of economics at Stanford; Benjamin Edelman, doctoral candidate in economics at Harvard; and Michael Schwarz, RWFJ Scholar at UC Berkeley, blames both the naivete of bidders and the “generalized second price” (GSP) auction mechanisms used by Google and Yahoo

"We want to educate advertisers about the fact that in some sense they are being taken advantage of," Ostrovsky said in a statement. "Under the current mechanism, if they don't think carefully about their bidding strategies, they can end up paying a lot more to the search engines than they need to."

Here's a link to the cited research paper:


We investigate the "generalized second price" auction (GSP), a new mechanism which is used by search engines to sell online advertising that most Internet users encounter daily.

Posted by: Sid Steward [TypeKey Profile Page] at December 8, 2006 01:28 PM

Thanks very much, Bertil. Very helpful. A few comments/questions:

Regarding the opportunity cost of forgoing revenues from other advertisers when someone clicks on a Google ad: It's not just a cost. There's obviously value to Google in having someone click on its ad, particularly when that ad is vying for attention against competitors' ads. Also, I'm not sure what effect Google's opportunity cost has on costs/benefits for other bidders. It still seems to me that Google's participation would increase costs to other bidders, particularly since, as you point out, Google has natural advantages in securing high-ranking slots (hence requiring other bidders to bid relatively more to compete).

Regarding AdWords rankings not relying on "mysterious principles." I agree that the general principles are fairly obvious, but, as you point out, the actual calculations are hidden from bidders. So while the principles aren't mysterious, the actual process is.

Regarding the fact that Google transfers costs/revenues for house ads between its units, that's of course true and would dampen any incentive to overbid. But, still, transfer costs are accounting fictions, not real costs. It's still playing with the house's money, unlike other bidders.

Posted by: Nick Carr [TypeKey Profile Page] at December 8, 2006 01:42 PM

I have another question for any accountant who may be reading. Given that Google Marketing is an active buyer of Google Advertising, how would the transfer costs be accounted for in Google's financial statements? Would some of Google's reported ad revenues consist of transfers from Google Marketing (and hence also show up as marketing costs)? It's probably trivial, but I'm just curious how that works.

Posted by: Nick Carr [TypeKey Profile Page] at December 8, 2006 01:51 PM

This is why there are strict accounting rules when companies buy products and services from themselves. The "transfer pricing" Wikipedia article on is actually pretty good on the subject.

So...Google biz units pay real money to advertise on Google. Thus, there's no opportunity cost.

The only point to make here is the *perception* of a conflict of interest.

It's a big yawner if you ask me.

Posted by: pwb [TypeKey Profile Page] at December 8, 2006 03:40 PM

Remember when Microsoft claimed there was a "Chinese wall" between its OS division and its application software divisions and that the app divisions got the updates to the OS at the same time as 3rd party developers...nobody believed it. It could be the same case here @google.

Bertil, one Q: If Google spent a lot of money acquiring a software company they think is strategic (?) to its future and this company/division competed with one of its advertisers why is it not outside the realm of possibility that they would give up some ad money to promote this division and at the same time starve a competitor?

Posted by: GaryValan [TypeKey Profile Page] at December 8, 2006 07:47 PM

A couple of comments: as to the "transfer accounting", Google does not report results granularly enough for this to be visible, therefore it just doesn't matter. Nick is right when he says the brands are playing with the house's money. A $100 ad "cost" for the spreadsheet division is also $100 revenues for the ad division... but since these are not broken out, who cares? (Publicly, at least... there may be internal P&L's)

But the bigger issue is the point of the post: This is inflationary. Here's how the ads work. Let's say there are 2 bidders, and one bids a max of $1 for the word "foo". The other bids $2 for "foo". Google would net external monies of $1.01, basically, because it takes the top money from the previous vendor, and adds a bit for the next.

Enter a third bidder, say Google. And Google bids $3 for "foo". They win, "paying $2.01", bidder #2 pays $2, and bidder # 1, ranked third, pays $1. Total net external monies to Google? $3.

The reason auction theory isn't deteministic in this case is because this isn't a zero-sum game... all the top bidders win some real estate, as opposed to bidding on a scarce resource (like ad time on television)...

What to do? It's certainly OK for Google to use its own real estate to advertise its own stuff. But advertisers competing with Google (and who therefore will be subject to this phenomenon) need to be aware that they are betting in a game where the odds aren't as good as they might think.

Posted by: Phil Gilbert [TypeKey Profile Page] at December 9, 2006 12:21 PM

On the transfer pricing thing - it is not uncommon and GOOG is one such that does this - to establish offshore entities that have lower tax regimes and shelter profit accordingly. It is impossible to say whether AdWords are considered candidates for offshoring. In any event, GOOG will tend to lump any 'required' distinction in GS&A.

GOOGs accounts are unusual. I assume they meet SEC requirements but they don't offer the same level of disclosure as most others.

On the specific issue of AdWords: is it any concidence that the words you've hightlighted are those most closely associated with M$? If this is part of a dual strategy then it would even be possible for them to cook up a way of writing this off as R&D. That of course is pure speculation and in any event would only apply to internal accounts. The ones they publish would have to 'net out' the amounts they bid at the internal market.

I may explore this on my blog.

Posted by: Dennis Howlett [TypeKey Profile Page] at December 9, 2006 10:35 PM

Hopefully we will not regret that we will have not considered Google as critical as others (like Microsoft).
See my small cartoon.


Posted by: Oliver Widder [TypeKey Profile Page] at December 10, 2006 02:59 PM

I am making the (strong) assumption that companies bid as much as what ads gives them: tracking allows this quite finely --- and its the key new aspect of Internet ads. Therefore, the actual process of the auction is not as relevant a priori for the long term as how much each company can expect from having some ads visible.

The detail of how the process deals with multi-criteria would be if (what I believe) people search terms that are not perfectly related to what they look for, and Google as a wider range of services.

On this matter, I concur the GSP auction problem, and I have seen a Google Tech-Talk where one of the authors mentioned it quite bitterly to a numerous assembly of mostly AdWords techies, that appeared quite aware of the problem; tough no one said anything publicly, if I remember properly, the most likely assumption was that the complexity of the optimal process prevents Google to offer it to the common buyer.

Unless anyone makes the perfect explanation on "Multi-dimentional bidding for dummies & average eBay users" ("Dummies" here including anyone but the ten or so people on Earth able to understand Ostrovsky's paper) this argument would be coherent with Google's "simplicity for everyone".


1. All what makes a click valuable for Google should be included in their bid for an ad; and all what makes a click valuable to their competitor in their opportunity cost. I am assuming that other bidders are rapidly converging to efficient competing --- but if Google's entry increases their bid, this cannot mean that they value users more than Google: whose product is better, then?

2. When mentioning "finer ("secret sauce") statistics", I mean that they use wisely information that is already available in the AdWords interface. You can't blame them for trying to increase everyone's profit, do you? E. g., they insist on people trying several texts, time of the day and words to find out the right sweet spot. Fact is, the house is playing with far more information (presumably all AdWords) than each player (only theirs) on this one --- and having a Information "Wall of China" (and detailed public information on what makes a good ad) should be a good thing, indeed.


Choking a competitor (whether this is prior to acquisition or as a monopoly strategy) is economically sound in most cases, however:

- any proof of that would cost Google a tremendous amount of credibility --- remember this is a company based on trust by many aspects; doing so seems by far more costly than profitable: who'd play with needles when sitting on nothing but billions worth of (according to Nick) over-inflated stock?

- in today's IT, most services have a low barrier to entry: choking one is maybe just a way to encourage more entrepreneurs to come; this was obviously not the case in the OS market for a long time, but I don't think Google is competing against a limited number of services rather than against ignorance, time or it-self.

I have never heard anyone from Google saying anything about competing against a given company. I guess no one here but me is buying this "Love -is-all" rhetoric -- but what might be a insight of how pervasive is co-opetition (Nick, you could attack this new buzzword) is the most typed (non-porn or starlet-related) search word typed in Google: Yahoo. And in Yahoo? Google.


I have no doubt there are opportunity cost for Google to take the sweet spot from a competitor: not from AdWorks point of view, but over-all Google. I've tried to read the Wikipedia article on Transfert pricing, but I really could not make sense out of it: graphs are clumsy, seem inconsistent with the text, and I can't figure out all the implicit and very strong assumptions made (linear, symmetric costs). Sorry about that.


I agree with you, except on the market considered (web-apps for office-like services), their are more than five players, hence the ordinated slots one the first page are in limited supply.

Sorry to have been so long --- but if any of you are here in Paris for the LeWeb3 Conference, I'd happily meet you here to talk about that in details.

Posted by: Bertil [TypeKey Profile Page] at December 11, 2006 10:36 AM

I wanted to raise some awareness with a problem using Goggle Ad (Placement Targeted) ads. I am a new business owner and made a very big mistake by changing my marketing strategy from CPC (Cost per Click) bidding to ad placement yesterday.

I set up my account on a 2/day budget, I was sure not to get many leads, however since this was a new marketing strategy I wanted to see how it went for a few days.

In just 8 hours of my campaign being active, I read my account summary and to my surprise my bill went from owing 80.00 to over 400.00 in just 8 hours!

I have yet to tell this to my partner, in fear that he will think I don't know what I am doing. However, I read the instructions and upon further research I have discovered that the ad I wrote only appeared on 2 websites, neither of the sites is the ones listed in my account summary as to having all of the clicks.

Further note, this campaign was not structured on a click basis; it was for impressions, this way a business owner could place their business logo out there to get known, verses worrying about spending in excess when their competitions clicked on their ads.

I tried to get in touch with Goggle several times today and have received no feedback from them. If anyone who reads this knows of a telephone number that I can call about this issue with Goggle, please send it my way.

I would suggest all ---ALL business owners finding a new way to market. This is not the way I chose to conduct my business, and I don't expect a company with such a "reputable" name to conduct business this way either.

Thank you for allowing the comments.

Owner of:
One Stop Write Shop

Posted by: Litwriter21 [TypeKey Profile Page] at November 16, 2007 07:20 PM

Post a comment

Thanks for signing in, . Now you can comment. (sign out)

(If you haven't left a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Thanks for waiting.)

Remember me?

carrshot5.jpg Subscribe to Rough Type

Now in paperback:
shallowspbk2.jpg Pulitzer Prize Finalist

"Riveting" -San Francisco Chronicle

"Rewarding" -Financial Times

"Revelatory" -Booklist

Order from Amazon

Visit The Shallows site

The Cloud, demystified: bigswitchcover2thumb.jpg "Future Shock for the web-apps era" -Fast Company

"Ominously prescient" -Kirkus Reviews

"Riveting stuff" -New York Post

Order from Amazon

Visit Big Switch site

Greatest hits

The amorality of Web 2.0

Twitter dot dash

The engine of serendipity

The editor and the crowd

Avatars consume as much electricity as Brazilians

The great unread

The love song of J. Alfred Prufrock's avatar

Flight of the wingless coffin fly

Sharecropping the long tail

The social graft

Steve's devices

MySpace's vacancy

The dingo stole my avatar

Excuse me while I blog

Other writing

Is Google Making Us Stupid?

The ignorance of crowds

The recorded life

The end of corporate computing

IT doesn't matter

The parasitic blogger

The sixth force



The limits of computers: Order from Amazon

Visit book site

Rough Type is:

Written and published by
Nicholas Carr

Designed by

JavaScript must be enabled to display this email address.