I recently had the pleasure of hearing Muthu Muthukrishnan talk about Google’s approach to designing algorithms for advertising auction design. Muthu basically described the process as being similar across all the major search engines in terms of their ad sales. In short, what the major search engines do with the right half of the page.
One major feature of generalized second price auctions (pdf) is to discover a good’s (in this case a keyword’s) true price and to minimize gaming the system. I won’t go into the mechanics of auction design, the talk does a much better job than I could. What seemed particular interesting to me is that Google wants keyword prices to be relatively stable. The introduction of the generalized second price auction is designed to dampen price fluctuations.
Google doesn’t want their advertisers to be programmers and mathematicians, they want them to be advertisers. In other words, create a market that’s fair and you enable people to devote their time and energies to do what they do best. To my ear, this sounds much like Adam Smith’s proverbial invisible hand:
By preferring the support of domestic to that of foreign industry, he intends only his own security; and by directing that industry in such a manner as its produce may be of the greatest value, he intends only his own gain, and he is in this, as in many other cases, led by an invisible hand to promote an end which was no part of his intention.
Smith wants the market not to be the object of someone’s machinations, rather it’s efficiency is effectively invisible. To this end Google wants their auction to be as true as possible in regards to price, and to do that they don’t want to tinker with it endlessly. Hence the motivation for Muthu’s talk; to invite academia to study the problem of an idealized auction. If Google was tinkering with their auction all the time, this would provide incentive for some advertisers to constantly be trying to game the system. While this would be good for a few, it would be highly annoying for many. People would lose trust, and drift away from the service.
Contrast this with the left half of the page. On this side, the search engines are constantly pushing out updates. And people are constantly trying to move up the rankings. Some, in the famed ‘black hat seo’ community, are trying to game the system; much to the annoyance of search engines and users. Oh, the crooked timber of humanity! To get to the top of the rankings is to be king of the mountain. On the right side of the page, it’s often better to be in the middle of the pack. Conventional wisdom has it that the top ad clicks are often the result of a casual seeker, while a real information hunter would be willing to read through the listings to find the most appropriate source.
While I’m sure a user’s trust and the stability in the system are important goals for search, relevance is king. If you can acheive a 1% or 2% increase in relevance, the search engines would not hesitate to try and capitalize on it. Here, the mechanisms may not be known to the public, but they are far from invisible. A small tweak to the algorithm is discussed endlessly.
I find this dynamism between the right half and the left half of the page particularly interesting. I guess when anything is seen and used by millions of people a day, interesting patterns will emerge.
One last note… someone in the Q & A period asked about personalization and auctions. Here, the jovial Muthu bristled a little. He stated that he didn’t like personalized news or other forms of personalization. And that all that we were talking about in this case was a relationship between bidders and keywords. In addition Muthu turned the discussion of personalization in the direction of recommendation engines; this is not necessarily a new development, but it ties in with the idea that recommendations and advertising are very closely related.
