Google's Secret Sauce Revealed

Google opened its kimono for Saul Hansell of the New York Times last week, and (for me, at least) the glimpse was a revelation in explaining how the company has maintained its rocketship growth trajectory for so long. 

About nine months ago, when I was first playing around with Google revenue forecasts, the most important assumptions seemed to be 1) number of searches, 2) percentage of searches in which paid links appear ("coverage"), and 3) price per click.  Back then, the growth of global searches was steady but stately, and U.S. coverage seemed to have reached the stage of diminishing returns.  The growth of international coverage was going gangbusters--this was the main driver--but price per click seemed to be flattening. These observations led me to conclude that Google would probably have 2-3 more quarters of strong growth before international coverage matured and the low-hanging paid-search fruit disappeared.

What Saul's article made clear, however, is that over the last year Google has benefited from a fourth major growth driver: improving relevancy of paid links.  The company apparently devotes massive engineering and computational efforts to analyzing which ads will be the most likely to be clicked on by any given searcher, and then displays only these.  According to Saul, the system was introduced late last year (when I was concluding that U.S. growth was about to slow), and currently analyzes the user's geographical location, the time of day, and other (unnamed) factors that might improve relevancy.

It was not news that Google determines which ads to present on factors other than the price an advertiser is willing to pay: the company has long said that an ad's popularity with searchers is a big factor.  Until now, however, I at least had not realized the extent to which this selection process is proactive rather than reactive.  I also hadn't realized that the company might present different ads for the same search based on geography, time-of-day, etc.  Online media companies have been talking about both since the dawn of Internet time, but until recently, the reality has been disappointing.

Without metrics with which to analyze Google's business (searches, coverage, clicks, price-per-click), it is impossible to figure out how much juice relevancy improvements have added to the company's growth over the last year.  Similarly, without knowing how much of the company's global network is tied into the system, as well as how much more complex and effective the selection process can get, it is hard to assess how much longer relevancy will keep turbocharging the business.  This explains why, in part, Google's results have been hard to forecast and suggests that they will continue to be going forward.

It is important to understand that this is a different "relevancy" issue than most industry participants have been obsessing about for the last five years.  Relevancy-for-organic-search, where most of the arguments and development efforts are focused, is a subjective, eye-of-the-beholder advantage that may (or may not) affect which search engine a given searcher decides to use.  Relevancy-for-paid-links, meanwhile, is a direct revenue and profit driver: the same ten ads available to be shown for the same ten searches on Google might generate significantly more revenue and profit than they would on Yahoo!--even if conducted by the same searcher. 

This competitive advantage is more profound than any organic relevancy advantage, especially in a world in which, for most web users, Google has become synonymous with search (i.e., the results could suck and they wouldn't switch).  More importantly, it may also be more defensible.   It also allows the company to fulfill its other mission, improving the user experience.  Another key question, therefore, is how much money Yahoo!, Microsoft, et al, will spend trying to match Google's system. 

And then there is the last question: why Google is blabbing about all this to the New York Times.  Presumably the answer is "to tell advertisers about it," but it's not obvious why they would want to do that, especially since some advertisers might object to having their ads rejected (and not presented) by the system. Also, any discussion like this will only increase the odds that privacy freaks will jump all over the company again. 

   

Comments

[W. Zimmerman]

Well, one obvious theory is suggested by the full-page employment ad for Google on the facing interior page (at least in my Midwest edition) -- they are looking for "scary smart" people to work in ad sales.

[W. Zimmerman]

See http://www.google.com/scarysmart/.

[Niki Scevak]

I think one highly under-rated factor here is the sophistication of search marketers themselves.

Spending soars as marketers quantify the value of keywords and bid with confidence. They then use bid management technology to expand their keyword footprint into the thousands.

The growth of Google and Yahoo has been in the volume of paid referrals not the cost per click price paid in recent times.

No matter how sexy 'Artificial Intelligence' they have, raw drivers in demand far outweigh it (remember this is still an immature industry).

As far as the cutting edge of yield management, the likes of Advertising.com and Poindexter are far ahead in that game. Google is simply copy and pasting ideas. Plus ad.com and Poindexter have access to the buy-side metrics to improve targeting algorithms even further.

[Ian Holsman]

One other thing that I don't think they mentioned in the NY times article was the differential charging model that they are using as well. (described here http://jensense.com/archives/2005/10/one_poorly_conv.html and here http://feh.holsman.net/articles/2005/10/27/google-now-knows-what-you-shop-for ).

In a nutshell, they can now charge the advertiser at 2 different levels. One if the user actually converts/buys a product, and the other for a regular clickthrough. I don't know of any other advertising network which does such a thing. This in itself will allow them to charge a higher CPM then a regular advertising network as the advertiser can tell google if it led to a sale.

[Dominic Jones]

Henry,

Thoughtful piece, but I was stunned that you would find the relevancy of google ads "new."

You know, there's the old adage (from Phil Fisher, I think it was) about the need for analysts to get out there and kick the tires, speak to competitors and employees, etc.

You might want to kick the tires on Google by getting adsense for your blog. This would give you an inside view and none of this would seem new at all.

[Henry Blodget]

Thanks. It wasn't that Google was trying to improve relevancy that was news to me--it was that they were doing it proactively based on ZIP, etc. What they had said previously was that they were dumping ads that didn't get clicked on and promoting those that did, but they've been doing that for years. At least as the NYT told it, however, they introduced the proactive technology at the end of last year, about the time that revenue growth on the Google Network reaccelerated.

As Niki points out, this may also have something to do with the improving sophistication of the marketers themselves.

Good idea re AdSense for the blog...

[Anthony]

Looks like you were a few years early on the $400 target price call -- and Google instead of Amazon -- either in my mind is absurd, although Google modestly less so. I cannot help but be skeptical given that everybody seems to be on board the Google express train at this point and $100bn+ is a big market cap for a 7-yr old business...

[Brad]

This doesn't surprise me at all. What does is that others weren't trying to do it. Its a very very logical extension. I always thought they were doing it.

Brad