Digital Conference 2010

Google's New Advertising Machine: the Bid Simulator

Search Giant's Chief Economist Explains How Its New Tool Can Analyze Auction Data to Estimate Impressions

By Published on .

Most Popular

NEW YORK ( -- Call him Google's brain. Officially, Hal Varian is the search titan's chief economist, but he is perhaps more accurately its high priest of big-think innovation. Academics, entrepreneurs, Googlers and the like have pored over Mr. Varian's numerous articles espousing the company's ambitions, from digitizing libraries to the nuances of copyright.

Hal Varian
Hal Varian
Mr. Varian, who will discuss the online future of information and data as a speaker at Ad Age's Digital Conference today, recently wrote about the value of auction pricing and audience measurement, but don't let these seemingly simple topics fool you. He sees them as an obvious outgrowth of a larger idea: computer mediation.

Ad Age: What is computer mediation?

Mr. Varian: I have written a lot about it, and for the people here, one example of that would be our AdWords tool. But this idea is not just for advertisers anymore. The simplest case would be between buyers and sellers, and there you can structure new kinds of contracts through computers. You can also have revenue sharing and point of sales -- which is all difficult to do unless you have computer mediating the exchange. Another example is between worker and worker in a corporation. We had this old model where people printed hard copies of documents and we circulated them around.

Ad Age: We still do that.

Mr. Varian: What makes more sense is to have a single document in the cloud and have people work on that instead of circulating around more paper. You need document sharing, version control. It's much more efficient, such as Google Docs.

Ad Age: So for advertisers, computer mediation would refer to auction pricing. What was the genesis of this?

Mr. Varian: At Google we ran across this problem of how we price at scale. We originally had a model where we're selling keywords by sales force. Then the number of keywords just multiplied so rapidly that we couldn't keep up in pricing. Auction was the best way to do it -- it was really a reaction to the problems of scale.

Ad Age: So is it only keyword search that fits the auction model?

Mr. Varian: Well, I guess the high-value stuff can be sold by sales force and the high-volume stuff would be done through auctions. But really I'm agnostic on that. We don't really have a firm position on that either way. It's going to evolve, and we'll work with that as it does.

Ad Age: Some folks would be concerned if it were all turning to the auction model.

Mr. Varian: [laughs] And that's why I answered the way I did. Really, we're agnostic.

Ad Age: You recently launched a new tool, Bid Simulator. How does it work?

Mr. Varian: It's a really exciting tool -- it shows you the real value between click and cost. It's extracting data from the auction but presenting it in a way that's easy to understand. We run it by what I call the "Groundhog Day" model. Remember that movie where you're living the same day over and over again? So you can simulate a bid over and over, and it shows you what kind of impressions you get at whichever bid. The estimates are based on historical data. So this way you can experiment. It's just like a spreadsheet where you try out different scenarios.

Ad Age: Despite that kind of transparency, there's growing concern that people aren't getting a firm sense of who's online. We're missing accurate demographics.

Mr. Varian: Well, we've got a new tool that came out this week, Ad Planner. You can see the interests of a visitor to any site. An example would be Of course a visitor to the site would be interested in wine, but it turns out many of them are also interested in travel. So if you're a travel advertiser, you would want to hit those people. We also have a way to do demographics -- it typically comes from the site operator who gathers that information.

Ad Age: But that assumes visitors to these sites are supplying accurate demo information.

Mr. Varian: That's the nature of the beast. But in many cases, you're more concerned about their interests than their demographics. And you want to see what works and what doesn't. You have to experiment and learn from it; it's the feedback loop, and it's religion at Google.

Ad Age: Does this apply outside of Google?

Mr. Varian: The future of advertising will employ this feedback loop: experimenting and learning from that experiment to alter your campaign. The old model was batch processing. You should be evolving the campaign every few days or weeks, not months. And because of computer mediation, you can measure audience reaction in real-time.