Google’s AdWords and AdSense platforms are the Web’s largest and most profitable network for keyword-based advertising, which means Google deals with a constant onslaught of hucksters trying to rope consumers into scams and questionable deals. The company has long worked to exclude ads for get-rich-quick schemes, harmful products like cigarettes and guns, and counterfeit goods like fake Gucci handbags. But it’s been pretty quiet about the technology behind those efforts—until now.
In a blog post this afternoon from the engineering director behind AdWords, Google is talking for the first time about how it screens out ads from fraudsters and counterfeiters, and about measures it’s taking to beef up those efforts. Sridhar Ramaswamy, a senior vice president of engineering at Google, says the company is “fighting a war against a huge number of bad actors,” and says his team is making improvement to both the automated and manual systems the company uses to identify suspect advertisers and advertisements.
As Ramaswamy spelled out in an interview this morning with Xconomy, the company is taking steps such as adding new levels of scrutiny for ads linked to sensitive keywords related to counterfeit goods. It’s also using a new software “risk model”—an idea borrowed from the credit card industry—to assess whether people opening AdWords accounts are shady or legitimate. And many accounts and ads flagged by Google’s models must be examined by real humans, so the company is also bringing on new staff in offices around the globe to speed up these manual reviews.
Such efforts aren’t new—they’re simply being enhanced—but Ramaswamy says the company can already tell the techniques work. The percentage of “bad ads” that made it onto Google search pages in 2011 was reduced by half compared to 2010, he says in today’s blog post (though the company isn’t sharing absolute numbers). The company disabled 130 million ads in 2011 and shut down 150,000 accounts for attempting to advertise counterfeit goods. Google is “catching the vast majority of these scam ads before they ever appear on Google or any of our partner networks,” Ramaswamy writes.
In a phone interview, I asked Ramaswamy to elaborate on the new measures Google is taking to stop the “bad actors.” Here’s an edited version of our conversation.
Xconomy: Why has Google decided to get more public about its anti-counterfeit-ads effort, and why now?
Sridhar Ramaswamy: This is an issue that we have thought about very deeply, pretty much since the beginning of AdWords. It’s for a really simple reason: we need our users to like and love the ads that they see on Google. That is what makes the company successful. So a corollary to that is that we want to make sure they don’t get scammed and don’t see crazy, misleading claims on Google ads, and also that we’re not doing anything that is illegal to do.
Part of the reason for getting this out there is that there is more news about [spamming and counterfeiting]. I think it will be a good thing for people to know that we take this issue seriously, and that we have a number of teams working on it, and that it is an ongoing effort. This is not a problem that we can just solve, it’s something that we make an ongoing effort at.
X: Virus and malware makers are always getting more sophisticated about their methods. Is that also true in the area of counterfeit advertising?
SR: Absolutely. In this realm of counterfeit and get-rich-quick schemes, there is an enormous amount of money that provides very, very powerful incentives for scammers to keep changing what they do. It’s absolutely a game of cat and mouse. So our systems and teams and enforcement policies have to evolve on a pretty much continuous basis.
X: The first specific change you’re announcing today is in the area of keyword or query watching. Can you explain a bit more about that?
SR: There are lots of ads that come into our system, and our goal is to get them out onto our platform as quickly as possible, while at the same time making sure that the ads that are not okay are caught very quickly. What we essentially observed is that there are a number of queries that are more susceptible to counterfeit advertisers than others. I’m not able to give you a specific example, but imagine high-priced luxury goods that are not all that hard to imitate. That is the perfect setup for a counterfeiting operation. We have taken a set of queries [for keywords like that] and have said that they have to be subject to
