At SES San Jose, representatives from the big four search engines talked about the persistent threat of click fraud, and how they are fighting back to protect their advertisers.
SES: Search Engines Talk Click Fraud
(Our on-scene murdok staff has passed along this latest news from SES San Jose 2007. If you can’t be there, you need to be here with murdok this week, for videos and reports.)
Paul Vallez of Ask.com and his fellow speakers focused on the discussion of how their firms look at clicks to find ones that don’t conform. For example, Vallez noted how a high density of clicks from a single traffic source to a small set of keywords could be suspicious.
Large variances between the expected CPC on Ask Sponsored Listings versus those on other Tier 1 networks, sporadic traffic spikes with click averages going up 30 percent or more, and large volumes of clicks with no traffic source raise red flags as well.
Performance-based pricing is becoming more granular, which will benefit advertisers. Vallez noted that Ask offers traffic source blocking, so advertisers can block referrers that drive down their metrics.
James Colborn of Microsoft noted how their adCenter service has a new click quality reporting capability. Advertisers can see ‘low quality’ impressions, clicks, and conversions.
If a conversion happens through a low quality click, adCenter does not charge for that click despite the beneficial conversion taking place. Microsoft deems clicks low quality if they show unclear commercial intent; exhibit patterns of unusual behavior; or originate from suspect sources like spiders or test servers.
Reggie Davis of Yahoo mentioned the several teams the company has working on click fraud issues. “We’ve moved out of the dark ages,” he said of cleaning up Yahoo’s ad network.
Quality-based pricing continues to be phased in, and Davis said they are on phase two of doing so. He reminded attendees that domain blocking will arrive soon for their advertisers.
Shuman Ghosemajumder of Google said the two main incentives for click fraud are to attack advertisers, and to inflate affiliate revenue.
“Click fraud is similar to email spam detection,” he said. “Once identified, you can send (clicks) through a filtering processing.”
That process differs since email spam is more sensitive to false positives. Google builds in a high false positive rate intentionally to help keep click fraud impact low.
Shuman also said Google routinely marks clicks worth hundreds of millions of dollars as invalid as part of their click fraud fighting efforts. He restated the less than .02 percent figure Google has mentioned in the past as the percentage of illicit clicks that Google has to refund to advertisers.
The high false positive detection rate built into their filtering contributes to the lower rate of traffic Google has to refund, he said.
