September 1, 2006, 12:00 AM

Good Clicks vs. Bad Clicks

(Page 2 of 2)

Google and Yahoo find fault with critics who say the search engines don’t do enough to protect retailers against click fraud generated by affiliates and others. Both say they have systems in place that catch most click fraud.

Layers of protection

Yahoo has three layers of protection against click fraud, including filters that look for suspicious activity, Slade says. The filters collect data on such things as the IP address of the computer generating the click, the browser used, what time the click occurred, and how long after a search it occurred.

The second layer of defense is a team of data analysts and statisticians who monitor the click stream looking for suspicious patterns that might indicate click fraud, Slade says. Yahoo also uses the recommendations of those analysts to refine the filters to look for those new patterns, he says.

The third layer of protection Yahoo uses to prevent click fraud is to marry data it collects with information gathered at the retailer’s site. A retailer’s web log captures information, such as how long a visitor spends at a site, that can be crucial in identifying click fraud, Slade says. For example, a visitor that clicks through to a web site but leaves before a page is fully loaded could actually be a robot, he says.

What’s more, Yahoo gives away free clicks if there is even a hint of click fraud, Slade says. Yahoo has not charged retailers for billions of clicks that were fraudulent or suspected to be fraudulent, he says.

Google, too, has real-time filters that use algorithms to detect clicking activity that might be invalid or fraudulent, Ghosemajumder says. “There are hundreds of different factors that we look at to make that determination,” he says, without giving details. “Those filters actually detect and filter out the vast majority of the invalid clicks that we encounter.”

Misleading activity

Google also runs clicks through another set of filters offline, using a different set of algorithms to look for suspicious clicking patterns, Ghosemajumber says.

But before the industry can get a handle on click fraud, it has to agree on exactly what constitutes a click. “It’s very difficult, if not impossible, for any search engine to be able to make a definitive declaration as to whether a click is click fraud or not,” Slade says.

That’s because invalid clicks can be generated by humans or by automated systems. And activity which might appear to be click fraud might be legitimate. For example, a consumer clicking on an ad five times in five minutes-action that might be interpreted as click fraud-might be comparison shopping, clicking back and forth between web sites, Larkins says.

In spite of all the focus on the problem, though, most in the industry don’t expect a solution to click fraud anytime soon. The IAB’s Stuart says reaching a consensus on a standard for clicks will be complicated and could take 14 months or more. “We’re just really getting started,” he says.


Winning the struggle against affiliate-generated click fraud

If you’re a retailer using affiliate marketing, you are just as likely to fall victim to click fraud at the hands of your affiliates as to a competitor. An affiliate has a very powerful incentive-the money it earns every time someone clicks your link at its site, observers say. Retailers can pay search engines and affiliates as much as $2 per click for the most popular search terms.

“That’s a pretty nice living for so long as you get away with it,” says Benjamin Edelman, an attorney and click fraud expert.

One way retailers can minimize their exposure to click fraud is by contracting with an affiliate network. The vast majority of affiliate networks eliminated click fraud several years ago by switching from the pay-per-click model favored by the search engines to revenue sharing and pay-for-performance advertising models, says Shawn Collins, an affiliate marketing consultant.

In a pay-for-performance model, also known as cost per action, a retailer pays for a click only if a certain action takes place, for example, a sale.

That’s the model at Performics Inc., a division of DoubleClick, “They’re not running up click charges with us because we only pay affiliates when they transact for a client,” says Stuart Larkins, vice president of search marketing.

Commission Junction, a ValueClick company, also uses the pay-for-performance model, says John Ardis, ValueClick’s vice president of corporate strategy. “We’re focused on what happens after the click,” he says.

Commission Junction and Performics, like other networks, also closely monitor conversion rates. “If we see any dramatic spikes or fluctuations, we dig into it,” Larkins says.

But retailers don’t have to go to an affiliate network to reduce the chances of click fraud. Just by switching to a pay-for-performance pricing model, retailers can remove the strongest incentive for affiliate-generated click fraud, Ardis says. “That’s a no-brainer fail-safe,” he says.

comments powered by Disqus




From The IR Blog


Bob Brodie / E-Commerce

How raw data enables retailers to know every customer

Here are five technologies that can unleash the power of all the data an online ...


Scott Allan / Mobile Commerce

The missing link for turning app-only promotions into revenue

Apps engage a retailer’s most loyal shoppers, but technical hurdles keep many merchants from taking ...