Pay-Per-Trick: Half Of All Ad Clicks Deemed Fraud

Pay-per-click (PPC) advertising models, which compensate networks or their affiliates each time a user clicks on a link, were always considered to be something of an honor system. But new data coming to light this week reveals that PPC fraud is far more significant than many industry observers would have imagined.

According to research released by Web analytics provider Clicklab, fraudulent clicks can account for more than 50 percent of all advertising fees attributable to certain categories.

The data provides a rare public snapshot into a segment of the industry that is controlled and rarely disclosed by companies that manage their own proprietary databases. And while a big player like Google, for example, does not disclose its fraud rates, the problem is significant enough that Google underlined it in its IPO filing with the Securities and Exchanges Commission as a potential risk that investors should worry about.

PPC fraud essentially breaks down into two primary areas: Competitor fraud, in which competitors run programs that repeatedly click on competing advertisers' sponsored links in order to deplete their daily ad budget, or affiliate fraud, in which affiliates utilize similar programs that repeatedly click on a link in order to increase their compensation.

Dmitri Eroshenko, Clicklab chief executive, declined to mention specific categories, citing potential client conflicts, but noted that fraud has been "much more than 50 percent" in some instances.

Click fraud is more prevalent in the B2B space, Eroshenko said, where keyword bid prices can approach $100 per click. Higher PPC rates often mean higher click fraud rates, he said, adding that the trend will likely continue in categories that garner expensive PPC rates.

"Fraudsters," as Eroshenko calls them, create automated programs that repeatedly go to Google or Yahoo! and click on the same sponsored link. These programs will click through to the landing pages, triggering the advertiser or affiliate fee, abandon the session, and then repeat.

Competitors will also run programs that repeatedly click on competing advertisers' sponsored links in order to deplete their daily ad budget. Eroshenko said some of these programs also adversely affect ranking criteria by abandoning user sessions after clicking through to a landing page, which results in a blank impression.

Affiliates, which receive commissions for referring users to merchant pages via links, write computer programs that repeatedly click on merchants' paid listings, thereby generating incremental, but illegal referral revenues for each click.

According to Eroshenko, the best way to combat the growing problem of click fraud is by monitoring analytics programs, many of which are equipped to stamp out some of these problems.

JupiterResearch Senior Analyst Gary Stein noted that "most analytics packages have some kind of fraud detection function, using IP addresses, timestamps, and so on. But it's an arms-race," he said. "Just like viruses."

Stein opined that the search categories in question probably included the affiliate-heavy pharma and health sectors. He said the results were not all that surprising, as "certain categories are loaded with scammers."

"I think click fraud is a problem," he added, "but one that's being watched. I'd be surprised if, over the enter category, you can say that [PPC fraud penetration] is anywhere near 50 percent."

Spyware research guru Ben Edelman seemed to think the problem needed closer attention, but noted the difficulty in obtaining the necessary data to conduct click fraud research. "I think the problem of click fraud is very real -- much bigger than Google et al ever want to admit," Edelman said.

However, Web analytics providers like Clicklab are the gatekeepers to the client data that researchers need to flush out these problems.

Edelman called Clicklab's methodologies "interesting," but "inadequate." He said Clicklab "will detect some, but by no means all" attempts at click fraud. "That said," Edelman added, "it's not immediately obvious how to do a better job here. It's a truly hard problem."

Next story loading loading..