When clicks are clinkers

When Danielle Leitch first started speaking about click fraud at industry forums a little more than a year ago, few in her audiences knew what click fraud was. Today, the vice president of marketing and analytics at search engine marketing firm MoreVisibility.com has marketers’ attention—in part because she’s done something about fraudulent clicks: gotten refunds from search engines when she’s demonstrated to engines that without their knowledge they’ve been a venue for click fraud..

Click fraud is an attempt to fraudulently derive revenue from or drive ad spending in pay-per-click advertising. Affiliates that exist solely to collect commission on clicks without delivering any value is one type of click fraud; the practice of clicking on PPC ads repeatedly and rapidly just to drive up ad spending is another. Sometimes perpetrators of click fraud are the advertiser’s competitors, seeking to deplete the advertiser’s marketing budget.

Estimates of the incidence of click fraud vary from as high as 50% of clicks to as low as 5%, with most somewhere in the middle. But to an online marketer managing a pay-per-click search campaign to budget, any clicks it pays for that turn out to be fraudulent are too many.

Leitch, who used what she’d learned in a previous job as an affiliate manager to identify click fraud in MoreVisibility’s own paid search marketing efforts last year, says there’s no established standard of proof that click fraud has occurred and no set rules on what results in a refund versus what won’t. Her experience has been that such decisions are made on a case-by-case basis. “There’s no handbook on what the engines will accept or won’t,” she says.

Leitch says the engines have recently boosted vigilance to prevent fraudulent clicks from occurring. Even so, MoreVisibility now routinely polices for click fraud on behalf of clients, recently securing a $4,000 refund from a search engine after identifying a week-long pattern of bogus clicks that had occurred in connection with a pay-per-click campaign. It’s succeeded in obtaining refunds on fraudulent clicking from both Yahoo and Google.

Leitch has used what she says are red flags in log files and analytic data indicating the possibility of click fraud to make a case to the engines.

As a start, it’s critical for marketers to first benchmark data and trends so as to be able to spot what falls outside of established patterns. With benchmarks in place, the signals of potentially invalid traffic include outliers—for example, repetitive clicks in rapid succession from the same IP address using the same referring string. “If it’s a disgruntled former employee or a competitor, chances are they don’t have a robot or sophisticated software to hide the IP address or the referring domain,” she says.

Other red flags include identification of the referring domain being stripped out or replaced with Xes in the server log. Another indicator of potential fraud is unusually high click activity on a particular keyword, coupled with low conversions, on the same day. Leitch observes that fraud typically occurs on one keyword rather than across a whole campaign.

Leitch notes that a bad sales day in pay-per-click campaigns has many possible causes other than click fraud, and that there are legitimate reasons why clicks can escalate very quickly without producing sales. “If you’re involved in contextual targeting, for example, and there is an event such as the London bombings and you’re bidding on keywords surrounding London, your clicks could go through the roof,” she says.

For that reason, she adds, it behooves marketers seeking to identify and substantiate click fraud to dig into log files as well as look at analytic data. In some cases, she points out, fraudulent clicks don’t even hit a page long enough for the analytics code to load up and capture them. “Taking a step further back, into the server log has, in our case, given enough information to warrant credit back,” she says. m