Internet advertising is now responsible for two percent of the current budget for all advertising. With pay per click responsible for the bulk of this money, click fraud (bogus clicks) has become a very important problem to advertisers and search engines alike. Both Google and Yahoo have made settlements to advertisers in court, and both have joined the IAB (Internet Advertisers Board) in an effort to standardize auditing and accreditation. Another reason the search engines joined the IAB is to show that they are actively policing click fraud.
Percentages don’t always tell the full story, so here are some hard numbers. It is projected by the investment bank Piper Jaffrey that online advertising will increase from $27 billion in 2006 to $61 billion by 2010, or 20 percent of total advertising revenue. If you look at the chart below, you will see that pay-per-click advertising is projected to make up an increasing proportion of online advertising.
Current and Projected Internet Advertising Spending
Google Does Some PR
Advertisers are getting antsy about click fraud policing. The loads of bad press resulting from Outsell Inc’s projection in 2006 that the incidence of click fraud could be as high as 15 percent certainly didn’t help. The industry also did not appreciate Eric Schmidt’s suggestion that advertisers should just "let it [click fraud] happen." Google has also been taken to task for its secretiveness, with advertisers wondering whether the biggest search engine currently cares about policing click fraud at all.
With Google getting bad PR from the press over click fraud, and Yahoo claiming to polices "bogus clicks" better, Google has quietly been working on policing what they call "invalid clicks." The search engine claims that its secrecy in its policing techniques is to prevent click fraudsters from reverse engineering their security protocols. Shuman Ghosemajumdar, Google’s product manager for trust and safety, explained sketchily how Google tracks invalid clicks (note the distinction Google makes, more on that later) and asked that the reporting press give a more balanced report on click fraud (instead of basing it all on Outsell Inc’s report). You’ll find this information on his blog at http://www.shumans.com/. First let’s take a look at the various forms of click fraud, and then we will see what Google has to say.
Click fraud consists of web site owners clicking on their own ads (network click fraud) so as to generate higher revenue fraudulently. It also consists of competitors clicking on each other’s ads (competitive fraud) to ensure that the competition pays higher bills for PPC advertising; this is done in an effort to reduce competing bidders. The current pay per click model, based on Overture’s PPC system, includes several layers to protect advertisers from fraudulent clicks. One filter screens out known fraudsters in real time, then statistical analysis is done by human reviewers to check for possible fraudulent clicks; finally, there is correspondence with the advertisers (reporting). Google’s method of screening is similar to this model.
Shuman posted on Google’s fight against click fraud after giving Andy Beal over at http://www.marketingpilgrim.com/ an exclusive interview during an SES conference, where a two percent click fraud percentage was seemingly inferred. The error made by Beal (www.marketingpilgrim.com/2006/12/google-click-fraud-rate-two-percent.html) has since being corrected.
Ghosemajumdar implied that Google has exact figures on click fraud percentages and on invalid clicks, but that the company would be unwilling to share the information over fears that it will reveal secrets about its competitive advantage in PPC. However, he said the figure for click fraud is not two percent, but that the figure is LESS than two percent. Actually the exact word on his blog was "negligible." When pressed by questions posted on his blog, he repeated "single digit" figures. For his articles and comments check out www.shumans.com/articles/000044.php and www.shumans.com/articles/000045.php.
Venn diagram courtesy of http://www.marketingpilgrim.com/
Shuman pooh poohed the methodology used by Outsell in coming up with its figures, claiming that the 407 companies simply gave a guesstimate of the figure. During an interview with Gord Hotchkiss, he said that "it’s like asking a random group of people what they estimate the average salary in the US to be, ….when they don’t even know how much their own salary is." Obviously Google is seeking to redress the damage done by Outsell’s report.
Ghosemajumdar made a clear distinction between invalid clicks and click fraud, stating that not all invalid clicks are fraudulent clicks. He noted also that even for fraudulent clicks, not all of them are passed to the advertiser (according to Shuman, only a "negligible single digit number" is passed on). Fraud is only perpetrated when damage is done; in other words, it is when the fraud is actually done to the advertisers that it is click fraud. I infer that what Shuman means is that maybe Outsell’s figures are inflated because they consist of discovered fraud attempts and invalid clicks, which are then added up and marked to reflect suspected fraudulent clicks that got through.
According to Shuman, Google claims that no advertiser is charged with a significant percentage of invalid clicks, and over at Webmaster World forums, marketers agree that reported cases are quickly dealt with and money paid back into accounts if a case is opened. Ghosemajumdar also insists that some (in fact most according to the Venn diagram) invalid clicks are benign clicks (not fraudulent) which occur when a user goes back and forth between an ad page and the ad landing page.
Invalid clicks make up less than ten percent of all clicks according to the Google manager; the percentage that filters through to the advertisers is less than two percent (again according to the Google manager). Gord Hotchkiss and Andy Beal agree with this figure. No other statistics were given except for the figures quoted by Shuman Ghosemajumdar.
Google has a point in claiming that Outsell’s figures are supported by vested interests who would like click fraud numbers to be high in order to promote their own businesses (ad agencies, fraud detection software and fraud detection services). Google also has a point in claiming that the press have helped in spreading Outsell’s story — then again, it talks about lots of money and big companies, so why wouldn’t the press spread it? However the same argument can be used, and to much better effect against Google and the figures Shuman Ghosemajumdar claims to be quoting. Not only are they internal figures, but they work better still as an excellent PR tool.
Google’s sixty something percent grip on pay per click revenue is worth $2.7 billion per quarter as of the third quarter of 2006. That is a very large figure to see being whittled away by bad "cyberbuzz." Google would not like to see their PPC model being replaced by a CPA model, and neither would web site owners (really, AdSense is good business). Some webmasters claim that when they checked the loopholes in AdSense, they couldn’t publish their findings due to the thought of what would happen if their data was acted on and the PPC model was changed to CPA. The losses they would experience would make life slightly less interesting.
Blowtorch (www.blowtorch.com) is one site that is aggressively marketing itself based simply on its anti-click fraud stance. It is a meta search engine that offers its advertisers a money-back guarantee on driving legitimate traffic to their sites. The senior vice president of marketing, Joe Halcomb, does go overboard with the issue of click fraud, however, claiming figures as high as 33 percent. Blowtorch is hoping advertisers will abandon other search engines and turn to it for their needs. How they will do this with increased incidences of suspected click fraud is unknown, as analysts who comment on the recent "botnet" phenomenon (bot networks which perpetrate click fraud) observe that it is almost impossible to distinguish between a human click and a "bot."
Nobody really knows how high the figures for click fraud incidents really are, but for Google to be discovering up to 9 percent shows that vigilance is essential, and that perhaps other models will be more effective for advertisers.
Bill Gross (who brought us Overture and the pay per click system) has introduced a CPA model with www.snap.com. Google is likewise exploring possibilities on www.turn.com. Losing PPC to CPA, however, will reduce earnings substantially, since actions depend a lot on the web site’s landing page and content. It is in Google’s interest to tout the current PPC model as the best way for advertisers to go.