But ClickForensics' numbers suggest the rate of click fraud across different ad industry sectors and different search providers varies widely.
Google on Thursday dropped a bombshell that calls into question years of hand wringing about the future of pay-per-click advertising: The rate of click fraud for which advertisers seek a refund represents less than 0.02% of all clicks.
Google has long maintained that the risk of click fraud has been overstated by online ad auditing companies, and even gone so far as to suggest they have a financial motivation to exaggerate click fraud.
Of course the same can be said of the entire computer security industry and that doesn't diminish the reality of the threat posed by spam and malware.
And it's also fair to say that Google has a vested interest in dismissing click fraud, given that by its own account, "every percentage point of invalid clicks we throw out represents over $100 million/year in potential revenue foregone."
So how to reconcile Google's numbers with those of, say, ClickForensics, which puts the click fraud rate at 14.2%?
First, it helps to define the statistics, because the 0.02% figure Google is talking about refers to reactively detected click fraud -- clicks an advertisers questions and Google, upon analysis, agrees should not be paid for.
The 0.02% figure differs from the number of clicks that Google's systems detect proactively that never make it to a customer's bill.
It's also differs from ClickForensics' overall average of click fraud across the search industry. In much the same way that the overall crime rate in a city may not reflect the safety of a particular neighborhood, the rate of click fraud across different ad industry sectors and different search providers varies widely. Those buying expensive keywords as triggers for their search ads, for example, are more likely to be targeted by click fraudsters than other advertisers.
To further complicate matters, trying to determine the intent of someone clicking on an ad is inherently unknowable. The question then becomes how to bill uncertainty? Should advertisers pay for clicks made by mistake, with no intent to defraud? Should they pay for clicks made by the curious or those with no plan to buy? There's no easy answer here, particularly given that search ad providers and advertisers tend to have different sets of Web metrics data.
Nor is there an easy answer to another question: How much click fraud is going undetected? The fact that it's undetected makes it hard to say.
Online ad models like cost-per-action, where advertisers pay only for specific actions, like the sale of an item, can be better evaluated in terms of return on investment but they've not taken off and aren't as useful for brand awareness campaigns. Search ad providers also aren't keen to give up profiting from the inefficiencies cost-per-click advertising -- therein lies a nice margin.
Google's version of the ClickForensics' figure of 14.2% is "less than 10%." Like objects in a curved mirror, these two figures may be closer than they appear, since the click auditing firm measures the average click fraud and includes a variety of second- and third-tier search engines, which lack Google's click fraud detection resources.
ClickForensics plans to start publishing click fraud rates for specific search providers in the second quarter of the year, which will make reconciling click fraud statistics much easier. It won't resolve the issue but it will represent a step toward increased transparency, which pretty much every online ad buyer would like to see.
5 Top Federal Initiatives For 2015As InformationWeek Government readers were busy firming up their fiscal year 2015 budgets, we asked them to rate more than 30 IT initiatives in terms of importance and current leadership focus. No surprise, among more than 30 options, security is No. 1. After that, things get less predictable.
Join us for a roundup of the top stories on InformationWeek.com for the week of December 14, 2014. Be here for the show and for the incredible Friday Afternoon Conversation that runs beside the program.