How bad is the click fraud problem?
There are no definitive answers. Search engine companies minimize the scope
of the problem, and companies that offer solutions
to click fraud paint a more ominous picture.
Fair Isaac , a leading provider of analytic
software used by major financial institutions, telecom firms and others,
just released preliminary results of research it’s done on click fraud
The early
results indicate fraudulent clicks can amount to 10 percent to 15 percent of an
advertiser’s billed click traffic. The term “billed” is important because
search engine companies say that any click fraud they identify is not billed
to the advertiser.
Google disputed the conclusions, however preliminary,
made by Fair Isaac. “We look at the clicks and decide if we bill the
advertiser,” Google spokesman Barry Schnitt told internetnews.com. He
said before filtering click results Google generally finds fewer than 10
percent are suspicious and doesn’t bill advertisers for them.
Schnitt
disputes Fair Isaac’s estimates, which would include those filtered out by
Google. “We see no data from advertisers anywhere close” to what Fair Isaac
is claiming, Schnitt said.
Advertisers also bring certain anomalous results to Google’s attention,
but Schnitt said that only amounts to about another .02 percent being
eliminated.
Fair Isaac said its results were preliminary and require more study.
The company also said its results came from a cross section of mid- to
large-size advertisers, but admits its sample research covered fewer than 10
firms. “You can’t claim four out five dentists prefer something if you’re
only surveying four out of five dentists,” said Schnitt.
But Fair Isaac said its research has value and it’s anxious to get more
advertisers involved in its study. Joseph Milana, chief scientist at Fair
Isaac, said by working with advertisers the company had access to data that
Google and other search engines don’t typically see.
“What we look at is what subsequent activity [after a click thru] happens
in the private area of the advertiser’s site,” Milana told
internetnews.com. While not everyone clicking through to an
advertiser’s site buys something or clicks further, there is a certain
baseline of expected activity, and Fair Isaac’s researchers looked for large
deviations from that norm.
The company launched the study in summer 2006 with support from the
Search Engine Marketing Professional Organization (SEMPO). SEMPO members and
non-member pay-per-click advertisers contributed anonymous click-stream data
to the study in exchange for analysis of their search engine advertising and
potential click fraud.
Jupiter Research analyst Sapna Satagopan said there is definitely a
concern about click fraud among search marketers. The research firm’s latest
survey indicates some 27 percent of search marketers experienced some kind
of problem related to suspected click fraud.
“But it is important to evaluate how many of these marketers are actually
measuring and filtering out what they consider to be click fraud, from what
is defined as invalid clicks or non-performing clicks,” Satagopan told
Internetnews.com. (For example, a certain percentage of ads are
clicked on by mistake). “So far we have not seen a deceleration in search
spending based on this concern,” she said.
Fair Isaac said it used the same patented, multi-entity profiling
technologies it’s used to successfully fight payment card fraud for more
than 15 years. In addition, the study used an “anomaly detection engine” to
analyze advertisers’ click-stream data, identify predictive patterns and
generate a “pathology” score indicative of click fraud.
“Marketers have embraced pay-per-click advertising, but ways to defraud
the system also have become more sophisticated,” said SEMPO Chair Gord
Hotchkiss. “Based on these early results, we believe the opportunity exists
for our members to participate in an expanded study that can help further
determine the extent of the problem and point the way to industry-wide
measures for dealing with them.”
