New Click Fraud Index Aims To Quantify Click Fraud Rate

Elinor Mills points
to a new Click Fraud Index service
that reports click fraud rate among those monitoring through its network to be
13.7 percent. The big caveat to keep in mind is that this rate only takes into
account data from those who are part of the Click Fraud Network, which is backed
by click fraud auditing company Click Forensics. The company does say this one
the Click Fraud Index page, but despite this, some might assume the figures
represent an industry-wide estimate.

The company says it has hundreds of
advertisers enrolled, but who exactly these are isn’t listed. If many
advertisers are in competitive, high risk areas, that could potentially drive
click fraud rates higher than the industry norm (whatever that is — no one
really agrees on a figure). Or it could drive the figure lower, depending on the
make-up. But without having a good breakdown of the network, it’s hard to take
this figure alone as proof of the state of the entire industry.

The rate is said to be less at top-tier search engines such as Google and
Yahoo, accounting for 12.1 percent. As you step down the tiers, the click-fraud
rate estimates increase, with 2nd tier PPC engines at 21.3 percent, and 3rd tier
PPC engines realize a 29.8 percent click fraud rate.

Postscript: Tom Cuthbert, president and CEO of Click Forensics emailed:

I wanted to take a moment and answer some of your concerns. You first
mention that the Click Fraud Index, "estimates the industry wide click-fraud
rate". We make no claim that the data reflects an industry of over 400,000
advertisers. In fact, the text below the graph on the Click Fraud Index home
page says, "Derived using average threat level across all industries and
keywords monitored by the Click Fraud Network?. Threat Level is identified as
having a high attribute rating score as measured by the Click Forensics?
rating engine using data provided by members of the Click Fraud Network?".

Our network is only a few weeks old and the numbers are growing rapidly. We
do not release information until we have reached statistical significance. For
example, we will eventually be releasing threat level data by vertical markets
and specific search provider statistics.

You are rightly concerned about who these advertisers are. You commented
that, "…but who exactly these are isn’t listed." We protect the privacy of
the members of our Network. In fact, our privacy policy states that we will
not release their names without permission. The Network members include some
prominent advertisers and agencies. I am happy to speak to you personally
about our member profile offline. Many are your readers and companies you will
recognize.

I thought it might be helpful if I give you an overview of our approach.
Click Forensics definition of click fraud is the combined scoring of three (3)
sets of attributes; technical, behavioral and Market (economic). We add to
that the collective intelligence provided by the Click Fraud Network member
data. Each click is scored by our algorithm and flagged in a high, medium or
low threat level. The total number of clicks falling into the high threat
level is the number we report on the Click Fraud Index site.

To improve the accuracy of our rating engine, Click Forensics is working
with Dr. Alexander Tuhzilin from NYU to further enhance our statistical
modeling approach. The company has a goal of producing reports with the
highest degree of accuracy in the industry. Dr. Tuhzilin said, ?Click
Forensics has good data and this is a source of their advantage over the
search engines. My role is to work with them to refine the scoring methodology
to improve accuracy. Their approach is to incorporate as much data as possible
to improve accuracy. The search providers simply don?t have enough data to
have the most accurate approach.??

One other key point so that the CFN aggregate number is not misleading. We
are counting number of high threat clicks. Bear in mind that the average cost
per click for a high threat level click is three to four times higher than low
threat level clicks. This should be obvious because of the economic factors
involved? the more expensive the click and therefore more competitive the
category? the greater likelihood for click fraud. Therefore even if only 10%
of clicks fall into the high threat category, the dollar value could be well
over 20% of the money spent.

I hope you would consider my feedback and invite your readers to take part
in the Click Fraud Network. We are providing our CF Analytics tool for free, a
point you failed to mention! This product scores up to 100,000 paid clicks per
month and provides summary and detail reports. This is a great service for the
thousands of advertisers trying to catch click fraud.

I agree with you, "…it’s hard to take this figure alone as proof of the
state of the entire industry." While that is clearly not what we are saying at
this point, it is a goal. As I mentioned, the more the Network grows the more
accurate the data will become. This is an industry problem and I invite you to
join our efforts.

Thanks, Tom! Bad wording on my part, and my apologies. I didn’t mean to say
the the Click Fraud Index was offering up an industry wide statistics. I meant
only to highlight that the data is specific to your particular network only, as
you correctly point out.

Meanwhile, another reader who specializes in click fraud analysis writes:

These figures have to be vetted – how are they calculating the percentage,
i.e. what does that represent?

For example, if 30% of the universe of all PPC clicks are fraudulent, and
the publishers "catch" and don’t bill for half of that 30%, then we would see
15% of advertiser traffic that is being billed for as fraudulent. Is that the
figure they’re talking about?

Or, is the percentage of fraudulent clicks 15% of the entire universe of
clicks? And if so, then what percentage of those clicks are being billed for,
and what aren’t?

Also, how are they defining "click fraud"

So much methodology to vet.

What do readers think? Please come over and comment in our Search Engine
Watch Forums thread,
Thoughts On
New Click Fraud Index & Network?