Just in case the Googles of the world ain’t paying attention:
This Small Business Technology Transfer (STTR) Phase I project will provide a commercial solution to click fraud identification and prevention. The current existing solutions can not detect the so-called software click. This STTR project proposes a real time collaborative click fraud detection and prevention system to detect these software clicks. The approach draws on data mining techniques for fraud identification using detailed user activities. An accurate and efficient classification method based on association rule mining and data stream mining will be formulated to identify the click frauds.
The system will protect Pay-Per-Click advertisers from click fraud and improve their return on investment. The new data mining techniques discovered during the course of this research will be applied in multiple fields related to online business marketing, user analysis and other fraud identification processes.
Let’s look at this in perspective.
The median yearly salary + benefits for a software engineer in the SF
bay area is about $150K. So if this much money could have been spent
on a “solution” to click fraud, why wouldn’t the top tier engines and
networks have dedicated a person to solve it? Seems implausible to
me. Of course, if the NSF wants to pay me for my solution (fixed fee
payments), I’ll be happy to take the money. 🙂
I can only speak for Google, but we’re very supportive of academic research into this area and have spoken with a number of parties who have expressed interest in conducting such studies, including FairIsaac and MIT. There’s no question that a better public understanding of the science and theory around click fraud will benefit everyone. That said, one of the big challenges is collecting usable data for third-party studies, for which there aren’t easy, privacy-friendly solutions.
Well, it seems to me that there’s already a lot of literature
available on the web that discusses the whys and hows of click fraud.
It’s really not that hard for anyone with a technical background in
the Internet protocols and architecture to figure out. People in the
Internet technical community, such as Bruce Schneier and Lauren
Weinstein, have already made public comments about the problem.
On the other side of the coin, (some) advertisers say they don’t care,
or don’t have time to worry about it, or feel it’s “the cost of doing
business on the web.” As long as they’re getting good ROI (better
than what they get on print or broadcast media), there’s no problem.
Anyone who’s not getting good ROI should just reduce their bids, avoid
low-conversion networks, etc.
So what will $150K turn up that isn’t already understood?
We own several hundred domain names and earn revenue from ppc which helps the names pay for themselves at renewal time. Due to how good some of our names are and how little we earn from the PPC revenue compared to other domain portfolios that earn huge revenue, We have no doubt that PPC fraud is very widespread.
The problem is that they can not and will not stop it. If we wanted to earn big revenue from our names we could simply build 20-30 inexpensive desktops/laptops pop in one of the various auto click software into each, enter 100 different url’s to be visited (randomizing the length of stay) in each computer and change the software from one computer to another every 24 hrs. So you would have 2000-3000 names (20-30 computers x 100 different url’s) staying just long enough to earn $2-4 per name per day. 2000 names x $2 = $4,000
This pays for the computers in (2) days. If you want, just sell the computers after (30) days and keep going.
Good luck for(STTR) in your article above.
Sincerely: Dave Crutcher / Twilight Research