It’s VERY hard to substantiate these claims, but here they are again, in a survey covered by WebProNews.
From the survey release:
The experiment was conducted in conjunction with Los Angeles-based Clicks2Customers.com and focused on three pay-per-click (PPC) campaigns running during a 10-day period in 2005. Duplicate clicks were determined by comparing IP addresses, language, browser settings, referring URL, time of click, operating system, browser plug-ins and country of origin.
“Our random sample of PPC campaigns uncovered as much as 29.5 percent PPC fraud and showed that Google was able to account for and credit only a tiny portion of those fraudulent charges,” McGlaughlin said. “Whether it is click fraud or the lesser known impression fraud, these fraudulent clicks can cause a lot of damage to advertisers because it drains their budgets. Companies should be aware of how big of a problem it really is and be equipped to more aptly detect it.”
11 thoughts on “Nearly 30% Fraud?”
It’s the pink elephant that no one wants to talk about. And of course, Google’s contracts leave the advertiser and the publisher without control or rights to understand just how big of an issue it really is. More of Google’s “just trust us, we’re not evil” mantra. There is an excellent class action case here, as I’m sure an audit of their AdSense/AdWords click streams would reveal a much higher click fraud rate than what Google charges their advertisers.
The class action was sent back to State court in Arkansas by a federal judge. This impacts (weakens) plaintiff(s) case to some extent.
Some second tiers like Miva, Blowsearch and LookSmart have acted in recent months to eliminate poorly performing partners hoping to curtail worthless clicks.
Snap has a sweet new program designed to eliminate fraud.
I suspect it will never be eliminated completely. Just an added cost of business?
30% is not out of the question for any one campaign for a 10 day period.
Based on my experience at LookSmart, where I was responsible for all customer reporting, the easy fraud to detect was the kind described above, where you are looking for “duplicates.”
The harder stuff to find are the clicks for a given listing that get distributed across a variety of IPs/User Agents/Days/Keywords by “clever robots.”
Finding those requires decidely non-manual techniques (trend and concentration anomaly detection) to find.
One thing to ask about the study though, what is the expected definition of a click? Many people want it to reflect a unique user, or a session, yet still call it clicks. Perhaps some of the difference could be explained by better understanding what it was each system was intending to measure.
Why are all duplicate clicks considered fraud? I just rented an RV from a website I found through Google. The site didn’t have a memorable name, so the easiest way for me to find it each time I wanted to tweak my quote was by clicking on its Google AdWords ad. I probably did this 15 times in the last few weeks before eventually renting the RV. My guess is that lots of people behave similarly, especially when searching for high value items, and that claims of “click fraud” being as high as 30% are wildly overstated.
they are re releasing month old news to promote their site
I *wish* this were true, but it’s not. If it were true, then I would look like a marketing genius instead of just really good at my job. If I could take a 30% reduction in cost-per-sale, then I’d be selling everything for 1% of the retail price. Right now I’m selling it all at 31% roughly overall.
Even though I wish this study were correct, I don’t see how it could be.
I had asked for feedback on this one on the SEM 2.0 list. A few of the list members agreed:
The study is questionable. It counts all duplicate clicks as fraud, which is a matter of debate amongst the key players in the industry. Moreover, there is no proof the advertiser was even charged for all the clicks!
When the experimenters tried to set up some sites and accounts to create click fraud themselves, the advertisers were *not* charged, so they in fact failed in their attempt to prove how easy it is to fool the fraud-detection systems. Content-targeting click fraud has declined noticeably in recent months thanks to sustained and coherent pressure by advertisers. The first rule of complaint in this realm is: “be specific.”
Calling some anecdotal evidence a “study” or an “Experiment” is really pushing the envelope of what counts as science.
Fraud happens every day. But the patterns are not so easy to explain as “it’s 30% fraud out there.” On poorly-designed, poorly-conceived accounts, 10-20% is not out of the question. On well-optimized, smart accounts, small flareups may occur that amount to 2%-5% of a monthly spend, and because they can be relatively obvious, you stand to get most of the money back.
Repeat: fraud is happening and it needs to be monitored, but this study does not appear to make a positive contribution.
Larry and Sergei should just buy someody like clickdetective.com and offer fraud protection as standard to Adwords advertisers.
…if they are knowingly profiting from fraud…oh my…now that’s a story!
well…….this just re-inforces my view that i shall never never EVER promote our site through PPC….however cheap they make it appear !
For the key words we bid on we have seen 100% clickfraud with our Mia advertising. The Miva customer service people don’t seem to want to come to grips with some of the completely bogus content partners they are getting their clicks from. I have presented bogus web sites to them and they claim it is quality content.
Here’s some of the bogus sources of clicks which are on Miva network. Judge for yourself if you want an ad with these people:
I have run a number of ppc campaigns now on facebook, and strangely enough, just over half of the clicks never make it from the click to my site(snap shot at the moment on one: 166 clicks 106 people made it to the website). Now I wonder whats going on there!?