With digital media comes the ability to tie advertiser payments to actual user actions, such as clicks. This led to the formation of the entire pay-per click ad industry and the formation of advertising giants Google AdWords and Yahoo! Search Marketing both of whom make money by inter-mediating between advertisers and websites.
With this added granularity, however, comes the ability of unscrupulous websites to skew the numbers. To see examples of some of the sites that do this you can see our post on autoblogs. Even more nefarious than the autoblogs, however, are bots that can jack up the number of clicks on an ad very quickly. This type of fraud is called “click fraud” and it can be difficult to avoid, but it is easy to spot.
A Recent Experience
We had a recent related experience with an ad network where we immediately suspected click fraud. What caught our attention was a cost per KPI (in this case not just clicks, but video views) that was too good to be true. The cost was much lower than every other vendor in our run and had a volume inconsistent with the size of the target group.
Concurrently, our Analytics team flagged that there was an increasing gap between clicks and visitors on the campaign. When we dug into the numbers we found that this vendor had the largest discrepancy—completely out of whack with the rest of the vendors and industry standards. We don’t quote industry standards as the absolute measure of success, but they are useful in cases like this where we need just an “order of magnitude” check.
What was interesting in this case was that it was not simply click fraud as the vendor was optimizing the buy against video views and a certain amount of video watch time is needed to trigger the metric. Needless to say we cancelled the buy and we’ve black-listed the vendor.
We also had an issue with our Klick Holiday Video last year when it got pulled down just as it was “going viral” and we couldn’t figure out why (and YouTube tends to remain silent on these issues, see our post on trusting your content to 3rd parties).
With the new YouTube metrics we’re now able to see where the trouble was. It seems that a hacker was able to initiate clicks from a farm of machines in Russia and Eastern Europe to drive up views without actually watching the videos:
So, why would a hacker do something like this? There isn’t any value in it. Except there was. It turns out we were running a contest for the number of shares that users could get through our microsite. Apparently the hacker didn’t know that direct video views wouldn’t help their count on the microsite.
So, the true winner had only double-digit shares, but the hacker still got us banned.
The moral of this story is beware of fraudulent activity if you are running a contest of any sort. Even activity that won’t help the participant win.
How do we Protect Clients from Click Fraud?
We have a number of programs that ensure that our clients are not exposed to click fraud and, if it happens, it is reported transparently and immediately.
- We focus on premium vendors with highly qualified audiences (click fraud is usually associated with networks against unspecified “run of network” targets)
- We use network vendors who have built-in protection against click fraud and a policy of crediting advertisers when it happens, like Google does
- Continued vigilance on the part of the media team in reviewing performance data and actively managing campaigns so that we can raise the flag is something untoward is found