Domain ad networks must stop fraud to survive in the long run.
Austin, Texas based click fraud prevention company Click Forensics just picked up a $10 million venture funding round, bringing its total raised to $15 million. Click fraud is a big problem, and I like Click Forensics’ two prong approach: help advertisers block (and get refunded) for fraudulant traffic, but also help publishers improve traffic quality sent to advertisers.
The second point is very relevant to the domain community. Regardless of how direct navigation traffic actually converts compared to search, there is a negative perception of domain traffic. Perception can be more important than reality. When advertisers see referrals they don’t like, they tend to cut back on advertising.
I’ve read a number of forum posts where people complain about Google “Search Network” traffic coming from domain names. They believe it should be considered “Content Network” traffic. And while investigating Yahoo’s (YHOO) decision to nix arbitrage for parking company customers, I saw the same sentiment over and over: advertisers don’t like seeing traffic come from sources that don’t seem natural. It doesn’t matter how much it converts; if an advertiser sees a high amount of traffic coming from some-lame-domain.com, they assume it’s fraudulant.
The net-net is they stop advertising altogether. Why did Yahoo kill arbitrage (except with direct clients, it turns out)? Advertisers were nixing the entire Yahoo ad network because they didn’t like where some of their traffic was coming from. Opting out of content network traffic didn’t help because many parking services use a search feed.
It’s the responsibility of the domain name industry to police itself and provide the highest quality traffic possible. I predict there will be a day when Google (GOOG) and Yahoo have a separate opt-out box for domain traffic. If people don’t trust this traffic, they will opt-out. Game over.
From a website this morning:
“Google apparently has a new Content Network evangelist, and from what I understand, his job consists of promoting how “cleaned up” and profitable the content network now is, and to entice advertisers back to Content. And one of the hooks they’re using is that parked domains (on both networks) can now be excluded.”
@ Chris – looks like they’ve already started letting people out of domain parking. See my new post, and thanks for the tip.
Here is the problem:
The industry is completely confused between invalid clicks and click fraud. Those are two seperate and very different things. Invalid clicks are clicks generated by robots, email harvestors, and systems that are not designed to specifically game the PPC advertising space. Click fraud on the other hand, is usually done by a group of scammers behind anonymous proxies, and ads are clicked in order to make money for themselves. Hence why it mainly affects domain parking – because most domain parking companies accept third-party client registrations.
Search is a bit more controlled. The number of search partners out there is much less for Google which include mainly massive companies that display ads only on their own portals. Scammers would have little to no reason to click on those ads since they are not in on the revenue.
Here is an article I like to show many: http://www.marketingpilgrim.com/2006/12/google-click-fraud-rate-two-percent.html
Clickforensics, AuthentiClick, and most other third-party click fraud services are based on invalid clicks – which Google and Yahoo do not charge their advertisers for in the first place – but they do send the traffic ultimately to the advertiser if it is invalid. The misconception comes in when advertisers look at their traffic and notice it is not converting. Well, Google and Yahoo usually does not charge for that traffic if it is an invalid click. If it is in your website referrer log, it does not mean you were charged for it as an advertiser.
Now to make things really bad, Clickforensics and most other companies will not look at your conversions, they look at your referrer logs and create a system that marks clicks as invalid.
For instance, when I was a publisher with them, one of the things they did to determine if a click is “fraud”, is that they’d compare the ip address within a session. If a visitor’s ip address would change from page to page, they’d mark the entire range of that ip address as invalid for 30 or more days. So if a visitor would come in our site with an ip address of 209.43.124.54, and then on a second page that ip address would change to 209.43.124.80, they’d mark the entire range 0-255 as “fraud” for 30 days. The problem here is that they did not look at where the ip was based from. AOL for instance (making up nearly 5% of traffic), changes the visitor’s ip address on each page request. They basically banned the entire AOL network since the AOL ips are so limited. This is just one of many of their mistakes, and at the end, the biggest mistake is that this is not “fraud” as they think it is, it is still invalid clicks.
Fraud is extremely difficult to detect in real time. It takes days, weeks, or months to detect fraud based on patterns of clicks from a given client. Why? Because it is done by scammers that usually hire people off-shore for a very low wage to sit behind anonymous proxies and click ads like a real person. That is hard to detect without conversion scores from an advertiser since it looks and behaves like any real person would. That is why most domain parking companies have closed registrations or are accepting manually.
I think advertisers are opting out of domain parking because yes, domain parking companies have a hard time dealing with fraud, usually accounting about 10% of their network, especially the biggest parking companies that bring in such a big volume that Google would probably let a 10% fraud rate slide due to the volume of traffic that is brought in. That is one reason.
One other major reasons is because domain parking is made up for high invalid click rate, usually 80%+. Advertisers may see an extreme amount of invalid clicks in their referrer logs and automatically assume they are being charged for it and opt out of domain parking after realizing it is coming from parked sites. Content networks and search networks usually have a 10-20% invalid click rate, my assumption for that is because content networks use JavaScript and many bots do not click on links via JavaScript. Also, search networks require a submission of a keyword through a form. And finally, domain parking is made up of thousands of sites. A bot may crawl 1 site once. For search it is just once since a search portal is usually made up of one site. And for domain parking, we may have 10,000 sites. If it is a big robot, that can mean all 10,000 sites crawled in a day. And if they click an ad on each, that’s 10,000 invalid clicks.
And I agree with your statement: “It doesn’t matter how much it converts; if an advertiser sees a high amount of traffic coming from some-lame-domain.com, they assume it’s fraudulant.”. That is exactly what I meant when speaking about high invalid clicks in referrer log.
What we need is for the “anti-fraud” companies to first understand the difference between invalid clicks and click fraud, and then to give a solution for each to only those that qualify. Giving an invalid click detection system to a publisher will drop the publisher’s revenue due to non-perfection of systems like ClickForensics. Giving an invalid click detection system to an advertiser will make the advertisers opt out of the domain parking network. Remember, many advertisers don’t even track conversions, but they do run a program like Clickforensics. The invalid click system should only be on an ad network side, just like Google and Yahoo already have. And the click fraud system, if ever developed and successful, could be on publisher or ad network side.
There should be nothing on advertiser’s side except tracking conversions.
Matt