How To Defeat Click Fraud - Turn


The ability to defeat click fraud has been around since tracking and attribution have been viable. CPA - cost per action or cost per acquisition, eliminates the overhead of click fraud detection and monies paid to fraudsters. But charging for clicks is a goldmine and not one that the search engines are likely to abandon anytime soon.

Look for more companies like Turn to appear. Turn bills itself as 'the world's first automatic targeting, bidded CPA ad network'.

If demand for CPA doesn't rise, it simply means that click fraud isn't as bad as a lot of people are making it out to be. Isn't it about time to step away from clicks and pageviews?More on CPA.


It's even easier than

It's even easier than that.

There are a couple of problems with CPA advertising on third party sites. First being the gap between what the ad host sees, (such as click throughs), and what actually happens once the click reaches the target.

With some of the testing we did with this kind of thing years ago, we kept running into problems where the host claimed to have had an action but the target not being willing to disclose all info regarding that specific click. So if anything happened other than a direct link from click to sale, and the target wasn't willing to share cookie info for example, it created as much distrust and animosity as click fraud would have in the first place. Not really a big issue except that it increased cost by having to constantly get involved in disputes where neither side is willing to compromise.

The other "obvious" issue was defining action and how to pay for different distinctions. Sales were pretty easy, (placing 30 day cookies and trying to credit hosts for return visit sales were a hassle), and even things like subscriptions to newsletters, (this was before there were blogs), weren't too bad. Determining the value of forum memberships, passing birthday Ecards, (remember those?), those kind of things started getting a little more problematic.

Of course it doesn't really matter what is problematic EXCEPT that there is a cost involved in each problem and figuring out the pricing and maintaining acceptable margins became a little bit of an issue. An issue, by the way, far different than straight forward PPC and refunding back small percentages desingned more to appease the bitcher than to actually eliminate click fraud.

BUT, there were three things, (two of which I'm still involved with), that did work. The cost to implement and maintain was acceptable and fraud, (while possibly impossible to completely elminate and guarantee), was virtually non-existent.

Pay-per-day, ad delivery based on referrers instead of context and simply sending the click to a page that required some type of human response.

DG is right!It never HAD to be this way.

CPA abuses screw the publisher

That's why CPC is also popular as many publishers get hosed on CPA deals with cookie washing and so forth. I dumped my biggest and best paying CPA program over a year ago as my traffic levels to them were not only sustained, but increasing, and suddenly the steady conversion rate went to ZERO. They could give me no good answers and no checks for the sudden zero sales so I pulled the plug on them and then got begged to put them back on my site.


If I'm not converting why do you want me back?

Sorry, I don't play games like that, traffic is money.

True that click fraud is a problem as well, but if someone is going to get screwed, I'd prefer it's not me with CPA.


I've got some calls in to try to see if I can find out how Turn plans to deal with cookie deletion/washing, etc. As soon as I hear something, I'll post it here.

I had a similar

thing a while back. It didnt work because the goodwill vs being screwed under the guise of 'not a genuine sale/lead' became a joke. The amount of control the advertiser has over what constitutes a CPA and how its implemented is going to create alot of hurdles for pure CPA not to mention that it isnt suitable for all advertisers anyway. Maybe better communication channels between publishers and specific advertisers might help to validate with some sort of accelerator payment like MSN offers higher CPC % for demographics, you could stipulate a better % payment for an advertiser you like but enmasse thats not going to help much.

>It didnt work because the

>It didnt work because the goodwill vs being screwed under the guise of 'not a genuine sale/lead' became a joke. <

I had a terrible time with exactly this problem. That led to quite a bit of testing with other kinds of systems trying to find ways to overcome this dilemma. We were testing concepts more so than software and with a limited and small group anticipating many more problems than we actually encountered.

Maybe it's just the brilliance of ignorance but one answer back then was to send the click to a page provided by the target that sat on our server. Then we made the stats available to both the target and the host.

We had a little resistance from the advertisers,(targets), because, naturally, they wanted control. We overcame that objection with the one question,"do you want sales/signups/subscribers/real leads or do you want control?"

We probably did lose some potential targets, but the reduced cost with being involved in disputes and for our entire testing period we had 0, (yes I said ZERO), instances of fraud complaints from either the targets claiming click fraud or the hosts claiming to be cheated by the "not a real sale" guise.

By giving both parties access to the same stats, everybody seemed to agree that we were all at least seeing the same stuff.

It wouldn't work for every instance, but I'm confident that we are talking about enough market share to make the the third party page with open stats profitable. I couldn't prove it was perfect, but I could prove it was verifiable and scalable.

We also had very good results with an ad delivery system based on referrers and rotating alternate tracking scripts. This forced bots to have to
#1 identify a referrer source, (such as a search engine) and a target keyword or url. In other words the bot has to go to Yahoo and do a search. Select the appropriate link from the serps to hit the page that might run an ad

#2.Wait for the page to load and then deliver the ad

#3. Identify a unique script with each ad based on each unique referrer and then click it.

Those steps coupled with the already fair amount of elbow grease demanded of randomizing IP's, spoofing headers, setting up multiple fake accounts, (doesn't do any good to get the click if you don't get the cash), AND building the host site and getting it accepted makes it better. I say better because I don't know for sure if it's foolproof.

As long as there are PPC systems than can be frauded easier, using a referrer based system is kind of like keeping a big dog in the yard. If someone wanted your TV bad enough they could justify the time and expense of killing the dog, they are going to get your tv. But if all they want is "A" tv, then grabbing the guys next door who doesn't have a dog is faster and cheaper. You keep your TV.

I believe It can be done and with less hassle and money than a PPC system that can be bot bit. You just have to want to do it.

Sounds good

so long as you can trust the broker holding the stats / pages. I can only imagine the temptation of large advertiser budgets has / would cause established brokers to tweak arrangements to help them get the results they need whilst hitting the publisher. Being cynical I would expect a small network like this to ensure they are clean & this would work, its when the network gets big and loses its vision that I would get paranoid. I guess it would be about then that they would start to attract a large number of publishers too.

Hosting the pages on the broker does make alot of sense though, do you think Google Base, Google checkout has this in mind?

The biggest problem you have

The biggest problem you have with a CPA model is trust. Trust that your traffic is being counted, trust that the sales are credited, and trust that the advertiser is doing everything in their power to turn your traffic into sales. In my eyes, a CPA model turns the fraud from the publisher to the advertiser. Advertiser fraud is much more difficult to combat.

First, how can you ever be certain your traffic is being counted properly? Besides cookie scrubbing, brower hijacking, and other dirty tactics played by others, the advertiser has to properly setup their site to track everything. Not only that, but how can you ensure their honesty. It isn't difficult to plug-in something that sends 5% of sales to a different confirmation page without the tracking code.

You're also at the mercy of the advertiser for how your traffic converts. If their website goes down for the day, you are screwed. If their billing processor decides to scrub harder for a month, you're screwed. If the usability of the site and process of ordering is cumbersome, you are also hurt. These are things that you can't control, yet rely upon as a publisher. This doesn't even account for other issues that can hurt conversions such as the appearance of a phone number, browsers that don't accept cookies, and traffic leaks throughout the site.

But the biggest issue with CPA is the fact that it's near impossible to police advertisers. Not every site is built in a way that every single sale that goes through is counted toward the publishers payout. There are fraudulent transactions that need to be voided, erroneous leads that don't count, and refunds/chargebacks that would have to be processed at a later date against the publisher.

Now I'm not completely against CPA. It is a great idea and works in select circumstances. I just don't think it can ever be "the norm". Publishers ultimately have the final say in what they are willing to advertise on their site. Good ones don't care to deal with the irregularities and issues involved in the CPA model.

While click fraud is a huge problem, it is also much easier to police in my opinion than advertiser fraud in a CPA model. Advertisers need to be given more statistics and tools to know where their good traffic is coming from. Let them know what sites are converting for them, let them choose which search engines their sites appear on within the search network. Let them block IPs, countries (which Google still doesn't do correctly), and have more customization in their campaigns. The true fraudsters are those that are facilitating the click fraud by not providing the information or tools to combat it.

0 MY

The cost of entry into the CPA arena is now so low that there are 500 +/- CPA networks.

They feed on each other through layered networks. The link to adotas is funny.

>>Advertisers need to be given more statistics and tools to know where their good traffic is coming from.

What tools would do that job?

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.