Dabble DB

The Dabble Blog

I thought this was supposed to avoid click fraud? Our Google AdWords PPA experiment

Like many online businesses, we use Google's AdWords service to reach new customers. We've been pretty happy with it, but we're always trying new ways to get our traffic up and our costs down, and we were intrigued by Google's new Pay Per Action beta.

We tried it in December. Wow, what a difference. We got twice as many trial signups as we'd ever gotten from a single campaign, and at half the price per signup we'd been paying with Pay Per Click.

Only, the demographic we were reaching was a little unusual. For example:

  • About 85% of our PPA signups came from India and Pakistan, with almost all of the remainder coming from Southeast Asia. None of them were from North America or Europe. For reference, India normally provides about 10% of our traffic. That didn't change in December, it's just the trial signups that went through the roof.
  • Nearly everyone (90%+) that was signing up from PPA seemed to use a free email service - mostly Yahoo, sometimes GMail or Hotmail. A large percentage had emails of the form jon123@yahoo.com. Invariably, they would also use jon123.dabbledb.com as their subdomain on our service, rather than a company or organization name. That's highly unusual for us, since most of our customers are using Dabble DB for work.
  • As far as I know, none of these users logged in a second time, and none have become paying customers. Many of them still have time left on their free trials, but I have to confess that I'm not exactly optimistic this will change.

So, what's going on here? I'm not worried about the money we spent on this campaign, we run adwords experiments all the time and many of them fail. None of them have failed this spectacularly, however. I'm also not worried if a small percentage of our adwords clicks are fraudulent, because I believe that market forces will just drive the price of a click down to balance things out. But Google might have a problem if, as I believe is the case here, none of the clicks are legitimate. Maybe we just had bad luck, and of course the service is still in beta, but it made me think about the whole PPA model and whether it might be easier, not harder, to game than PPC. On the one hand, PPA is difficult/impossible to game with bots; these were real people sitting down and filling out forms to try out our service. On the other hand, we're paying enough per signup that if someone is getting a significant cut of that, it's economically worthwhile to use real people - and on our end, it's much more difficult to detect these people than it would be to detect bots. At least, it would be if they put just a little bit more effort in than they already are.

We're suspending our PPA use for now, obviously. We'll probably try the experiment again later this year, and see if we get different results. Has anyone else given it a try?


  1. SecretSanta says:

    Congratulations - you've stumbled on the biggest secret ... of 2001.

  2. Rachel Keslensky says:

    For what it's worth, I stick to Project Wonderful. It's actually pretty damn hard (if not impossible) to game if you know what you're doing.

  3. Auston says:

    We've tried to use it, but we do not receive a significant number of impressions...

  4. Francis Vallieres says:

    The only time we used it for a client, we had a similar problem. I really think there is a big problem with Google's PPA security.

  5. Rational Beaver says:

    If there are any legit sign-ups to be had at all, this is my bet for the most likely way to find them:

    1. Geotarget your campaign to the US only.

    2. Run Placement Performance reports to see which sites are sending you traffic/Actions. Exclude sites that seem shady/irrelevant or that you can see are sending fake sign-ups. Repeat frequently throughout the test period.

    Good luck!

  6. RogerM says:

    Very high percentage of ad sense and ad words clicks are *click arbitration* or clicks by people getting paid to click, only a small percentage is real customers clicking to buy products. It will take about 1-2 years when most advertisers will understand the click arbitration secret!

  7. Stephane Grenier says:

    That's a very interesting article. The question becomes how can Google figure out who's gaming and who's real. With PPC they have the details of the person clicking. They can find patterns. With PPA they don't. They don't have the data from the forms people filled on your site. They can't possibly know what you know...

    So what's the solution? I don't know. Maybe it's worth trying to even reduce your price further, making it uneconomical for these people trying to game the system.

    Alternatively, you could exclude certain countries from your PPA campaign. Unfortunately this will exclude some of your potential customers, but not every campaign is international anyways. If you advertise in a magazine for example, it might not be global...

  8. Jack says:

    This makes me laugh. Google delivers, ProjectWonderful is a total scam with crappy traffic being sold. Bad real clicks are wasted money. Spend a month on ProjectWonderful and you will see you are paying for search engines and other very low quality leads on sites that have no ranking what so ever. Their system was a nightmare to use. Campaigns would not refresh. The CPD model means you’ll be out bid constantly and end up paying for ads to run only at early AM hours. We saw nearly 100% bounce rates because most of the leads were search engines and crawler. The same money spent on Google got less visitors but 43% of them stayed compared to none from ProjectWonderful.

  9. Brian Adkins says:

    As someone who has developed a "pay per sale" advertising system, my personal opinion is that the only way to avoid fraud is to not pay for "actions", but to pay for legitimate sales.

    If merchants only pay for referrals after a legitimate sale is made, and after the money back guarantee period, fraud will be minimal.

The comments to this entry are closed.