Google Launches CPA Ads – Problems with the Model

Google is now experimenting with CPA (Cost Per Action) advertising on it’s AdSense network, according to David Jackson and Search Engine Journal. Very interesting move, and in my opinion and taking us closer to “The Future of Search Engines” a blog that I wrote last year (and as you can see on the right hand side, one that I link on the top of my links!), and I think has been read by quite a few people so far (or that’s what my logs tell me).

I suspect that Google may start using the data in the CPA Ads to assist in determining relevancy at some point in time, not now but at least it’s moving closer to it. They’re going to have to work out all the fraud issues first – CPA engines have been a dream of mine since 2003, and given the fact that I’ve spent so much time trying to figure out this dream, that I’ve also encountered numerous pitfalls.

For instance, whilst using CPA ads in a distribution network like Adsense, you might find that you are paying CPA to Google for transactions that are fraudulent, but you only discover this after 90 days when the user chargebacks their credit card. It would be frighteningly simple for a rogue publisher from some backwater country to do the following, and get away with it:

1. Setup a site
2. Put up the ads
3. Click the ads on their own site (no click fraud)
4. Purchase using the stolen credit card (CC fraud)
5. Receive a check from Google for purchases on a CPA basis (alot more than a few clicks, because now the clicks are low, but the conversions are high – so no click fraud)
6. Receive the goods (or not)
7. Merchant receives chargebacks and is out of pocket for both the credit card charges & the goods.

That’s one scenario, and I could go into dozens more!

This knowledge really comes from experience in the affiliate marketing world as to the fraud that occurs on the affiliate networks. The CPA method in my opinion would only work on Google’s own site and selected high volume partners, as they have no incentive to commit fraud – the moment it’s opened up to the broader public, it’s becomes problematic. This is also one of the issues that networks like Commission Junction do very well at containing as they have had to deal with issues like this for year now and are equipped to assist merchants – I’m not sure if Google is equipped to deal with Chargebacks, Returned Items, etc, so I must disagree with David Jackson’s analysis in Internet Seeking Alpha (linked above).

Don’t get me wrong, Google are a smart bunch of guys, I’m just not sure if they have enough experience in the affiliate marketing world, or the systems, checks & balances in place to deal with the contingent liabilities of running a CPA Ad Network – but I’m sure they’ll figure it out!

Comments

  1. Jamie says:

    Seems like all they would have to do is align their payments with the merchants’ return policies and their risk is minimized?

  2. It’s not quite that simple. Given that at the point in time when the auction for the position is run, the advertiser who has the highest effect earning per click, given estimated CPA, may not even pay for new customers down the line, if the reversals are too high. We’ve seen reversals of up to 32% on some merchant programs (where some months they only average 15%), this fluctuation is quite unsettling in any calculations that Google may have to run at runtime.

  3. Hi Vinny,

    Your Future of Search Engines’ post was definitively prophetic and I’m happy to have it translated a few days ago!
    Now I speak about this kind of CPA revolution here. Cheers,
    Jean-Marie

  4. Jamie says:

    Seems like all they would have to do is align their payments with the merchants’ return policies and their risk is minimized?

  5. It’s not quite that simple. Given that at the point in time when the auction for the position is run, the advertiser who has the highest effect earning per click, given estimated CPA, may not even pay for new customers down the line, if the reversals are too high. We’ve seen reversals of up to 32% on some merchant programs (where some months they only average 15%), this fluctuation is quite unsettling in any calculations that Google may have to run at runtime.

  6. Hi Vinny,

    Your Future of Search Engines’ post was definitively prophetic and I’m happy to have it translated a few days ago!
    Now I speak about this kind of CPA revolution here. Cheers,
    Jean-Marie

  7. Massimo says:

    Hi Vinny,

    I’m not sure, it still seems safer to me than the current ppc model. Only true criminals steal credit cards, whereas I suspect ppc fraud is a lot easier to do, for example – not to mention that even competing ecommerce sites sometimes try to bust other sites’ budgets on adwords and adsense…

  8. Massimo says:

    Hi Vinny,

    I’m not sure, it still seems safer to me than the current ppc model. Only true criminals steal credit cards, whereas I suspect ppc fraud is a lot easier to do, for example – not to mention that even competing ecommerce sites sometimes try to bust other sites’ budgets on adwords and adsense…

  9. Hey Massimo,

    I certainly agree with you – but the credit card fraud is just one way of creating issues. If Google mixes CPC & CPA inventory for the purposes of yield, then people will click the ads, to generate lower yields – creating a similar click fraud scenario to what we have now. I’m sure that CPA is the way to go, and I’m even more sure that the road will not be easy!

    V

  10. Hey Massimo,

    I certainly agree with you – but the credit card fraud is just one way of creating issues. If Google mixes CPC & CPA inventory for the purposes of yield, then people will click the ads, to generate lower yields – creating a similar click fraud scenario to what we have now. I’m sure that CPA is the way to go, and I’m even more sure that the road will not be easy!

    V

  11. Massimo says:

    Hi Vinny,

    I understand what you’re saying, but I don’t agree with you when you say…

    >If Google mixes CPC & CPA inventory for the purposes of yield, then people will click the ads, to generate lower yields – creating a similar click fraud scenario to what we have now.

    I think that for CPA ads the fact that people will make false clicks on them or not does not really count because in the end what Google will want to match is how many impressions they give a certain advertiser (no matter if he pays per click or per action) vs. how much money they make. So, on CPA ads you can’t cheat by just clicking, you can only make click-to-action stats look funny :-)

  12. Not sure that I agree with you either!! :-)

    If the clicks are high, and no conversions, then Google will most likely determine that the ad is not relevant – their ultimate measure of whether an ad is displayed or not.

    High CTR & low yield will lead to a scenario where Google should derank the ad, due to non-interest shown by the user, especially in a CPA scenario where the user is not completing the specified event.

    Google really doesn’t care that much about money if the ads are not relevant, but when it’s relevant – they will milk you for maximum CPC/CPA – make no mistake!

    In my mind, Google will use a mixture of CTR & Conversion rate to determine whether or not an ad is relevent, so “false” clicks should affect yield and therefore ad rotation.

  13. Massimo says:

    Hi Vinny,

    I understand what you’re saying, but I don’t agree with you when you say…

    >If Google mixes CPC & CPA inventory for the purposes of yield, then people will click the ads, to generate lower yields – creating a similar click fraud scenario to what we have now.

    I think that for CPA ads the fact that people will make false clicks on them or not does not really count because in the end what Google will want to match is how many impressions they give a certain advertiser (no matter if he pays per click or per action) vs. how much money they make. So, on CPA ads you can’t cheat by just clicking, you can only make click-to-action stats look funny :-)

  14. Not sure that I agree with you either!! :-)

    If the clicks are high, and no conversions, then Google will most likely determine that the ad is not relevant – their ultimate measure of whether an ad is displayed or not.

    High CTR & low yield will lead to a scenario where Google should derank the ad, due to non-interest shown by the user, especially in a CPA scenario where the user is not completing the specified event.

    Google really doesn’t care that much about money if the ads are not relevant, but when it’s relevant – they will milk you for maximum CPC/CPA – make no mistake!

    In my mind, Google will use a mixture of CTR & Conversion rate to determine whether or not an ad is relevent, so “false” clicks should affect yield and therefore ad rotation.

  15. [...] Google Launches CPA Ads – Problems with the Model of Cost Per Action advertising [...]

  16. [...] * Google Launches CPA Ads – Problems with the Model [...]

  17. MM says:

    You’ve still got the problem with the fact that it’s adsense that this system will be running on. People who click adsense ads are more curious than ready to buy. If we’re talking CPA, to make any real money, the low low number of people who actually click these ads would have to be buying something big. Most people who click an adsense ad to get off some worthless site don’t just have their credit card laying around ready to enter the number to buy a product they didn’t know they needed. I think Google makes more money just charging per click.

    Let the arbitragers make the money. They’re the ones buying majority of the clicks.

  18. MM says:

    You’ve still got the problem with the fact that it’s adsense that this system will be running on. People who click adsense ads are more curious than ready to buy. If we’re talking CPA, to make any real money, the low low number of people who actually click these ads would have to be buying something big. Most people who click an adsense ad to get off some worthless site don’t just have their credit card laying around ready to enter the number to buy a product they didn’t know they needed. I think Google makes more money just charging per click.

    Let the arbitragers make the money. They’re the ones buying majority of the clicks.

  19. [...] PPA system. Google is still tweaking their infrastructure as they continue to receive feedback from advertisers. What I’m trying to get at is, Pay-Per-Action is in a BETA test and should be handled as [...]

  20. You make some great sense there Vinny

  21. Your Future of Search Engines' post was definitively prophetic.
    If the clicks are high, and no conversions, then Google will most likely determine that the ad is not relevant – their ultimate measure of whether an ad is displayed or not.

  22. High CTR & low yield will lead to a scenario where Google should derank the ad, due to non-interest shown by the user, especially in a CPA scenario where the user is not completing the specified event.

  23. If Google mixes CPC & CPA inventory for the purposes of yield, then people will click the ads, to generate lower yields – creating a similar click fraud scenario to what we have now.