I was just reading Barry Schwartz’s report that Google is opting-in some AdSense publishers into Pay Per Action (CPA)  ads. He poses the question of why would Google push these ads on the publishers who haven’t asked for it? The immediate answer I come up with is that this could actually be a test to try to detect fraud, since CPA is thought to be less prone to exploit. After all, the publisher would only get paid for these ads if someone buys – not just clicks on the ads on their sites. Perhaps the publishers that are getting opted-in are ones for which Google has had some question about the quality of click-through in their regular PPC ads.
I’ve been thinking that an unpublished problem with Google’s pay-per-action product is that Google itself is likely to become more a victim of fraud with these types of ads. Read on and I’ll describe…
I haven’t signed up for Google’s pay-per-action ads, so I’m basing this on what I know about how CPA ads would be integrated with an advertiser in general. As I understand it, Google needs to be able to detect when a particular action such as a sale in an online catalog has been completed. In order to do this, Google likely has the advertiser install a little bug onto an exit page in their shopping cart. So, when a person who has clicked through an ad completes an order, this bug would detect when the customer passes through the “Thank you for your Purchase” page or something similar at the end of the advertiser’s shopping cart process.
So, Google and the publishers of Google’s ads, depend on that reporting bug in the shopping cart exit page to be present in order for them to account for each billable “action” to be completed. If the advertiser leaves that bug off the page, Google and thead publishers would not be paid for all the customer referrals they’re sending to the advertiser.
Now, Google likely has some sort of validation process at the beginning to ensure that the bugs are installed at the beginning of pay-per-action ad campaigns. But, after initial validation, how would they tell if the bugs are present? Naturally, it would be questionable if Google is sending lots of click-throughs/referrals to the advertiser’s site and none of those converted. But, what if the advertiser is randomly displaying the tracking bug on their checkout pages only half the time, or one-third, or one-fourth? Google surely couldn’t always detect the lack of the tracking bug.
If an advertiser did this sort of an exploit, they could get a lot of free advertising at the expense of Google and AdSense publishers.
Now, I posed this exploit theory to my colleague, Alan Rimm-Kaufman  who is an expert in the paid search industry, and it only took him about five seconds to find a major flaw in my thinking. According to him, this could indeed potentially give an advertiser some free advertising exposure, but it would eventually bite them, since their apparent conversion rate would be lower than Google’s threshold, which would ding the advertisement’s Quality Score . Once this happened, the ranking of the ads would start dropping and their cost to the advertiser would increase, counteracting the value of cheating.
Alan’s surely right, but I don’t think this entirely discounts my concern. There will surely be other people who would realize the same chain of reasoning I had: if the tracking bug isn’t present at checkout, an advertiser won’t get charged anything. There are likely to be some number of advertisers who are unethical enough to do this while also being so unsophisticated as to not realize that they may eventually end up getting charged more. Some of them might even not care about the eventual impact, and be entirely happy with getting loads of advertising at half-cost, third-cost, or quarter-cost.
If I’m right, pay-for-action ads appears to turn the big fraud problem with internet advertising completely on it’s head – advertisers wouldn’t need to be concerned so much with paying for invalid clicks, but Google and AdSense publishers now have to worry about being cheated by advertisers!
If you’d like a great overview of the older pay-per-click problem, check out the BusinessWeek article on Click Fraud  from last year: