I predicted months ago that click fraud would cause advertisers to lose faith in pay-per-click advertising, and so it is coming to pass, according to a new study by Outsell that estimates click fraud at $800 million (not so far from the middle of my back-of-the-envelope estimates):

The perception of pervasive fraud has prompted many advertisers to change their spending. Many are asking why they should fork over money – significant amounts, in some cases — for phantom shoppers.

The study found that 27 percent of advertisers reduced or stopped spending on click-based advertising. An additional 10 percent said they intend to curtail spending.

Among the predictable responses to this news was this fascinating analysis from Michael Rogers on the potential of cost-per-action to replace cost-per-click:

But CPA is a very slippery slope. The issue goes back to why online publishers traditionally resist cost-per-click pricing for display advertising: your revenue becomes dependent on the advertiser’s creative. A good ad is going to get more clicks than a poorly executed one. CPA action takes that dependency a step further: the search engine’s revenue depends on both a well-designed pitch after the click plus an enticing offer. If the advertiser fails on either of those points, you’re not going to get paid and your inventory isn’t generating revenue.

If I ran a search engine, I’d be spending a lot of time and energy trying to maintain the credibility of my cost-per-click business. If the market really does turn to cost-per-action, we may end up looking back on these as the Golden Days of search engine advertising, when the money just fell from the sky.

First, he’s right that the days of “easy money” from pay-per-click are numbered — it doesn’t matter (Danny) how big click fraud actually is — the system can’t escape the inexorable death spiral of negative advertiser perceptions. Google knows this, and that’s why they’ve been chasing offline media and experimenting with cost-per-action.

But what about Michael’s theory that cost-per-action might not work because of the publisher’s “dependency” on the advertiser having a “a well-designed pitch after the click plus an enticing offer”?

If you simply converted the current AdWords model to a cost-per-action payment, then I think you would have a big problem. Publishers would still be dependent on advertisers writing good ad copy and they’d be even more dependent on what happens when someone clicks off to the advertiser’s site.

But what if there were an approach to cost-per-action that could overcome these problems?

Sounds like an opportunity for someone to eat Google’s lunch.