Peter Caputa linked to a barnburner of a post today about the future of advertising. Peter Hershberg argues that the future of advertising is in cost-per-action models – a model where payment is made only when the advertising leads to a specific action, like making a purchase. Peter’s expanded on this notion, pointing to some of the most interesting companies in advertising. (He was too modest to mention his own.) It all sounds very plausible, but I’m not quite buying into it. I think we’re in agreement that the CPM model – in short, banner ads – doesn’t have a strong future; although it’s going to make Google a bushel of money in the near term as old-school companies transfer their money from the networks to the ‘net, buying up eyeballs indiscriminately is becoming less and less effective over time. But cost-per-action advertising models have several disadvantages, both for advertisers and publishers, that cost-per-click models don’t.
First of all, CPA models aren’t something new that the market is ‘moving to’ – they’re already here, as affiliate marketing programs. And affiliate marketing programs have big problems with tracking. Programs that set and read cookies are susceptible to cookie deletion. Programs that include affilate IDs in the URL are susceptible to affiliate-ID replacement by scumware. Not so for cost-per-click, since it’s a simple thing to track people exiting your site through a link.
Next, cost-per-action models force you to compensate for the quality and effectiveness of the advertiser’s own site. Even if their products are identical, you can’t just prefer a $6 CPA advertiser over a $5 CPA one, since the $5 CPA advertiser might convert like gangbusters, and the $6 CPA one could be a dog. You have to experiment, constantly montioring advertiser performance. The advertisers you decide not to run could become more profitable than the ones you’re running at any time – and unless you’re constantly returning to the clients you’ve rejcted, you’ll never know when this happens. Ugh. Not so for the cost-per-click model. If two advertisers’ products are identical, your choice is easy – pick the one with the highest cost per click.
Finally, cost-per-action models have one huge disadvantage for the advertiser – they give away too much information. The advertiser has to transmit his conversion to the publisher – it’s right there in the amount of the cheque divided by the number of leads sent. If the deal’s a revenue sharing one, the advertiser’s sending back even more sensitive information – the exact return on investment for the advertising spend. This is information the advertiser doesn’t necessarily want to give away. On the other hand, successful publishers love it – when campaigns are going better than expected, it allows them to push back and raise rates.
There’s one advantage of cost-per-action over cost-per-click, of course – it’s a whole lot easier to commit ‘click fraud’ than it is to commit ‘action fraud’. I suspect that’s why cost-per-action is being touted as the next generation of internet advertising, even though affiliate marketing has been around for years. But the click fraud problem is a temporary one, one that will evaporate as soon as the Internet’s pervasive anonymity glitch goes away – as it surely will, probably sometime in the next five years. As your personal internet appliance grows ever more portable, as devices increasingly come with personally-identifiable information baked right into the hardware, and as the quality of personalized services grows to the point where you wouldn’t dream of remaining anonymous, advertisers will be able to spot fraudsters with increasing ease. A decade from now, advertisers simply won’t bother to serve ads to individuals they don’t already know something about – the portion of non-identifiable traffic will be so insignificant and poorly-converting they’ll insist on having nothing to do with it. Cost-per-click’s inherent disadvantage will have disappeared, but it’s advantages will remain. That’s why I’m predicting CPC will continue to have a long life ahead of it – it’ll exist quite healthily alongside CPA models. That variety should be good news for advertisers and publishers alike.
