You are here

How only Google can pull off pay-to-perform ads


Bruce Schneier today compliments Google on trying out pay-to-perform ads as a means around click-fraud, but worries that this is risky because you become a partner with the advertiser. If their product doesn't sell, you don't make money.

And that's a reasonable fear for any small site accepting pay-to-perform ads. If the product isn't very good, you aren't going to get a cut of much. Many affiliate programs really perform poorly for the site, though a few rare ones do well.

However, Google has a way around this. While the first step on Google's path to success was to make a search engine that gave better results, how they did advertising was just as important. At a time when everybody was desperate for web advertising, and sites were willing to accept annoying flash animations, pop-ups and pop-unders and even adware, Google introduced ads that were purely text. In addition, they had the audacity, it seemed, to insist that pay-per-click bidding advertisers provide popular ads people would actually click through. If people are not clicking on your ad, Google stops running it. They even do this if there are not other ads to place on the page. They had the guts to say, "We'll sell pay per click, but if your ad isn't good, we won't run it." Nobody was turning down business then, and few are now.

Sites of course don't want to be paid per click, or a cut of sales. They want a CPM, and that's about all they want, as long as the ads are otherwise a good match for the site. Per-click costs and percentages are just a means to figuring out a CPM. Advertisers don't want to pay CPMs, they want to pay for results, like clicks or sales.

Google found a great way to combine the two. They offered pay per click, but they insisted that the clicks generate enough CPM to keep them happy.

The same will apply here. They will offer pay for performance, but those ads will be competing with bidders who are bidding pay-per-click. Google will run, as it always has, the type of ad that gets the highest results. If you bid pay per performance, and the PPCs are bidding higher, your ad won't run. And even if there are not higher PPCs, if your ad isn't working and convering into sales and generating revenue for Google, I suspect they will just not run it. They can afford to do this, they are Google.

And so they will get the best of both worlds again. Advertisers who can come up with products that can sell through ads will pay for actual sales, and love how they can calculate how well it does for them. Google will continue to get good CPMs, which is what they care about, and what Adsense partners (including myself) care about. And they will have eliminated clickfraud at least on these types of ads. Once again they stay on top.

(Disclaimer: I am a consultant to Google, and am in their Adsense program. If you aren't in it, there is a link in the right-hand bar you can use to join that program. I get a pay for performance credit if you do. Unlike Google's PPC ads, where Adsense members are forbidden by contract from encouraging people to click on the ads, there is no need for such strictures against pay for performance ads, in fact there's evey reason to encourage it.)


This ultimately blurs the distinction between PPC and PayPerAction -- pay whatever you want for any click, but better make sure you pay enough to continue making it worth Google's while to find you more such clicks. You could pay for all clicks, pay only for those resulting in conversions, or any mixture based on your own internal lead-valuation/yield analysis.

To be really diabolical, Google could let you decide how much you want to pay long after the click was delivered, when you've had a chance to assess its true value. (This might initially take the form of an easier-to-understand 'money-back-guarantee' on all clicks, offering no-questions-asked refunds on any clicks you don't like.) As you say, "They can afford to do this, they are Google."

I wrote about this initially here:

Killing click-fraud (and the competition) with one stone

...and several times since, developing the idea to here:

"Cost Per Whatever" to displace CPC and CPA

In some ways, your proposal hinges on a bigger question -- who actually figures out the best places to run an ad? The advertiser and the ad-seller both want the ad to run in the best places it can, to generate the most results. The advertiser of course wants to generate sales (or whatever their ad campaign is meant to generate) and the more sales they generate, the more they'll pay for a campaign. The ad-seller, Google, wants to do all this with the least inventory, or the least inventory that other people aren't willing to pay more for.

In the end, all Google cares about is CPM. All the advertiser cares about are the results the campaign is geared for.

Most of the time, though, Google is in a much better position to figure out what ads should go where. It's aware of all the various bids, and their performance history, for all advertisers. The advertiser is aware, one hopes, of what clicks actually performed. With cost-per-action, they're now letting Google know that.

Economies of scale suggest Google should be far better at figuring out what to do with the information, once the advertiser shares it as they do in CPA. Google of course has ways to spot click fraudsters too, which the advertiser can't do as readily.

In your proposal, the advertiser has a more detailed way of telling Google what clicks they liked. Of course, Google starts this process by having the advertiser pick the keywords, and refine them with a detailed system. (In the old days you would have had an ad sales rep who advised you on your campaign in a newspaper.)

In some ways, Google could go back to CPM (and is doing so in some cases) if the impressions are well chosen. No click fraud by competitors then. To stop click fraud by adsense partners is more difficult. (Go with CPM and they can just do "impression" fraud.)

I expect most advertisers will just want Google or their other media buys to figure out best what to do, and the market will pick the company that can deliver the best return for the buck.

Agreed, though this statement needs to be qualified: "Google of course has ways to spot click fraudsters too, which the advertiser can’t do as readily."

The advertiser has the best insight into what clicks are 'fraud' (or of negligible value for any other reason). Google is only guessing; the recipient of the clicks can, with suitable effort/tech, essentially *know* what value a click had to their business. It's just that a lot of advertisers aren't that sophisticated, while Google has a concentrated incentive to address (some kinds of) click fraud.

But over time, the expertise the advertisers need can be shifted towards them by software (Google Analytics and other lead tracking/valuing packages), and the knowledge of ultimate click value Google needs can be shifted towards them by pay-by-value pricing (the CPC to CPA continuum).

Some interesting wrinkles are:

  • Google would love to keep all the info about click origins to itself, so that advertisers remain dependent on Google for optimal placement. Perhaps, my hypothesized 'money-back-guarantee' would only be available for 'blinded' clicks. If you insist on seeing the origins yourself before giving Google feedback on what clicks were good, no on-demand click refunds.
  • Time/space rentals based on neither CPM nor CPC nor CPA. To at least some extent, ad inventory is like real estate. You don't find commercial retail real estate charged by the number of walk-bys/walk-ins/sales -- at least not directly. (If you did, you'd get walk-by and walk-in fraud.) Instead, a rent is set by perceived value, as discovered over time and through competition with alternate providers/purchasers. The walk-bys/walk-ins/historic-sales are factors, but contributing factors moderated by judgement and competition.
    In this view, CPM/CPC/CPA are all just transitional pricing models to help determine (and convince conservative purchasers of) the value of a new media and new inventory, in a risk-shifting manner. Once the value is well understood, demonstrated by successful professional exploitation, and traded by a diverse marketplace of sophisticated buyers and sellers, the volatility/variance of CPM/CPC/CPA will be a bug rather than a feature, and most ads will be sold via contracts in time/space rather than traffic.
    (Of course, lots of sites have offered ad placement by time/space for a fixed price, but large brokers for such space haven't yet taken off. It appears a company called Madhens is making a go at it, though, with the fraud-resistence being a major selling point.)

Add new comment

Subscribe to Comments for "How only Google can pull off pay-to-perform ads"