Will Google extend the CPA-Model to Search?

For German readers:
Es gibt einen ähnlichen Artikel auf deutsch. Da es aber Interesse an diesem Beitrag seitens amerikanischer Leser gab, haben wir den Artikel ins Englische übersetzt und dabei noch einige neue interessante Aspekte beleuchtet. Es lohnt sich also diese Version zu lesen, auch wenn man den deutschen Artikel schon kennt.

Google recently introduced a new pricing mechanism to Adwords which is currently available for ads targeted to the content network. Rather than paying for clicks or impressions, advertisers can choose to pay when a user clicks on an ad and performs a specified action, such as generating a lead or purchasing a product. Search, Google’s main business which contributes to over 90 percent of its revenue, remains unaffected from the new pricing model up to now.

Google’s entry into the affiliate marketing arena created a lot of buzz and was extensively discussed in many blogs and forums. The following citations clearly describe the differing evaluations of Google’s move:

This is Google’s ValueClick (VCLK) killer. Google has greater resources than ValueClick, a larger advertiser base, and the advantage of being able to offer publishers a full range of ads based on page views [CPM], clicks [CPC] and now actual purchases or leads [CPA].
(David Jackson, SeekingAlpha)

Would Google’s extension into the CPA Affiliate Network world spell the end of CJ? I doubt it.
(Sam Harrelson, ReveNews)

I believe the second statement is more down-to-earth. I don’t think that the major, well established affiliate networks will be squeezed out of the market in the near future, because..

• Affiliate Marketing is more about relationships than technology, and relationships require communication between merchants, publishers and affiliate network providers. Google is a first-class technology provider, but its service quality is at least average.
• CPA implies that publishers bear the whole risk, because they only get paid when a user performs an action. Risk averse publishers might prefer to place CPM or CPC Ads.

Well, this was a short summary of what already happened. Has anyone thought whether Google could extend the CPA pricing mechanism to search? Would the adoption of CPA benefit or harm Google? What would this mean for SEM companies or agencies? Below I will explain the five most important effects associated with the implementation of a CPA-model in search.

Auction mechanism and ranking

Today Google ranks ads by revenue. The advertiser which contributes to the highest revenue receives the first position. Thus, with a relatively low bid and a high Click-trough-Rate (CTR) an advertiser can rank higher than a competitor with a high bid but a somehow irrelevant ad and a low CTR. Implementing CPA in search would not force Google to redesign the current auction mechanism from scratch. Similar to the current ranking mechanism, Google would not rank the ads by the CPA-bid, but rather by revenue, which depends on the CTR of an ad, the conversion rate and the CPA-bid.

Important Note: With CPA-bid I do not mean a fixed amount for any action/purchase, but rather a variable amount for every product. Thus, advertisers have the option to set higher CPA-bids for products with higher margins and vice versa (it’s not a necessity). So this kind of CPA would be slightly different from the current CPA-pricing in affiliate networks, where CPA is often fixed for all products. It would also be slightly different from a common revenue sharing model, as a revenue share is usually fixed for all products. That’s the idea, no matter if you call it product-specific CPA or variable revenue sharing or whatever.

A short example to explain the mechanism I have in mind:

Assume we have 5 advertisers, who bid on the keyword “antivirus software”. In the table below, you find the relevant metrics of that specific auction.

To determine the ranking, Google would have to multiply the CPA-bid, the CTR and the conversion rate of the keyword and the corresponding ad. Thus, the advertiser with the highest ranking factor would receive the first position. The ranking factors indicate which of the advertisers would contribute to the highest revenue for Google. Like today in CPC, Google would rank by revenue.

Google would probably implement a second-price auction, which means that the advertiser does not pay the CPA-bid. To determine the actual CPA, Google would use a similar logic as today. The Quality Score would be the product of the CTR and the conversion rate. The actual CPA of advertiser E would then be calculated by dividing the ranking factor of advertiser C (one position below) by the Quality Score of Advertiser E plus a certain increment (here we assumed one dollar). As we can see in the table below, advertiser E who has the lowest CPA-bid is ranked first, because of a high CTR and a high conversion rate. Advertiser B’s actual CPA would be a minimum amount set by Google.

The implementation of this mechanism would reinforce the trend towards performance-based search marketing (first CPM, then CPC and now CPA). Although CPA in theory is superior to CPC, there are some major challenges associated with the computation of the rank of an ad. It is a widespread strategy of search marketers to target the long tail. The problem is that many of those long tail keywords receive only a few clicks per year. For those keywords it takes a long time to estimate the CTR of the corresponding ad. If the ranking mechanism was based on CPA, it would take much longer (in some cases up to several years) to compute ad ranks.

Information asymmetry is – besides the complexity of ranking the ads – a severe problem. Information asymmetry occurs when one party to a transaction has more or better information than the other party. In this case the advertiser knows much more than Google about the products he sells and about the expected conversion rate. The advertiser would be able to take advantage of Google’s lack of information. He could submit a CPA-bid for a large number of keywords which he knows will never convert (products not stocked, prices too high,…). As Google does not have the vaguest notion of the advertisers’ products and corresponding conversion rates, they would start showing ads for testing purposes. Until Google finally figures out what is really going on, the ads might have shown up thousands of times. By this means the advertiser could get a whole branding campaign for free!

Risk distribution

Today with the CPC-based pricing mechanism the advertiser bears the risk (although the risk is much easier to predict than under CPM). The continuously increasing competition during the last years led to bid inflation. Particularly in the initial phase, when no conversion data is available, PPC campaigns can incur a loss. Some campaigns never turn profitable. This high risk combined with risk averseness and constrained budgets represents a high entry barrier for new advertisers.

A CPA-Model could encourage all those advertisers who were not willing to bear the risk associated with the CPC-model. The advertisers’ risk of losing money could be significantly reduced by adopting a CPA-model.

Click Fraud

I think click fraud is for PPC advertising what page rank is for SEO: extremely overrated. Companies like Click Forensics intentionally create fear to sell their click fraud detection software. In fact, Google already filters a large portion of all “invalid clicks”. Their systems even seem to be advanced enough to detect “low noise attacks” by clickbots. Furthermore, the residual fraudulent clicks – that without any doubt do exist – contribute to an increasing Click-trough-Rate (CTR), which tends to have a slightly positive impact on the CPC. Thus, to a certain extent click fraud is inherently limiting its own impact.

Altogether click fraud not only causes Google a lot of trouble, but also tends to limit the investments in PPC advertising, as click fraud has a negative impact on ROI. Adopting a CPA-mechanism in search, Google would eliminate click fraud immediately. The long term increase in trust and the higher ROI would result in higher investments in PPC and therefore favour Google.

Reach

Replacing the current CPC-mechanism by a CPA-mechanism would cause advertisers who are unable to specify conversion goals to drop out. Many B2B-Businesses for example engage in CPC, but they cannot measure their conversions (long buying cycle, customers don’t buy products online and don’t contact them via email/contact form,…). There also might be many advertisers who just place sponsored links without any tracking at all. So I assume that Google would not dare to loose all those customers by replacing the proven and successful CPC-mechanism. Therefore I assume that Google will – if at all – offer CPA as an option, rather than forcing anybody to adopt CPA.

Disintermediation

For agencies and affiliates the possible adoption of a CPA-model is a major threat. The main benefit of their service is a) the assumption of risk and b) the efficient campaign management (set-up, optimization). As mentioned before, risk reduction would not be a major reason any more to outsource PPC advertising. With regard to campaign management and optimization, things wouldn’t change significantly as the complexity remains the same. Today it is essential to solve the trade-off between higher bids which lead to an increase in volume and the resulting lower profit margins of the products (because of higher bids). Advertisers have to estimate whether the negative effect of higher CPCs is offset by an increase in volume. Basically the same logic still holds for CPA, where advertisers would have to decide whether the expected increase in volume would justify a higher CPA-bid. Moreover, the whole process of keyword research, keyword clustering and copy writing which requires skills and technologies doesn’t change. However, I expect the reduced risk to play a decisive role. Many companies don’t understand the process of PPC campaign management and therefore don’t see the benefit of outsourcing the task to specialized companies when they can do PPC in-house and risk-free.

Conclusion

The adoption of CPA would come along with the following effects:

• Some advertisers who have not engaged in PPC advertising because of the underlying risks would make an attempt to advertise with Google. On the other hand, a change from CPC to CPA would force some advertisers to drop out. Thus, Google would not give up the CPC-mechanism. They would rather offer advertisers to select one pricing mechanism.
• Under CPA Google would bear a higher risk, because they would not get paid for clicks any more.
• Ranking CPA-based ads would be a challenge for Google, because it would take a long time to gather the required conversion data.
• Information asymmetries would allow the advertiser to take advantage of Google’s lack of information. Savvy advertisers could get branding campaigns for free (to a limited extent).
• Click fraud could be eliminated immediately.
• Agencies and PPC affiliates would loose customers who decide to manage campaigns in-house because of the reduced risk.

Altogether there is a multiplicity of factors that give rise to several opposed effects. As it is unclear which of these effects is likely to predominate, it is difficult to predict if the extension of CPA to search would benefit or harm Google on the long term. I assume they will try to find out if it works for them in the course of a closed beta test.

Bill Gross, the founder of GoTo, might be afraid that history could repeat itself. Gross came up with an innovation ahead of time, when he invented the PPC model which afterwards led to Google’s fascinating breakthrough. In 2004 the same guy started Snap, a search engine which allows advertisers to place ads based on a CPA model. Sad but true: By adopting CPA, Google could dash Gross’ last hope to build the worlds best search engine.

ShareThis

Der Beitrag wurde am Donnerstag, den 16. August 2007 um 21:23 Uhr veröffentlicht und wurde unter Search Marketing abgelegt. Sie können die Kommentare zu diesen Eintrag durch den RSS 2.0 Feed verfolgen. Sie können einen Kommentar schreiben, oder einen Trackback auf Ihrer Seite einrichten.