I have been fascinated by all the debate around the inefficacy of Cost-Per-Click (CPC) model prevalent with search engines and shopping comparison sites. In his December 4 article in the New York Times, Bob Tedeschi highlighted some of the online shopping newcomers who are attempting to differentiate themselves from the three leading, multi-million dollar shopping sites, Shopzilla, PriceGrabber and Shopping.com. These start-ups have taken up arms agains the Cost-Per-Click (CPC) model and generally argue in favor of a Cost-Per-Action (CPA) or Pay-For-Performance (PPC) model where an advertiser pays only when a user takes a desired action after arriving on his site such as a commercial transaction, registration and such. While the ability to pay only for specific, measurable actions seems a natural evolution of the CPC model, the proponents of CPA are pushing several advantages which may only exist simply because CPA is in its infancy or because they refuse to see the similarities and direct parallels between the two models. Here are the myths and realities of Cost-Per-Click and Cost-Per-Action models:
1. Cost-Per-Click model has gotten very expensive. Cost-Per-Action will alleviate this pain. Price is determined by the marketplace and how much a seller is willing to pay to acquire a lead or a transaction. Cost-Per-Action will be no cheaper or more expensive than Cost-Per-Click once there are enough buyers and sellers participating in the marketplace.
2. Cost-Per-Action model offers a better user experience than Cost-Per-Click. Proponents of Cost-Per-Action model claim that this approach yields a better user experience since shoppers are shown products more relevant to them, not simply the items from the highest-paying bidders. Wrong. User experience is driven by the quality and relevancy of the advertisement and not the mechanics of the monetization. Advertisements that are hand-picked by a publisher for a given piece of content (whether that publisher is paid using CPC or CPA model) tend to be of higher quality than those picked using an automated system. An automated system is always prone to fraud, mis-representation, mis-interpretation of the user context and aggressive campaigns that can compromise user experience. Problem is that people use Googles AdSense and CPC interchangeably. Google could implement monetization using CPA with AdSense and that isnt likely to improve the quality of the user experience.
3. Cost-Per-Click model is prone to click frauds costing advertisers money and Cost-Per-Action model will eliminate that. This is the more interesting discussion. In a steady state, the money an advertiser pays for a click on an ad (CPC) is directly proportional to the quality of the channel. If the quality of a channel degrades (resulting in a lot of unqualified leads), conversion rate goes down. Lower conversion increases the cost of customer acquisition as a percentage of revenue (and reduces margins). This puts downward pressure on CPC because advertisers need to maintain certain margin to sustain the business. So, lower the quality of a channel, lower the CPC that the marketplace is willing to pay for a click. Or conversely, marketplace will pay higher CPC for a lead from a higher quality channel. Marketplace takes care of the click fraud.
The above argument works when the noise (fraud) is uniformly distributed across a given channel. If an advertiser is getting its unfair share of unqualified leads (compared to its competition) then we have a problem. This advertiser may have to (i) live with the higher customer acquisition cost and therefore lower margins (scarcity of qualified channels will demand this) or (ii) find a more effective and efficient channel that doesnt isolate this advertiser and helps him meet his margin goals. The reality is that the CPC model is a win-win for both publishers and advertisers/sellers. This model makes publishers focusing on providing the desired information on their site and referring a user to an advertiser for complementary information, products and services. Advertisers focus on providing the best advertisement to attract the most qualified leads and then converting them efficiently to meet their desired margin goals. Better job you do as a marketer, more you gain with improved margins (or lower acquisition cost). Paying for your actions (clicks) cultivates good behavior in general. Just like any other business offline, you build your channels, nurture the best ones, make them effective and efficient (with minimal or no hand-holding) and make it a win-win for both sides. If publishers cheat by compromising user experience, they pay with reduced CPC determined by the marketplace (or the advertisers will move to a higher quality distribution channel). If advertisers cheat with irrelevancy, they pay with higher acquisition cost (because their competition is going to do a better job of acquiring more qualified users).
Of course advertisers love the CPA model. Why? Because it bypasses all the risk and hard work for them. Since an advertiser doesnt pay for a click, it creates an incentive for them to focus more on getting "free clicks" by being aggressive with their advertisement sometimes at the cost of user experience (that demands understanding the user need, the context and satisfying that need in the most efficient manner). More energy is spent at the front end of the funnel and not in acquiring the most qualified users and satisfying their need effectively and efficiently. An advertiser can do a really lousy job, give a terrible experience to acustomerand not pay a dime for all the hard work a publisher did in sending him a valuable lead. This is good neither for customers nor publishers. I call it a win-lose situation.
Yes, the deficiency with "free clicks" can be overcome over time by using a combination of bid price and conversion rate to determine the relevancy/position of an ad. If you thought using the bid price and click-through data was hard (that took Yahoo multiple years to understand and use with their Panama release), think again. Determining the conversion rate (for CPA model to be viable) and using this data to eliminate spam automatically is an extremely hard endeavor. The best that can happen is that the quality of ads will be as good as with the current (CPC-based) systems. The worst that can happen is that the system will reach a lower-quality steady state because majority of the sellers will be somewhat aggressive about getting (free) clicks, compromising customer experience. Google Checkout is a critical piece for CPA model to work but I remain skeptical about its promise of building a higher quality system. Of course, we at Retrevo will use whichever model results in a better marketplace and user experience. You pick your religion!