Research

Should Online Advertisers Pay for Clicks or Sales?

In recent years, the longstanding “cost per click” model for online advertising has started to give way to a “cost per action” model, in which advertisers only pay for a sale or other customer commitment. A new paper by Jiwoong Shin, a professor of marketing at Yale SOM, and his collaborators provides insights into how both advertisers and publishers should think about which model to choose.


By Roberta Kwok

When most people read print newspapers and watched traditional TV networks, ad pricing was fairly straightforward. Advertisers paid for the estimated number of readers or viewers, even though much of the audience probably didn’t notice the ads. “Half of the time, it’s just a waste of money because people don’t pay attention,” says Jiwoong Shin, professor of marketing at Yale SOM. Early online media followed the same system, with publishers charging based on the number of visitors to the web pages where an ad appeared.

Around the turn of the 21st century, a new model for online advertising emerged: cost per click (CPC). By tracking how many people clicked on ads, advertisers pay only for those users who, at least, noticed their ad and likely had some interest in it.

More recently, CPC has started to give way to another model: cost per action (CPA). Under this system, the advertiser pays the publisher only if the user also performs an action subsequent to the first click, such as purchasing an item or signing up for a newsletter.

Online publishers and advertisers are now struggling to settle on one of these two pricing models. “There’s a big debate in the industry: Which one is better?” Shin says. The general assumption is that CPC benefits publishers, while CPA favors advertisers. But a study by Shin’s team in Management Science suggests the issue is more complex.

Using a theoretical model, the researchers found that both parties might fare better under CPA when the product’s sales are uncertain and advertisers want to avoid risk. “In this case, CPA actually makes all the players in the industry better off,” he says.

CPC has some clear downsides for advertisers. The publisher might put little efforts to send the right types of customers to the advertiser’s site. People might click an ad accidentally or only to get rid of an annoying window. Small companies run the risk of click fraud: A competitor could simply click the ad over and over until the firm’s advertising budget runs out.

The CPA model, on the other hand, increases the risk for the publisher. What if the publisher sends many users to the advertiser’s site, but the site is so poorly designed that people give up on making a purchase? Or a customer might click an ad and learn about a product, then visit the advertiser’s site directly a week later to buy it—in which case the publisher wouldn’t get paid. While some technological solutions to this problem are available, the issue hasn’t been fully addressed, Shin says.

Shin and his collaborators, Jeffrey Hu of the Georgia Institute of Technology and Zhulei Tang of CAN Capital, created a theoretical economic model to simulate online advertising. In the model, advertisers bid for spots and decide how much effort to put into designing their sites well; a publisher awards spots and decides how much effort to put into sending the right customers to the advertiser’s site.

The team found that the CPA model could sometimes backfire for publishers because it attracts advertisers who just want to spread brand awareness, without much interest in generating immediate sales. These advertisers might bid high to win spots but not put effort into encouraging purchases. However, publishers can partly address this problem by running a test campaign before awarding the slot to estimate likely sales, Shin says.

The researchers also identified some important conditions. If a product was high-risk—say, because it was new or seasonal—then a cautious advertiser would bid high under the CPA system. That model “shifts the risk from the advertiser to the publisher,” Shin says, so advertisers are “willing to pay a lot for this insurance.” This might be particularly true of small companies that want to avoid risk. Because the advertiser pays a premium, the publisher’s profits increase as well. If a product’s sales are predictable—for example, breakfast cereal—then CPC is better for all involved, Shin says.

Publishers should consider these factors when deciding how to charge for advertising, Shin says. If their users tend to buy products with uncertain sales such as fashionable clothing, then CPA is a better option, he says. “Choosing the right pricing model is really critical,” he says. “Billions of advertising dollars are at stake.”

Professor of Marketing