Adsense and Agency

Google's AdSense program (the program which is serving ads on this site) presents a twisted version of the agency problem, where Google is acting as an agent for parties (advertisers and content providers), but has an incentive to act against the best interests of the parties it represents.

An agent is a person or organization hired to represent some other person or company, and carry out certain activities on behalf of the represented person. A classical agency problem arises when the financial incentives of the agent run counter to the best interests of the person the agent represents. For example, a stockbroker with trading discretion (the right to make trades on behalf of a client) is supposed to act in the client's best interest, but gets paid on commission when trades are made. The all-too-common result is that the broker makes unnecessary trades in the client's account, racking up commissions but doing little good for the client.

AdSense presents this problem with a twist. Now there are three parties involved: the advertiser, Google, and the content provider. Google is acting as an agent for both the advertiser and the content provider, and keeps some fraction of the advertising fees the content providers pay. I don't know what the fraction is, because Google decides, but I'm guessing it is half or more. The problem arises in that Google can make more money by shortchanging either the advertiser or the content provider (or both).

Google can act against the interests of the advertiser by presenting inflated click-through rates, or by charging more per click than the market rate. Inflated click-throughs are relatively easy to audit and detect, since the advertiser knows how many surfers arrived through a particular ad, and can measure the quality of the clicks fairly easily by tracking the conversion rate (i.e. what fraction of click-throughs resulted in a sale). A sudden jump in clicks, combined with a drop in conversion rate, is a strong indicator of someone monkeying with the ad. In fact, my own company stopped advertising through a particular online company a couple years ago when we suspected the company of inflating the click-through rate, and they wouldn't provide us with the data to audit their numbers.

Charging more than market rate for a click is harder to detect, however. Google's ad programs work by essentially auctioning off ad slots to the highest bidder. Advertisers specify a maximum cost-per-click they're willing to pay, and Google is supposed to charge a fair market price based on advertisers' bids for a given slot. If there is only one advertiser for a slot, the click goes for a nickel. The potential for abuse arises because no advertiser knows what anyone else bid for a given click, or even what ads filled the other slots. In theory, no advertiser should pay more than a penny above the second-highest bid for a slot, but there's no way to know. Google could easily be using a bid increment of a nickel, or even thirty-seven cents, and none would be the wiser.

Google also gets to choose the percentage paid to the content providers. Since this percentage is never disclosed (and could be different from site to site, day to day, or even slot to slot), there's no way to know if the percentage is going up or down with time. If Google, as a newly-public company, ever needs to boost its profit a bit, they could drop the payout from, say, 50% to 45%, and few would notice the difference.

Right now, smaller content providers have no real options other than AdSense, but in the long-term, competition is the only way to keep this abuse in check. Since switching from Google to a competitor would be simple (if such a competitor existed), content providers would quickly go wherever they could make the most money. Many would probably even use both, and allocate ads based on who's payout was highest last week.

The really interesting agency dilemma has to do with fraudulent clicks from content providers. A content provider using AdSense owes no agency duty to the advertiser: the content provider is acting in his best interest only. So what is to prevent a content provider from clicking on every ad in sight to make more money?

Google has to prevent this fraud (by its duty to the advertiser), even though Google makes more money if the fraud continues. Fraud is controlled by kicking off content providers who appear to be inflating their clicks, though Google also owes a duty to the content providers to avoid falsely accusing them of fraudulent click-throughs.

Earlier in the AdSense program, several content providers complained about being inappropriately kicked out of the program, without explanation or recourse (and often, without the money they felt they had legitimately earned). This is where Google's twisted agency problem really gets ugly.

How, exactly, is Google supposed to reliably detect click-through fraud and eliminate cheaters without violating its duty to either the advertisers or the content providers?

Google can never release details of its Fraud Detection Algorithm, because that would tell would-be fraudsters how to get around the safeguards in the system. A side-effect of this is that content providers kicked out of AdSense can never be told why, which means they never get the opportunity to defend themselves. Furthermore, no Fraud Detection Algorithm can be perfect, so some false accusations are inevitable.

For example, the most obvious thing to look for is a sudden spike in impressions and clicks for a given site. But, as any webmaster knows, spikes in traffic of several orders of magnitude are common. When this blog was slashdotted a few months ago, we went from a few visits per day to 20,000 visits in 24 hours, and back again.

Other obvious warning signs, such as an unusually high click-through rate, can be just as innocuous. The click-through rate might jump because Google started serving a more relevant ad, or because the traffic mix to the site changed.

What it comes down to in the end is trust. Advertisers need to trust Google to police the content providers and price ads fairly, and content providers need to trust Google to give them a fair cut and not be arbitrary in their decisions to label someone suspect.

As Google becomes a public company--and all the quarterly financial pressure that implies--the incentives to violate the trust of both advertisers and content providers will only grow. It will be interesting to see how Google deals with this position.

Previous
Previous

Not Too Much Taxes, but Too Many

Next
Next

Why can’t the web be more like television?