Can changing the unit of the account save publishing?

That is precisely the claim in a new report, The Economics of Online Advertising, released by ComScore CEO Magid Abraham today. Here is a summary of his claim:

The very interesting thing about the online ad market is that it’s the only market where the accounting mechanism that you use will radically change the economic structure of the market. By that I mean, you have one accounting method that makes the market have infinite supply, which is basically the served impression method. If there is more demand for advertising, people can put more ads on a page very easily with very low incremental costs. Those kinds of markets are very unattractive because they become very commoditized, and the prices drop down to the marginal cost of an additional ad unit. On the other hand, if you look at a viewable impression metric, then that turns the market into a fixed supply market or a nearly fixed supply market. People are not spending unlimited time surfing the web. They spend on average one hour and 10 minutes a day, and in that time there is only so much real space they can cover. Therefore there is only so much advertising that will come into view, and we get essentially a limited supply market versus an infinite supply market. The equilibrium position for a limited supply market is radically different. In an infinite supply market, prices fall to the lowest marginal cost of producing something. In a limited supply market, pricing is highly sensitive to demand, and every increase in the value of an advertising unit will immediately translate into the price.

This is a little hard to parse but his claim seems to be that using current measures of advertising there is infinite supply but if you change to a new measure there will be limited supply. But how can there be infinite supply? One way or another an ad is only valuable if it is viewed whether we measure that accurately or not.

So what is going on here? Let’s start with the apparent problem. For online media sites, many ads are price per impression served. That means that if someone loads a page with an ad on it, the advertiser pays a price. The problem is that ads might be loaded and not viewed. Now that is surely an issue with lots of ads. You might pay for an ad in a newspaper or a flyer put in someone’s mailbox but it may not be viewed. That isn’t what Abraham is worried about; he doesn’t have a technological solution for that. Instead, what happens if you load a long page and there are ads all over it but you never scroll down to see half of them? What Abraham believes is technologically possible (and it should not surprise you that ComScore believes it possesses that technological solution) is that they can measure when an ad is viewed. For instance, if it is actually visible on the screen or if it is viewed by someone in the right geographical area or it is actually served up to a human (as there is potential for bot traffic to distort accounts). Abraham argues that by changing to this method of accounting for ads, both prices and profits will rise because there will be scarcity introduced.

Suffice it to say, I cannot really see how this can be the case; at least, as a straightforward matter of supply and demand. Let me try and illustrate it with a simple example. Suppose that there are two advertisers each of whom would value having an ad actually viewed by a customer at $1 each. In scenario 1, a webpage is served up containing both ads but one of them is ‘below the scroll’ and advertisers do not know which one will be there. Also, suppose that the customer in question never scrolls down. In this situation, each advertiser will have a 50 percent chance of a paid for impression being seen by the customer and so would only be willing to pay $0.50 for that impression. The maximum the publisher could earn in this scenario would be $1 = 2 x $0.50.

Now imagine scenario 2 where publishers only pay for ads viewed. In this case, they know what they are paying for and so are willing to pay $1 for a viewed ad. So the price per ‘unit’ is now higher. But what about total revenue? Well, with our non-scrolling customer, that customer still only sees one ad. We just know which one it is. So total revenue based on viewed impressions will be $1; the same as under scenario 1.

The ComScore paper provides a more complicated example than this and arrives at a conclusion that revenue will rise. But the reason this happens is that while I assume that if a share, x, of impressions are actually viewed when then the price per impression, call it p, would relate to the price per viewed impression, call it P, by the relationship, p = P*x, they assume that p < P*x or specifically that p is much lower than would arise when advertiser maximises expected profits and is risk neutral. So they artificially assume that the current prices are very low.

It is important to emphasise here that improved measurement can improve some things. For instance, it is surely much easier to measure the impact of an advertising campaign if you can determine whether ads are actually viewed or not. But what this cannot do, in any first order way, is change the fundamental supply and demand conditions in advertising. That is, that demand is driven by advertisers who base their willingness to pay on expected viewed impressions (regardless of how they pay for ads) and that consumers ability to view ads is constrained by their limited attention. Changing the unit of account will not change, as a direct effect, the overall revenue going to publishers. In that sense, the paper is far from an accurate depiction of the economics of online advertising.

For more on the supply and demand for advertising (in particular, working through the geographical targeting case) see my American Economic Review paper with Susan Athey from 2010.