Some historical context

In the early days of internet advertising (at least as far as I can remember) banner space was sold on per-impression basis. The prices were quoted in CPM – cost per thousand impressions. This model is basically the same model used in print, TV or radio advertising, just adapted for the web. The concept is simple and straightforward – publisher sells ad space and views, advertiser decides if the platform is a good match for his products and services.

Since it’s way easier to track results and effectiveness of internet ad campaigns than in “traditional” media, it’s understandable that advertisers are eager to pay for the ads on per-result basis. For businesses selling goods, subscriptions, information, etc. the result is an actual sale, so the perfect metric for them would be an event of money changing hands. This materialized in affiliate programs. The problem with affiliate programs is that publishers have to trust in their ability to sell advertiser’s products and in most cases this proves to be one step too far.

Moving one step back from the actual sale is a fact that a user actually came from publisher’s web site to the advertiser’s web site – in other words clicked on a banner ad. The model is known as Pay-per-click. There still has to be some belief on publisher’s part that the products or services being advertised are a good match for it’s audience. Otherwise no one is going to click on these ads and the only asset publisher has will go to waste (even though zero-click campaign could still be valuable to brand advertiser).

This problem was partially addressed with proliferation of contextual advertising led by Google with it’s AdWords product. The system matches relevant ads to the content of publisher’s web site, which (in theory) results in ads with higher probability of being clicked (good for the publisher) while advertiser pays only for some sort of a result (click through). Problem solved. Everyone is (theoretically) happy. After all there’s basically an infinite amount of ad space on the web and a finite amount of advertising dollars. So the market has to lean on the buyer’s side.

Mobile advertising is a new rage. It’s new, it’s hip, it gets better click-through rates (more on this below). And it brings the pay-per-click model from the desktop web to the mobile apps.

Problems with pay-per-click in mobile apps

No context

Most mobile apps have no content or content never changes. So the ads are mostly targeted by categories and/or keywords selected by the app’s developer. On networks with low(er) fill rates publishers are led to select categories and keywords with more/better ad inventory. And the contextual matching is basically in the hands of publishers (via category/keyword selection). So it doesn’t change automatically when new advertising campaigns launch. This results in less relevant ads, less clicks and less revenue. Bad for publishers.

Fraud

Click fraud is a big problem in PPC advertising. Even the Wikipedia article on click fraud is longer than articles on pay-per-view and pay-per-click advertising combined. Networks are able to fight major fraudulent activities to some extent, but this mostly matters to networks. For a regular small advertiser, with cost-per-click reaching more than $1 in some hot areas, even a few fraudulent clicks could be hard to swallow. And how often do we see something like “please click on this ad to support future development” or something like that? Do we know how many developers asked their friends and family to click on an ad from time to time? This is in no way specific to mobile advertising but is an inherent problem of the pay-per-click model. Bad for advertisers.

Accidental clicks

According to a survey conducted by Harris Interactive on behalf of Pontiflex:

… 47 percent of mobile app users say they click or tap on mobile ads more often by mistake than they do on purpose.

There’s not much I can add to that. Except that I believe this to be true. Bad for advertisers.

What to do?

The same article mentioned above proposes basically the same old affiliate program model. This model could work for campaigns negotiated on case-by-case basis when there’s a good match between the app and the advertised product, but in general ad network scenario it would probably result in very low revenue for the publishers. And it’s useless for advertising mobile apps unless app stores create APIs to track these leads.

Microsoft Advertising uses a mix of pay-per-view and pay-per-click model in their pubCenter network for Windows Phone. As far as I understand advertisers bid on per-click basis, but publishers get paid per-impression with rates calculated by some clever algorithm. Publishers are happy. Not sure about advertisers. As far as I know you can’t buy an ad campaign to run on the phones exclusively (unless this changed recently) so this is probably something no one really knows at the moment.

Unfortunately I don’t see anything more balanced than the old pay-per-view approach. It’s simple, predictable and difficult to abuse substantially by amateur cheaters (aka not very honest publishers). That’s what we use at AdDuplex for the time being. Until some bright minds create a model that suits everyone better.