It seems like every few weeks I hear a pitch for a new pay-for-performance ad network of some sort. Many of these pitches involve clever algorithms to duplicate the success of search in the realm of text or display ads on small to midsize content publishers.
A few weeks ago, I heard a pitch from a Cambridge Ph.D. and his business pal regarding a contextual targeting scheme. He has developed a means of contextual targeting that relies on a semantic database to determine relevancy, rather than keyword frequency and proximity.
“Who hasn’t?” you ask. Well, his twist is that the semantic relationships are dynamic and specific to a user. The relationships adapt as a user interacts with the system. Great stuff. A few days ago, I heard another pitch for an ad network based on another slew of proprietary algorithms. The crew behind this one had curriculum vitae that were, let’s say, humbling. Despite the good ideas and well-developed business plans, I don’t give either network good odds for success.
The problem with performance-based ad networks is that they all compete for a limited supply of prospective buyers. There is this idea that every visitor to a blog site or content publisher is a nascent buyer. It just isn’t so. Evidence of this is found in that click-through rates on contextual ads pale compared with their search counterparts. In a recent analysis of Google advertising, I found that Google’s AdSense contextual ads had an average click-through rate of 0.3 percent versus 3.3 percent for its AdWords search ads.
If we look at advertising as a game, with consumer dollars representing points earned, there are roughly a fixed number of points to be earned from a given population. One advertiser’s gain of consumer dollars is another advertiser’s loss. A game with fixed total winnings regardless of outcome is what is known in game theory as a zero sum game. A game of poker with a fixed cost of entry is an example of a zero sum game.
Even if every player draws flushes and straights all night, the number of hands won and the total winnings don’t change. A blackjack table at a casino is not a zero sum game: It is theoretically possible for everyone to walk away a winner after a night of play.
To understand the dynamics of the online advertising game, we have to look at how the winnings play out. We know before the game begins that a lot of points go to travel, cars, financial services, big-ticket electronics, etc., and most of those points will go to established brands. Advertisers selling niche products, unknown brands and smaller-ticket items end up with fewer points no matter how clever or sophisticated their advertising strategy. To continue the analogy of the poker game, it’s as if some players can beat a royal flush with three of a kind.
Why does this matter to the pay-for-performance ad network? A contextually targeted ad for a niche product may still not perform as well as an ad that isn’t targeted at all but represents a big segment of consumer spending. A targeted ad for the Segway scooter, for example, will be hard pressed to outperform an ad for a Honda Accord simply because a lot more people buy Honda Accords. Contextual targeting has its place, but it is just one factor in the mix that includes behavioral targeting, demographic targeting, reach and frequency optimization and plain old brand equity.
Still, a few clever algorithms will come along and make a difference. I thought goto.com (now Yahoo Search Marketing) was a load of malarkey when I first heard about it. Shows what I know.