You can’t manage online ad inventory like a stock market

George John is CEO of display advertising technology company Rocket Fuel.

It’s fashionable to talk about online ad inventory as a commodity that can be traded on exchanges and priced fairly through a market mechanism, just like trading stocks on Wall Street. The analogy misses important differences between online ad inventory and commodities, ones that create opportunities for companies that have the technology to capitalize on the distinctions.

[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":162816,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"business,","session":"A"}']

Here’s the thing — online ad impressions are more like snowflakes than stocks: no two are exactly alike, and they melt.

For example, a share of Time Warner is a share of Time Warner, but a single ad impression on Time.com is not exactly the same as any other: not only will future impressions occur at different times obviously, but they will be to different users with different interests and different previous ad interactions, in different geographies, with different demographics, different browser versions, different ISPs, etc. The value even depends on the buyer, and the relevance of that buyer’s ads to the current user — the opportunity to serve an ad to a luxury retail shopper on Time.com is worth more to Nordstrom than to IBM. Even worse, a share of Time Warner will exist in perpetuity, but a single display adserving opportunity comes into existence in real-time when a browser visits a page, and it spoils fractions of a second later when the ad is required to load.

This makes estimating the value of an online ad impression exponentially harder than estimating the value of a financial security, in the true sense of the word “exponential”, since the addition of each new factor describing an online ad impression multiplies the number of statistics requiring estimation by the effective cardinality of that factor. Certainly, quantitative trading houses also build multi-factor models to forecast prices on financial securities, but the point is that ultimately there is in fact a CUSIP or ticker symbol that identifies the security, and there is no such analog in display advertising except for the vector of factors that describe an impression.

The online ad marketplace has progressed through significant phases of evolution, from the early days dominated by relatively indiscriminate forward contracts on bulk bundles of ad space negotiated via publishers’ direct sales teams, to today’s increasing prevalence of exchanges with real-time bidding that allow parties to express bidding strategies based on their own mathematical functions applied to all of the descriptors of a single online ad impression to calculate a bid. In theory, the creation of this kind of market, that is perfectly liquid and allows expression of arbitrary complex bidding algorithms, should have resulted in amazing value creation for advertisers. In practice, the challenge is that the invisible hand of the market is at this point controlled by the immature brain of the market, but this will change.

On a recent day across a handful of real-time exchanges, we saw that over 35% of winning bids were multiples of $1, and over 54% were multiples of $.10 or $.25. Participating in a real-time bidding exchange but only doing this kind of manual bidding is like having a Ferrari and only using it to drive to the grocery store — you can brag about it to your friends, but it’s not really adding any value. This doesn’t indicate that real-time exchanges are a failed enterprise; instead it shows there’s a massive opportunity here for someone who can develop the rocket science that can run rings around the kind of unsophisticated players who are bidding $1 on every impression.

Of course, it’s self-serving to say that technologies like the ones my company offers are the solution to the problem, but they are: What unlocks the full value of the ecosystem for both advertisers and publishers is an intelligent arbitrageur, with a twist. In finance an arbitrageur adds value only insofar as it creates wealth for itself and generally serving to equalize prices to “fair” levels; but otherwise arbitrage is a zero-sum game. By contrast, in online advertising, value is created by optimally matching demand (ads) with supply (ad space), which fulfills the “right ad, right user, right time” promise. By doing this well, everyone wins — publishers are more fairly rewarded, advertisers can get more responses for the same spend, and users see more relevant ads.

The rocket science involved in doing this well is a combination of a number of components such as predictive modeling, adaptive control, and portfolio optimization. Predictive modeling means using all of the information about an impression to predict the likelihood of response for a given ad, so that an intelligent system can determine the impression’s likely contribution to performance goals. Similar technology is used in many marketing contexts from direct mail to Netflix movie recommendations. Adaptive control involves micro-adjustments to campaigns in flight to keep key objectives like performance and delivery pacing within a desired operating range. This is also the technology that keeps a Segway upright and allows pilots like Captain Sullenberger to glide a 150,000-pound Airbus 320 to a gentle river landing. Portfolio optimization considers an array of potential investments and allocates assets optimally to achieve a desired risk/reward blend, and as the name implies, is similar to its financial analogy.

So the solution is intelligent bidding and adserving. Like Jimmy Carter telling us 30 years ago that we need a coherent energy plan including conservation, the solution has the ring of truth but is also hard, which is where full-service media buying services like my own come in. Instead of forcing an agency media team to jump into the deep end and trade in its day planners for lab notebooks and ditch Photoshop for Mathematica, such services handle the technology and campaign management, and let agencies focus on clients’ strategic needs.

[aditude-amp id="medium1" targeting='{"env":"staging","page_type":"article","post_id":162816,"post_type":"guest","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"business,","session":"A"}']

George John is a bona fide rocket scientist who left a career working on artificial intelligence software for Mars rovers at NASA to work on quantitative finance and then optimized marketing. After stints at Stanford, Lockheed, IBM, E.piphany, Salesforce.com and Yahoo, George founded Rocket Fuel Inc as its CEO in 2008. Rocket Fuel provides an adserving and bidding engine to agencies and advertisers primarily via a full-service managed offering, the Rocket Fuel Network, that is sold like an ad network. Its advertisers include many of the top brands in the world such as Infiniti, Lord & Taylor, and BlackBerry.