- Serve ads
- Measure and track performance and billing
Lately, two new functions evolved from these basic functions:
- Ad delivery optimization
- Ad media exchange
Let’s evaluate the evolution of these two functions in relation to the ad exchange business model.
Ad delivery optimization evolved for two major reasons: The first being the ad server’s need to distinguish themselves from their competition, thus drawing more publishers and advertisers (and later ad networks) to their system rather than to the competitions.
The second reason was that more and more advertisers demanded result-based models, and since data was abundant, ad servers could use some basic methods of targeting and optimizing using the data they collected. This would enable them to provide the result-based environment that their clients demand.
The second function (ad media exchange) evolved for the simple goal of increasing the traffic going through the ad server, thus increasing its potential (and eventually, actual) revenues.
Let’s evaluate now how these two additional functions are performing today and what are their advantages and disadvantages.
I will start with the easier one to evaluate, ad media exchange. This function is performing extremely well for the ad server. It has increased traffic dramatically and generated much higher CPMs, a product of the fact that matching between impression and campaign is done on a much larger scale, allowing for a much higher probability of good matches.
As for ad delivery optimization, well, this is performing much less effectively than the previous function discussed. It is, undoubtedly improving conversion rates but it raises two questions — at what cost? And how well are they doing it? Meaning, can other people (technologies or systems) do it better?
The answer to the first question is “high.” The cost is very high on three levels: The first is obvious — manpower, the second is also quite obvious – CPU and disk space. The third is a bit less obvious — performance and latency. The extremely complex and data-heavy algorithms involved in optimizing huge amounts of data cause performance and latency issues for the whole network at times, and as we know, the internet and this industry in particular is very sensitive and unforgiving to latency and slowness issues. On various occasions we have witnessed revenue decreases which can directly be attributed to latency issues.
The answer to the second question of how well ad media exchange is doing is that the ad servers are doing a fairly well job in optimizing campaigns with their limited resources (data and algorithms) but there are people and systems and software out there that can do an exponentially better job of it. I am sure that as I am writing these lines, hundreds of developers and mathematicians are working tirelessly on developing targeting and optimization methods designed to do this task 100 times better than it is done today by the ad servers.
The reason I am convinced that this is the situation (other than actual performance increases I saw for myself) is that the ad servers use general algorithms for optimization for all types of verticals and all types of campaigns with mostly all types of media, when all around us new technologies emerge; Contextual targeting, behavioral targeting, semantical targeting, user generated targeting, re-targeting etc. – and the ad server’s ability to integrate these new technologies in order to improve their targeting and optimization capabilities is very limited due to a million other considerations they have to take into account.