Sifting The Programmatic Conflation and Conflagration

jed-nahum-no-usethis“Data Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Jed Nahum, Senior Director of Programmatic Sales at Microsoft Advertising.

At the recent IAB Annual Leadership Conference, I attended a cocktail-fueled discussion of programmatic.  The discussion started with a noted publisher asserting “Programmatic is not RTB, it is automation.” It was an assertion I’ve heard before –at Microsoft, also a publisher. Other publishers went on to assert that there is a conflation of RTB with programmatic that they don’t like. Tom Shields also discussed this topic on a couple weeks back.

I agree: there is conflation between the terms, but I believe it’s slightly different than the participants identified.

I believe there are two dimensions that tend to get confused when talking about programmatic offerings.  One is: Programmatic versus Manual (or “Hand Sold”). The other is one of the big issues facing our industry right now, and it asks the question “who gets to define audience: Buyers or Sellers?” Consider the following diagram, which crosses the two dimensions:


While I’ve put examples of offerings into each quadrant, I mostly want to concentrate on the axes themselves, and hopefully “de-conflate” them. In particular, I want to acknowledge and explore the tensions in the audience axis above.

RTB was display’s first automated, programmatically connected offering. When it arrived, the industry was captivated by its ability as a protocol to virtualize much of the intelligence of an ad network. Over time, we’ve realized that RTB lives at the intersection of automation and buyer defined audience. But, because RTB was the first programmatic offering, there is a natural tendency to conflate those two great attributes. But I see the automation and audience dimensions as independent.

If you think about it, the vertical axis of Programmatic vs. Manual is pretty much motherhood and apple pie. Everybody stands to benefit from automation, which solves problems of efficiency across demand and supply. Eric Picard argued recently that programmatic isn’t all about automation and in fact represents an opportunity for better demand and supply matching. I think he’s right. But when we limit our analysis to programmatic vs. manual the conversation naturally turns to efficiency and normally polarized constituents relax and embrace each other a bit more.

Much more divisive and complicated is the audience axis above. As advertisers and buyers tag up sites and build up their stores of consumer data, RTB has gained momentum. And it has triggered a looming battle over audience. The DMP segment in our industry is built on the premise that buyers must accumulate information on the users that they want to converse with. Publishers, on the other hand, react to buyer defined audience as commoditization of supply. The language they use to describe buyer defined audiences leans towards game theory (e.g., “cherry picking,” “cream skimming,” “asymmetrical information,” and “adverse selection.”

Developments in the programmatic axis are easy to forecast because programmatic should not divide the key constituents of our industry: buyers and sellers are united in seeking efficiency.  Things are going to get more automated. On the other hand, getting a fix on the buyer defined audience vs. seller defined context axis is much more difficult. For execs to predict along this axis they need to consider and forecast issues such as:

  1. The relative value of buyer data versus publisher context in driving results for advertisers
  2. The impact of potential regulation on collection of consumer data
  3. The impact of browser providers’ attempts to cater to consumer privacy needs on collection of consumer data (e.g., DNT defaults, Firefox and Safari 3rd party cookie policies, etc.)
  4. Response from publishers to items 2 and 3 above (e.g., potentially ignoring DNT from certain browsers, blocking browsing from browsers that disallow use of third party cookies by default)
  5. Other stuff we haven’t yet identified

It’s reasonably easy to have an opinion on item #1. It’s much more difficult to voice a forecast for items #2-#5 because of their inherent variability and because they interact with each other. Were it not for the complications of #2-#5 above I would argue that buyers defining audience is an inevitability. Outside of registration data, buyer defined audiences tend to be an accumulation of context and so present a greater signal to buyers.  As time passes I think we'll see consumers grow more accepting of online tracking and the value it generates in terms of greater access to content and better relevance in advertising.

For all the reasons above the trend of buyers defining audience will continue. And we’ll even see new ways in which forward markets address the buyer defined audience business. In the meantime, as offerings appear which illuminate the difference between programmatic and audience, programmatic will become a less controversial term.

Follow Jed Nahum (@jednahum), Microsoft Advertising (@MSAdvertising) and AdExchanger (@adexchanger) on Twitter.


  1. Great post, Jed, and thanks for the shout-out. I think your audience axis is pretty interesting, but seems more like a broad spectrum than just two buckets, given the merging of first, second, and third-party data into segments these days. I'm also confused about the lower left "manual buyer-defined" box, not sure how that differs from the top left. But anything that helps bring clarity helps grow the market for everyone - thanks!

  2. Jed Nahum

    Tom thanks for the kind words. And you're absolutely right: I took a few liberties collapsing and compressing a bit along the audience axis. That said, I was trying to represent the relative control exerted by buyers vs. sellers along that axis. You're right that first, second, and third party data becomes merged. But I meant to convey that the buyer defines the target rather than the publisher (using, as you point out, a merging of various data sources).

    WRT the lower left corner, maybe manual is less appropriate than hand sold. Retargeting on ad networks or against publisher ad servers, for example, has been less automated, but uses buyer controlled data. Holding yield constant, a publisher should probably prefer the RTB quadrant because it's more automated and higher efficiency. I said holding yield constant. :)

  3. Great article Jed, especially the part about me. :-)

    I still would argue that RTB as an ad serving mechanism sits outside of the discussion of the source of the data, or the point of view of the data. The tendency is to label the technology as driving the behavior, but I think the two issues are orthogonal. I know my point of view is a bit academic so far, but that will change.

    Especially with the 3rd party Cookies issue you mentioned above, the information assymetry problem publishers hate about RTB will collapse a bit. With the move to first party Cookies and domain mapping publishers will gain much more control and access to the demand goals. This is a good thing, more transparency between buyers and sellers is valuable. But this is the topic of my next article so I'll hold my cards until then.

  4. Jed Nahum

    Thanks Eric. Putting stuff in the quadrants may have wound up being distracting. And RTB is not a product offering -- it's really a protocol that enables offerings we haven't named except to call them RTB. So I agree with your statement that "RTB as an ad serving mechanism sits outside of the discussion of the source of the data, or the point of view of the data," or, in my parlance "whether the buyer or the seller defines the target."

    My thesis above was really "people confuse programmatic with the polarizing issue of defining audience and therefore they resist programmatic more than they should."


Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>