What We Talk About When We Talk About Programmatic

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.

A friend of mine once lamented, “As an industry, we ruin words.”

It’s true.

A few years ago, we all conspired to ruin the word “exchange” with ad network models that diluted its meaning. Now we’re in the process of destroying the meaning of “programmatic.”

Let’s face it. We’re talking about different things when we say the word programmatic. I’ve attended several good forums on this topic and was alarmed to realize that people who think they’re talking about the same thing when discussing programmatic are actually worlds apart.

I want to create an inventory of these meanings as best I can. But first, let me say that “programmatic” alone simply refers to ad offerings that can be automated through technology. It’s broad, and will mean more than just display.

A Proposed Taxonomy Of Programmatic Offerings

Under the overall umbrella of “programmatic,” I consider there to be four main offerings: RTB, Deal ID/private marketplaces, programmatic direct and another offering that some are discussing which, for lack of a better name, I’ll call programmatic forward. Here are some quick definitions:

Programmatic offerings


Deal ID and private marketplaces

Programmatic direct sold

Programmatic forward market

RTB has come to refer to the spot market for ad impressions that is enabled by the real-time bidding protocol. Features on top of RTB which enable sellers to designate privileged or specific buyers Refers to the efforts to address automation of existing forward buys (guaranteed or reserved) sold around the industry. Refers to the as yet undefined forward and futures markets for impressions.


Notice that I am not using the phrase “programmatic premium.” This phrase is simply too ambiguous and translates roughly to “automated good offering.” I can’t really define it and therefore recommend avoiding it. Since I won’t be addressing them, I’m omitting all reference to the manual lines of business that can be either direct or indirect, as well as reserved and remnant.


Technically, the RTB protocol is distinct from the marketplace it has spawned. But since this market remains unnamed, we have unwittingly conspired to call it simply “RTB.” We should probably call it “programmatic spot market.” I won’t go into much detail here, since presumably if you’re reading AdExchanger, you know something about RTB. However, I will point out the following:

  • This may be obvious from the last paragraph, but what we call RTB is not just a protocol but also a marketplace.
  • RTB has improved liquidity for publishers and advertisers alike in the spot market.
  • From the perspective of the buyer, there is fungibility of impressions and audiences across marketplaces. In other words, the stuff you buy from one RTB marketplace can more or less be replaced by the stuff you buy from other RTB marketplaces. Significantly, the price of impressions will vary against quality similarly across marketplaces.
  • RTB, while capable as a protocol of delivering content adjacent advertising, has been used mostly to enable buyer defined audiences across a broad range of content.

I think much of the time when people talk about “programmatic,” they’re really just thinking about RTB. I don’t think that’s specific enough.

Deal ID And Private Marketplaces

Deal ID, and to some extent, private marketplaces are efforts to automate existing direct buys. They allow a fair approximation of hand sold direct using RTB plumbing. One advantage of these solutions is that they enable buyers to utilize the same buying systems for both RTB and direct deals, and therefore concentrate data in one buying system and DMP. Another big advantage is that these offerings are available now and everything else below is more or less on the horizon. The disadvantages are that they aren’t yet effective for guaranteeing placement and it can be difficult to replicate ad server placement granularity in the RTB environment. Further, as currently defined, these buy types create significant overhead for publishers that limit their scalability. I suspect these are gap-bridging offerings that are neither fish nor fowl and won’t be actively sold in the long-term. The technology that enables them, however, may be repurposed for other goals.

Programmatic Direct

Direct or hand sold advertising exists in both remnant (spot) and reserved (forward) forms. Programmatic direct, on the other hand, refers to the automated versions of these advertising offerings.

The emerging solution set consists of API access into ad servers with automation against RFPs, I/Os and creative management. This product is coming together as you read this with publishers, SSPs and third-party vendors, such as ShinyAds and iSocket, adding the requisite feature set. Another good example is the recent feature additions touted by Rubicon for its Revv platform. There will almost certainly be multiple acceptable solutions that will sit in this space. Programmatic direct will offer efficiencies of process, such as removing Non Value Added steps and redundant activities in the order process (repeated input of the same data for I/Os). It will shorten cycle lengths and “appify” the selling process, if you will. It doesn’t offer increases in liquidity and price discovery. On the other hand, I’m pretty sure this offering will, over time, entirely supplant our current way of selling brand on display since it should wind up being a more efficient substitute. Most excitingly, it will be the means by which we sell native and very customized advertising.

Programmatic Forward Market

Programmatic forward will be the forward market that attaches to our current spot market. Its acolytes tend to advocate for liquidity and price discovery, indicating that pricing information must flow smoothly between the spot and forward markets. Such a market would be real-time from the perspective of order placement but would deliver impressions in the future on agreed-upon dates. This marketplace/offering would supplement the direct and programmatic direct offerings. In fact, while programmatic direct looks like a feature set to me, programmatic forward looks like a marketplace. As it requires integration with the existing spot market, it would mostly involve contracts getting written that are then delivered through systems built on top of the RTB protocol. This would allow the spot and forward marketplaces to trade pricing information and balance demand.

Envisioning this marketplace is fun, but defining it is a hairball. In order to sell forward contracts, the writers of those contracts must be able to forecast availability. Contract writers would forecast the availability of the item being sold, including audience or content placements, and later find the audience or placements necessary to fulfill the order in the programmatic spot market. This is analogous to the way contracts for commodities are written.

Since programmatic forward will support and require automation, it should see many of the process efficiency benefits I mentioned above for programmatic direct. However, it will additionally bring greater price discovery and liquidity, and should eventually make us better buyers and sellers. It will be very well suited to deliver standardized forms of advertising, but will have difficulty supporting customized and native executions. The pricing information that it and the associated spot market delivers will be invaluable for direct offerings.

Let’s Be Specific

My goal here has been to offer terminology and definitions that allow us to be specific enough to avoid confusion when we’re talking about “programmatic.” Below, I’ve distinguished between major types of offerings and added fairly standard names.

RTB / programmatic spot

Deal ID / private marketplaces



Programmatic forward











I’m not inclined to argue about the relative value of each of these ad offerings. The future is big and will incorporate all of them. Each offering will each provide unique value for sellers and marketers.

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

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. What about Facebook, LinkedIn, PPC etc? We’re choosing to include these under the programmatic umbrella as well.

  2. This is a great article Jed! Thanks for bringing more clarity to the conversation.

    I think you have started to uncover something here that isn’t clear to many people, specifically related to programmatic buying and selling and the pipes. I’m going to think more about this, and since I have a deadline this week – I’ll perhaps write a longer follow up.

    But – one thing that’s really interesting in your way of laying out the market is that each is its own island, which makes sense given that vendors today are each attacking the problem as if they are islands. But I see this as a continuum, and ultimately all of them should be solved through one toolset. The question is – one toolset that binds multiple vendors together, or one toolset that replaces existing vendors? I think the former is more likely to succeed. Of course, that’s what can be done using Rare Crowds’ open source framework. 😉

  3. John:

    Good questions. WRT FBX, I would simply throw that in the RTB spot market. WRT the other stuff you mention, I would add in Search to it and ask the question back at you since I haven’t thought that part through well enough (i.e., not sure about publisher hosted auction systems for native ad placements).


    Thanks for the kind words! I’ll look forward to your further thoughts. WRT your premise, I agree that there’s opportunity for companies to integrate the various offerings and that there can be tremendous value derived that way. For example, the APIs of Programmatic Direct are really just a layer in the stack. You can imagine buying system integration into the Programmatic Forward market as well. Another integration dimension might be multichannel buying systems which will share data across channels.

  4. Andrew Casale

    Great piece!

    How far away do you see us from a programmatic forward market protocol that is supported by buy/sell side? With DealID now well installed across the ecosystem, this remains one of the last key pieces to bridging the gap that still exists between what can be done programmatically, and what can be done with paper.

    • Thanks Andrew. I’m not sure how far out we are on the forward market, but I’m optimistic that we’ll see it take place in the next 24 months. I’d like to clarify that I see Deal ID not as the forward market but rather as a tool that might be used to lock inventory to support the contracts that get written for the forward market. I know you get that, but the issue is so confusing I thought I’d clarify.

      • I agree with Jed (although I’ll be perceived as having a bias) – the Programmatic Direct technologies are closer to ideal of a futures market than the DealID / RTB pipes.

        Jed, would you agree that a programmatic futures market is just a programmatic direct deal (direct, reserved, etc.) that takes place pretty far in the future and can be traded to others?

        FWIW, buyers are already using isocket (and I presume the most similar platforms) to do reserved deals up to a year in advance. But we haven’t opened up formal reselling of those reserves from buyer to buyer.

    • Mike Davis

      The protocols for the forward markets are already being defined. It will certainly be awhile before there is live, actionable forward contracts being made through them (the next 24 months sounds about right), but this is how advertisers want to buy media and what will drive the highest demand for top publishers.

      It seems to me that the road to getting to forward markets goes through all the pieces that Jed mentions here. We have the spot market and private marketplaces. The next step is automation of deal discovery and deal negotiation (sort of a next step for private markets). This removes much of the friction that is part of current private markets and direct deals. From there we have to begin to see the merging of the spot market and the private & direct markets that enables better price discovery as well as audience discovery. That happens with a significant portion of all inventory being made available via the spot market. That merger allows both sell side and buy side to be able to get better at forecasting which is key to being able taking on the risk that is inherent in forward markets.

      There’s a lot to be done here for sure, but there are economic forces that are already pushing the industry in that direction and it is going to happen very quickly once the pieces start to fall into place.

      Great article Jed!

    • Yes. Pretty sure he just rolled over in his grave. But then, “What We Talk About When We Talk About [Noun]” is already a meme. I’m just the latest to use it.

  5. John Ramey:

    You are bringing up futures which kind of takes things to another level versus forward contracts. I agree that in the futures market, the contracts would be resalable. In the programmatic forward vision I shared, I didn’t speak to the ability to resell contracts and I don’t consider it required there. Making contracts available for resale, in either of programmatic forward or programmatic direct across some standard market then starts the engine of speculation. Speculation is not inherently bad, in my view. But reselling contracts introduces extreme complexity in advertising due to the issues like creative acceptance, blacklisting and whitelisting, and other stuff I’m sure I haven’t thought of.

    I don’t recall saying that programmatic direct is closer to ideal, though, versus the programmatic forward market as I defined it. I don’t opine because I don’t honestly know that answer. Maybe this is a useful clarification: the APIs that publishers use for connection to their ad servers do not create a market. However, companies like yours, which effectively federate access, can become a marketplace for forward deals.

    Mike Davis:

    I think you’re right and I can’t imagine putting it any better than you did above. Thanks for your kind words as well.

    To you both: it’s really fun to think about and discuss this stuff. My main point was that we should always be clear about what we’re discussing. I don’t think anybody on this comments thread is “worlds apart,” so I hope that the terminology and taxonomy I’ve proposed are workable. I’d love to hear comments on that as well. 🙂

  6. Per John Ramey’s comments: There are many ways to skin various cats in this space. The “RTB Pipes” are a fine way to do just about any of the things done through the “Non-RTB Pipes”. Literally anything done inside DFP could be done in AppNexus (as an example.) And without much difficulty, I might add. It’s just a question of bang for the buck. I predict that inventory prediction services across RTB (and therefore the ability to have forecasting that can be used either to make decisions, or even to sign “forward contracts” as Jed suggests) is less than a year out. At that point, what’s the difference?

    The good news is that “programmatic direct” players can act across all these platforms. It isn’t “either or” it’s AND. 😉

  7. Saikrishna Chavali

    To Jed and to the others active in these comments,

    How will the possible block to third-party cookies affect the programmatic space?

    As I understand it, companies like AppNexus rely on matching their cookies (cookie ids i guess) with their partners’ cookies to track users and provide value to their clients.

    If third party ad networks cant rely on cookies, won’t above companies lose some chunk of their cookie-matching ability?

    P.S. Highly likely my understanding is partial, but comes from Eric Picard’s somewhat recent article https://adexchanger.com/data-driven-thinking/we-dont-need-no-stinkin-third-party-cookies/

    • I think that’s a bit off topic as pretty much all of these offerings can operate without cookie matching at all (including RTB). Programmatic Direct, for example, is no more or less impacted that any other publisher ad server. Further, a careful read of Eric’s article indicates that there’s a way forward for audience based offerings through the use of subdomain mapping, DMPs, and first party cookies.

    • Andrew Casale

      The act of buying and selling media through programmatic means transcends the availability of a cookie. While yes, cookie syncing is very much a part of the RTB protocol, and is an incredibly important element when we think of the value it brings to the table, the protocol is not dependent on it to survive. Impressions are already traded programmatically today on browsers that lack 3rd party cookie support – Safari being the largest example.

  8. Brian Burdick

    Great article Jed! It is fantastic to reduce many of the redundant, manual steps in premium sales. Reservations/guarantees still somewhat problematic when the most valuable audiences to the advertisers are the ones they won’t share the information about to publishers. Perhaps more detailed audience standards will emerge and act as a currency. Age and gender won’t cut it from a big research company but a free for all of data providers is also difficult to forecast / reserve inventory for any publisher. How do you see audience definition playing into this?

  9. So, listen. Everyone claims they invented RTB. I don’t claim I invented RTB. I think Brian Burdick invented it and I have access to the patents he filed on behalf of MSFT in 2005 (!) for Open Listing Exchange that back up my claim. So, that’s background for this question and my response. If anyone reads this far in the comments thread they may take exception to my assertion, but what I’m saying is: I’m humbled. 🙂

    Brian, your question opens up a can full of a bunch of worms. I agree with the concern but maybe not the analysis. First, you imply that the guarantee needs to be made (contract written) by the publisher in a buyer defined audience scenario (if I’m reading you right). There are others that can do this, and, perhaps alarmingly, there’s a way to do it that doesn’t actually require standards.

    Second, WRT Age and Gender, this is very interesting and is maybe a gateway drug for Brand campaigns, if we can engineer futures/forward contracts around this that buyers a) can understand and b) accept.

    I believe there’s a way to do both these things, so let’s get together and talk about it over coffee.