Header Bidding Killed Programmatic Direct

jed-nahum"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, an independent consultant and former head of Microsoft’s global programmatic sales team.

Every year in digital advertising, we seem to have a new issue or two bubbling up and occupying our collective consciousness. For 2015, I think of ad blocking, viewability and header bidding.

Back in 2013 and 2014, there was a lot of talk about programmatic direct, a method by which a publisher ad server – or really any ad server – exposed APIs in order for buyers to upload campaigns and line items.

We in digital get more and more items on our plate every year so let’s agree to stick a fork in this one and call it done.

Back in the day, there was a lot of buzz about programmatic direct. Remember when Microsoft, AOL and Yahoo agreed on an API standard? And even a year ago it was in the news when Rubicon bought iSocket and ShinyAds. PubMatic exposed a similar API, too. This was a big deal.

And what has transpired since then? A lot of cricket song.

Why We Thought Programmatic Direct Would Be Helpful

Programmatic direct is mostly about API access to the publisher’s ad server offerings. Because publishers are well positioned to forecast, their ad servers are good at offering forward contracts, essentially the reserving and guaranteeing of inventory. Since we’d been doing guarantees in our business since the 1990s, we thought it was probably important. We need that functionality, right?

Media planners enjoy the simplicity and lowered risk of guarantees. Programmatic direct also enables automation that should lower the overhead associated with insertion orders, RFPs, negotiation and the like. Beyond guarantees, marketers get access to the publisher’s context and segments. They can push them to marketers in an allocated fashion, giving marketers nice stuff like smooth, predictable delivery.

But maybe the most valuable attribute of programmatic direct is the ability to offer a first look at inventory. When I was selling, the No. 1 thing partners asked me for was first look. Until header bidding came along, first look was difficult for programmatic.

RTB Pushes Power And Control Toward The Buy Side

I’ve argued in other articles that RTB has created a war over whether the buy side or the sell side defines what is being sold. The nutshell of my argument is that RTB has made it possible for buyers to select specific impressions for their buying. This is “decisioning.” In RTB, the buyer does it, and in traditional ad sales the seller does it. I think the real battle is about who gets to decide. Will buyers define the audience they want to reach and bid on it or will sellers define the context they wants to sell and package it up?

When buyers decide, every buy potentially delivers data to their data management platform (DMP), enriching their knowledge of their audience and making them better marketers. The problem with programmatic direct is that it doesn’t incorporate the DMP in the same way RTB bidding does. And since buyers control the spend, they’re looking for ways to feed the data beast.

I learned this talking to buyers. They told me that the real value of programmatic to them was being able to buy only the stuff they wanted and none of the stuff they didn’t. With programmatic, there are two kinds of efficiencies at play: automation and selection. The word “programmatic” makes it sound like automation is the big deal, but I argue impression selection is far more important.

Hence, demand-side platforms (DSPs) and agencies realize they might not like programmatic direct because it takes decisioning away from them. They aren’t interested in being “order takers”; they want to be order makers. When DSP vendors talk about “powerful bidding algorithms” or “expressive bidding,” they’re pitching their ability to engage in good decisioning. Similarly, agencies and media services companies want to manage that decision-making to provide more value for their clients. They don’t want to limit their influence by putting control back into the hands of the sell side with programmatic direct.

And now there’s this thing called header bidding. It enables a publisher to offer first look in a bidded RTB environment by moving supply-side platforms (SSPs) up in the delivery priority of their ad server (it’s more than that, I know). Because bid requests are still being made and users are still being mapped, the data beast gets fed. Though header bidding has some pitfalls for the buy side, it’s much preferable to the programmatic direct alternative.

Where Programmatic Direct Still Fits In

All that said, there are still a few places where programmatic direct may prove valuable. For example, environments where user identification is difficult, such as mobile, can stand to benefit from programmatic direct’s efficiency of automation. Further, when scarcity drives up the price of a reservation (video) and pushes power into the hands of the sell side, programmatic direct might fit in. And seriously concentrated publishers, including Facebook or YouTube, can do whatever they like and may use APIs to sell their stuff.

I’m not saying slapping an API on an ad server or a DSP is a bad thing; in fact, it can be quite helpful. But I am saying it’s not as revolutionary as enabling buyers to define audience and bid against it. Buyers can take note when publishers expose APIs. They can decide case by case whether to integrate. But the programmatic revolution is being driven by audience, not automation.

Follow Jed Nahum (@jednahum) and AdExchanger (@adexchanger) on Twitter.


  1. Robert W.

    I'm not sure you fully understand what Programmatic Direct offers vs. Header Bidding. With programmatic direct, buyers can automate the buying of all types of publisher ad inventory (IAB units, high impact/custom sponsorships, etc). Some DSPs are able to offer brands the ability to utilize audience data to cherry pick which of those audiences they want on a direct guaranteed buy (typically called Programmatic Guaranteed or Guaranteed Plus). Header Bidding is not moving up the priority in the ad server, it's being able to respond to the impression before it even hits the ad server. It's code placed in the header of the publisher home page. Programmatic Direct, as with Header Bidding, will take time for the market to adpot. The only thing keeping Programmatic Direct from growing faster is publisher adoption, which has been slow due in part to the impact it may have on a publishers' ad sales business and structure. Header bidding is a publishers way to increase yield and, frankly, poking DCM/Google in the eye.

    • Jed Nahum

      Robert thanks for your comment. It's true that programmatic direct offers more ad sizes and more features than RTB/Header bidding. My point isn't about what offerings can potentially be offered through PD vs Header Bidding so much as what offerings buyers care about. I think they are mostly interested in being on an equal or improved footing relative to other buyers, including a first party sales force. I see Programmatic Direct as a sell side response to a buy side development, and therefore less in tune with buyer needs. BTW, correct in that I misspoke WRT the technical implementation of header bidding in the ad server. You do wind up being able to prioritize SSPs against your internal demand. Ben Kneen explains how this works nicely over at adopsinsider.com: http://www.adopsinsider.com/header-bidding/header-bidding-holistic-ad-serving/

  2. Eric Picard

    Jed - this is a great article, and it gets right at the heart of the ongoing and evolving issues in the programmatic space. I have often referred to programmatic direct as a bridge between the 'old' world of direct buys and the 'new' world of RTB. And you've given me a great idea for an article I'm going to write in response to yours - so I'll keep my comments to a minimum. BUT...

    Header bidding is the "rest of the industry"'s response to Google's offerings where enhanced dynamic allocation enables publishers to apply the demand from the exchange directly against the demand from direct buys. Both are solving the same problem, but much to Google's chagrin header bidding in many ways is 'winning'. By this I mean that header bidding enables publishers to operate with immense control, and to give buyers the other thing they greatly desire (besides as you described, being able to define the inventory in the buy), which is privilege.

    When a buyer asks for a 'first look' they're asking for privilege over their competition. This is not a new thing - every marketplace in the history of marketplaces offers various forms of privilege (non-price based differentiation). The simplest is that big spenders want discounts and access to the best stuff - so if you're a buyer spending $100M a year with a publisher, you want privileged access to the best stuff, regardless of what you're paying on a per-impression basis.

    To counter Robert's comment above - header bidding is sure as hell about moving up in the priority stack - even though header tag gets access to the page and impression information before the ad server - the header tag drops the demand *into* the ad server where it is assigned against a standard line item. That line item can be assigned anywhere - at any priority level, so yes, priority still matters a ton. You might imagine that publishers are doing numerous header deals and setting the various line items from those deals at various priority levels. That's specifically the kind of control ad ops teams and sales organizations desperately want.

    Oh - and btw - I can tell you from experience that DSPs are getting access to every imaginable type of inventory - everything from home page takeovers to every non-standard and even native ad unit on a publisher's site. Programmatic Direct had some valuable benefits when it was first brought to market - but there's no value add any longer. I agree with Jed's assessment - the market has spoken - and the strong value proposition many of the P-Direct guys had behind them has evaporated. Mainly because publishers used it as a clearinghouse for inventory that wasn't selling - and to be clear, not remnant. Most publishers just dropped all the inventory they couldn't sell at a rate card price through to the PD provider. So inventory that nobody wanted at prices nobody would pay. That's not a slam on the programmatic direct guys - it's sad that publishers had no real vision for how to use the channel.

    • Jed Nahum

      Thanks Eric, I think we're pretty well aligned here. I agree that priority and first look are the same thing. I'll look forward to your ideas when you publish your thoughts on this matter.

    • Jed Nahum

      Michael: I wouldn't say it's more valuable, I would say that programmatic direct is less of an inhibitor. Because user identification is complicated in mobile and the sellers may know more about audience than buyers, often in bidding environments the user id either isn't transmitted or isn't mapped effectively. This gives the buy side less assistance in amassing buy side data stores. Therefore, I'm positing that PD doesn't suffer some of the setbacks in mobile that it does in other channels.

  3. Brian Burdick

    In general, I agree that the purchasing decisions have moved to real-time buyer driven including 'Header Bidding'. That said, there are a few places where programmatic direct could become much more important:

    The first area is in buying programmatic traditional media - such programmatic TV, Radio, Print, etc. The reason for this is the thing that is delivered is a bundle of impressions that can not be individually served and the audience can be estimated a priori with panel or set top box data. In these instances, there is no 'header bidding' yet and the programmatic efforts are competing with traditional sales methods that lock inventory up well in advance. In order for Programmatic efforts to out-monetize the traditional sales it will likely require a form of Programmatic Direct that enables buy-side targeting but still works with seller guarantees (in order to compete with the yield provided by traditional guarantees before they are committed).

    The second reason it may have a resurgence in digital is privacy. With the EU and other governmental bodies increasingly scrutinizing consumer 'informed consent' and 'protecting users from excessive personal data-tracking' things could go where identity can not easily be mapped for buyers on an individualized basis at all. If this happens, much of the advantage of header bidding will dissipate and things like Programmatic Direct may become more important again assuming there are appropriate 3rd party targeting escrow companies that manage the buyer still gets to use complex targeting attribute rules of the consumers in aggregate they want to reach. If this happens we will be back to audience profiling pages and placements - just at a much more granular level than before. The situation that would cause this would be if the publisher could use the individual consumer data to ad decision but could not share it with any 3rd parties due to privacy regs. I did see today where some experts think the Facebook 'Like' button infringes on the EU 'informed consent' privacy law - and that is a first party publisher collecting data.

    The third reason is media strategy and has to do with inventory scarcity. If it is possible to share an anonymized list of consumer identity that an advertiser would like to reach - then they could put in a Programmatic Direct order for those consumers. If this list happened to be a highly overlapping list of consumers needed by the advertiser's competitor then it may be possible to starve the advertiser's competition of impressions of a small but important audience segment. There are only so many people with income over $500k for example. Blocking seasonal peaks where the advertising elasticity of demand is stronger could be another example where it would be worth it and the publisher may demand it. The reason for this is simple - if the programmatic direct order could be of higher priority than header bidding then this would be possible. The reason the publisher would like this is because they could collect the money up front to enable this (yes a change in payment structure but it would be motivational to the sellers to enable the invevtory to be committed) using their existing inventory prediction on the advertiser's custom segments which will spread the publisher's revenue more evenly across quarters and make it more predictable. This obviously makes cross-site frequency management more difficult for the buy side but might be worth it in some scenarios.

  4. Jed Nahum

    Brian: there’s a lot to respond to in here. First, I appreciate and agree with your additions to my list of places where PD may still fit in. Your point regarding privacy and whether it evolves to match the facility for identity mapping that RTB and Header Bidding have is also true. The privacy part is a hairball, and bigger by far than the issue of Header Bidding vs. PD. Finally, hard to argue with the scarcity argument increasing the value (to buyers at least) of a reservation. I believe forward contracts/reservations will become more important again if the balance of supply and demand normalizes a bit. And there are places where they are in greater equilibrium (video being one example I think of, as above). However, I wonder if forward contracts will exist harmoniously inside a unifying auction.
    Imagine instead of the header tag dropping into a line item in the ad server (as Eric explains above), we have a scenario in which a reserved contract is converted into bids continuously during delivery. An intelligent ad server could actually adjust its bidding up when it starts to slip out of SLA for delivery. And of course, adjust downward when supply meeting the reserve criteria is abundant. This would enable anyone with good forecasting (not just the publisher) to issue reservations. Very good forecasting would lead to more arbitrage profit in the issuance of the contract, and bad forecasting could be very very expensive. This would also have the effect of highlighting that Programmatic Direct is not a system for reserving inventory so much as the API that enables you to access the campaign entry of any given ad server.

  5. Ian Trider

    Very interesting commentary. I've long wondered, in a world of huge numbers of display ad impressions, why buy anything on a guaranteed basis at all? Certainly moving to a non-guaranteed basis allows for smarter buying per the reasons you note, and also actually enabling proper universal frequency capping. In theory, this should be good for the publisher as well as they are not sacrificing any revenue by making suboptimal pricing decisions due to a lack of ability to know who is willing to buy a given impression for the most at a given moment.

    I do think that advertisers and agencies need to change how they think and operate for such a world to thrive, however. The desire to be able to guarantee that a budget will be spent (after all, you don't get paid if you don't spend the budget), as well as existing workflows built around a guaranteed buying world (where there is a "placement" that is planned for X impressions at Y CPM for a total cost of Z dollars) do create drag that keeps a lot of buying happening on a guaranteed basis for now.

    It's not impossible -- search has of course always operated in this way. It's just a lot of mindsets and workflows to change.

    • Jed Nahum

      Ian, good thoughts. There is no doubt, as you posit, that an excess of supply vs. demand drives down the value of the guarantee on the buy side as you have implied. It also drives up the value of selling in advance for the sell side :). And I agree, frequency is a key variable for the buy side to manage.

  6. Neal Richter

    Good piece. Buy side decisioning via RTB will continue to revolutionize.

    Programmatic Direct is two things in my opinion. An automated marketplace for buyers and sellers, and a protocol. The protocol never really stood a chance versus RTB, long term. The concept of a marketplace of packaged premium inventory (and data) is the part of PD that can survive.

    Header bidding is yet another tactical mechanism to get all in inventory in an exchange to be sold via RTB.

    But the larger point is not that header bidding killed programmatic direct. It's that RTB and dats enabled buy side decisioning will kill the ad server.

  7. I don't understand the argument about programmatic direct taking the 'decisioning' away from buyers/agenices...they're still controlling where, how often, when, and at what price the impressions occur for their client aren't they? Would you mind clarifying if you have a sec, I'm interested in understanding this better.

    Thank you,


  8. To Publishers:

    Use all Broker & re-brokers feeds and then NET each feed for highest ROI (pretty much useless or limited at best) or bid amounts. We invented RTB in 2000.
    This is not new technology.

    Publishers need to learn how to "Control their own traffic."

    Current market traffic delivery systems are ridiculous -built off of OLD technology.. Just look what happened to the stock market.

    See supporting articles here if you like:

    Kind Regards,


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=""> <s> <strike> <strong>