Header Bidding: Dangerous For The Evolution Of Programmatic

maorsadraData-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 Maor Sadra, managing director at AppLift.

There’s been a lot of prattle about the potential of header bidding. Some have even called it waterfalling by a different name, and that with some tweaking to header bidding’s implementation it could be a positive development for the industry.

But I don’t believe that header bidding will ever be a positive development. It is not a waterfall – it is Niagara Falls. Header bidding’s adoption will thwart the advancement of programmatic advertising.

The aim of programmatic is to create a more efficient marketplace, one in which prices are set by supply and demand. Header bidding allows publishers to offer inventory on multiple exchanges rather than filtering demand through a single stream. Most DSPs are connected to most of the leading exchanges. If a DSP registers the same inventory from four different exchanges, it is going to bid the same price across all four, potentially increasing its price.

In other words, the DSP is bidding against itself and inflating the ad space’s value. That may make publishers more money, but it also wastes server resources. Furthermore, when advertisers realize prices are going up, they’ll be motivated to cut out the programmatic loop entirely and strike direct deals with publishers. This will essentially set us back to the early 2000s, with advertisers signing and faxing insertion orders to buy impressions in bulk.

One of the benefits of programmatic is that it makes it possible to evaluate each individual impression and lets the market determine its worth, much like the stock exchange. But header bidding tips the scale, allowing publishers to generate unfair prices at the expense of the demand side.

The inflated prices are obviously bad for advertisers, but they are also bad for the entire digital ecosystem. When advertisers opt out of programmatic, they forfeit its improved targeting and efficiency-enhancing capabilities. Publishers may gain a bit in the short term by squeezing more yield from their existing inventory, but in the long run they are hurting themselves.

One of the biggest challenges mobile advertising faces is that there is more supply than demand. Programmatic makes it easier for companies to buy digital advertising, thus expanding the demand side of the ecosystem. It commoditizes media trading via an open exchange and objectively evaluates each impression to determine its true market value.

When publishers force higher prices for their current demand, they are potentially slowing down the progression of programmatic for all of the demand that isn’t there yet – the future buyers who could help even the playing field and address mobile’s impression surplus.

Because the inflated prices that header bidding generates will encourage advertisers to cut out the exchange and strike direct deals with publishers, programmatic’s growth will slow and mobile’s supply-and-demand problem will continue unchecked.

Let’s take a common example that happens daily. A publisher monetizes its app in the open marketplace through a single supply-side platform (SSP). Ad impressions are auctioned by three different demand-side platforms (DSPs). The highest bidder wins and pays on a second-price auction.

But if the publisher includes a header-bidding tag on its unsold inventory, the next time an impression becomes available, it is sent to two SSPs. The first SSP runs an auction across DSPs A, B and C, while the second SSP simultaneously runs an auction across DSPs B, C and D.

In this scenario, which is indicative of a typical header-bidding transaction, DSPs B and C bid on the same potential impression twice, once on the first SSP and again on the second SSP. In doing so, they’re wasting server resources, but worse, they are bidding against themselves.

Let’s pretend DSP A would have been the highest bidder if the impression was only auctioned on the first SSP. Instead, it loses to DSP D and will never know the actual price the auction cleared at since the SSP only has data pertaining to its own platform, not transactions occurring on the second SSP.

If DSP A still wants inventory from the publisher, its best bet is to reach out directly to propose a direct deal in which they bypass the exchange entirely. If the publisher agrees, the entire RTB marketplace misses out on that traffic.

This temporarily solves a problem since the DSP gets its desired impressions, although at a potentially higher rate than necessary. The publisher also successfully sells its impressions but in the long term, it only exasperates the industry’s supply-and-demand issue by removing quality supply from the marketplace.

The downside of header bidding for existing advertisers is blatant. They stand to pay more for the same traffic. The risks for publishers are more subtle, but just as alarming. Header bidding could thwart the growth of programmatic. If advertisers abandon the exchange and start striking direct deals, we forfeit programmatic’s advanced targeting abilities and deep data, as well as its efficiencies.

Programmatic makes it easier to access the digital advertising marketplace. It makes it easier to purchase advertising. Why build barriers? Why risk shrinking the demand-side pool?

Publishers need to remember the long game. We all win by fostering an efficient, transparent advertising ecosystem.

Follow AppLift (@applift) 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. Mike Gartner

    Maor, Your article is misleading and incorrect. A transparent and fair marketplace is the only proven way to achieve efficiency and price discovery. If demand for the inventory is there, then the prices will go up, and that is not a bad thing. Inventory is virtually infinite at this stage in the progrmattic explosion. Think about all the social platforms that haven’t monetized their inventory yet; All the big websites, and publishers that have not even made a dime off their viewship. The facebook exchange is a great example of the bounty of supply that exists. Demand is HUGE for facebook because of the strong performance, but the massive amounts of inventory that can spawn in one second means that FBX CPMs are the lowest around!

    Please read more economic theory before you go spraying nonsense all over the industry.

    Cheers, Mike G

  2. Maor, interesting perspective – I agree header bidding has ramifications for buyers that are worth thinking through, but I’m not sure your point on buyers bidding against themselves is accurate.

    For example, because publishers have no way to price reduce across platforms a single buyer can’t bid against themselves. Rather, header bidding lets the publisher select the highest closing price from a handful of SSPs, but it doesn’t change the auction dynamics within those SSPs. In other words, if you bid $1.20 in SSP A and the auction closes at $0.80 and you bid $2.00 in SSP B and the auction closes at $1.10, then the publisher can certainly select the highest auction you won ($1.10), but they can’t use your bid from SSP A (or any buyer’s bid for that matter) to price reduce your bid in SSP B (which would close the auction at $1.21). In fact, I’m not aware of any header bidding connection that passes the buyer’s bid at all, rather each platform simply passes the closing price.

    You could say that it’s possible for buyers to lose in this situation as well, and price reduction across SSPs would actually be preferable for buyers – imagine another auction in SSP A where the high bid is $3.00, but the auction closes at $1.00, and an auction in SSP B where the high bid is $2.00, but the auction closes at $1.20. In this case, the publisher would deliver the impression to SSP B, even though the highest bidder was in SSP A. Because of the 2nd price auction and bid density dynamics, the buyer in SSP A is effectively penalized.

    I think the key problem to solve in this scenario, and a point you make which I agree with, is that it probably doesn’t make sense for buyers to bid for the same impression through more than one channel. And, in my (admittedly publisher centric) view, header bidding is making the market more liquid and therefore more fair. You might say that buyers have historically been the beneficiaries of illiquid waterfalls, since not all buyers who want to bid were able to bid based on the platform the publisher was working with. Isn’t it possible publisher inventory is simply worth more than buyers thought it was when it’s valued by a liquid market?

    • I agree with the first part of your argument that bidders won’t second price themselves in the prebid auction– but it very possible that the bidder who won the prebid auction will have that clearing price sent into DFP– then, that same bidder will bid via AdX, which will then price reduce to 1 cent above the prebid line item– hence 2nd pricing themselves.

  3. But quality inventory is moving into deals anyway. Publishers want to have direct contact with buyers and if header bidders are supporting this then its even better. And also programmatic deals offers all targeting capabilities that exist in open market.

  4. One of the worst articles on the topic… non-sense… You can have a lot more applications to header-bidding and your arguments against are definitely not the relevant ones…

  5. Why doesn’t the industry just finally move to first price auctions? Would remove most of the financial incentive to pit one platform against another and be much more efficient in terms of delivery.

    Then the only incentive pubs would have to use multiple platforms might be differentiated demand, so that can be the next challenge to tackle!

  6. Can’t the systems be build so that DSPs B and C in your example can recognize that it’s the same inventory from 2 SSPs? Would still use a bit of server resources, but avoid the double bidding.

  7. I think you might be missing the point that DSP’s bidding algorithm does attempt to detect duplicate bids. The scenario presented of duplicate bidding is no different from any open marketplace today because publishers are still selling and integrating with multiple platforms as opposed to 1 single ad server.

  8. Sebastiaan Postma

    Or the buyer sets up a private marketplace based on Private Auctions or Preferred Deals and only bids against those deals having full control, never bidding against itself. Also this is a way to minimize ad fraud as deals can be made with trusted partners guaranteeing a valuable PMP. Sure enough this will drive price up as well, but it is paying for quality rather than possible air inflation. And of course buyers will not switch completely to non programmatic direct deals, as they would give up the ability to control and use data. One of the big drivers behind programmatic.

  9. Great article and thought provoking.

    There may be a different spin on all of this.

    In the early stages of programmatic buyers were in charge. Given there was more supply then demand in display, mobile and other sources buyers had the ability to constantly fill all orders with very low cost inventory.

    Unfortunately, the inventory had a lot of fraud in in and this also drove down prices.

    In the end it the industry started to sink into a cesspool.

    By moving back to direct deals, advertisers will pay more but now they will at least know they are buying real inventory.

    Prices rise, publishers are better off and the industry moves away form a pool of murky inventory.

    In my opinion, header bidding and private deals are the way to make programmatic clean and effective and efficient.

  10. Programmatic is the problem with advertising, not the solution. Doesnt matter what tech is in favor, advertising should not be a per impression business nor one that is transacted by machines in such micro segments. All one needs to see is the avg. efficacy and cost of digital compared to any other medium to see this in play… Its not working!

    But when the industry is driven by engineers looking for a problem to fit their solution, you will see this charade continue. But it will not fix the problems with advertising and specifically with digital.