Header bidding — also known as pre-bidding, advance bidding, holistic yield management and even “tagless integration” (even though it requires tags to function) — can be a great way for publishers to maximize the sale of unsold inventory left over from guaranteed campaigns and programmatic deals. However, publishers should not expect it to maximize inventory revenue.
Using this advanced programmatic technique, publishers solicit bids from demand sources before ad-server callout directly in the page header. The inventory is offered to multiple exchanges simultaneously, and these are up against direct sales. The idea is simple: The more people to whom you offer your inventory, the more money you’ll make, and the better chance you’ll have of increasing your yield. With the demand side doing everything it can to pay less for impressions, header bidding is the supply side’s attempt to level the playing field.
Header bidding is only as complicated as the settings a publisher wishes to put in place. If it’s “just” the creation of a meta-SSP, it’s simply the creation of a fake automated auction. The parameters taken into account to ensure optimization are limited and the maneuver isn’t extremely complicated. However, the implementation of header bidding, the follow-up and the analysis are all quite complex for a simple open auction mechanism, and it is simply unable to address the preferred deals and guaranteed campaigns challenge.
Most agree that the current waterfall system, in which publishers offer impressions one-by-one down the line, is not only splintered and inefficient, but also wasteful. The loss may be only a penny here and a penny there, but those pennies add up. Some, like John Potter, CTO of Purch, purport that adding just one header bid can drive up yield by 10%. Yield is king, hence the excitement about the possibility of a programmatic revamp thanks to header bidding.
But the technique is not all that it seems.
In certain cases, it can be a good alternative to waterfalling. Header bidding enables a publisher to create a preliminary single unified auction — a sort of meta-SSP — in which all SSPs have access to the same information, and place their bid accordingly. Publishers can instantly sell each impression to the highest bidder, avoiding the latencies of waterfalling. This is why the technique is so popular among ad networks, especially in the U.S.
However, if you’re a publisher selling via direct campaigns at a fixed volume, or preferred deals with first-look access for the buyer, the benefits are less apparent. In these cases, choosing the best buyer is not based solely on the price they’re willing to pay for an impression. There are a host of additional parameters to consider. True yield optimization needs to take into account not only the offered price but also the guaranteed volumes in comparison to the available inventory. In some cases, it may be relevant to deliver a $1 CPM non-guaranteed ad rather than a $10 one. If the $10 ad desperately needs this impression to reach its goal, it’s not an option. Header bidding doesn’t take into account all the parameters, nor does it consider all the campaign distribution constraints. Those holes lead to imperfect optimization.
If third-party SSPs are at the head of the pack, then, according to header bidding, the primary ad server can’t see the whole picture. If that’s the case, it’s impossible to implement truly holistic management.
It’s in this context that header bidding can be a great way for publishers to maximize the sale of unsold inventory left over from guaranteed campaigns and programmatic deals. It’s a great backfilling framework. However, it’s unrealistic for publishers to think they can use the technology to maximize total inventory revenues. Only a truly programmatic (server side impression level) implementation can enable us to reach the mathematical maximum. It would also give publishers impression-level control and holistic reporting. The ad server can do that. What are we waiting for?