Header bidding continues to grow, with some SSPs reporting an impressive 202% year-on-year growth in Q3 2017. This growth explains why so many tech companies have jumped into the space. In this piece for ExchangeWire, Jonnie Byrne, managing director UK & France, Fyber, looks at the state of header bidding as the new year begins.
As with all markets that experience rapid adoption and growth, confusion abounds. What exactly does header bidding mean? If you plough through the bylines and website content penned by the 70+ providers, and talk to enough publishers about how header bidding is implemented on their pages, it becomes obvious that there’s very little consensus on how it’s defined, much less how it works.
For example, header bidding is often a two-part auction, first between RTB exchanges (which are connected to DSPs), with the winner going on to compete with mediation networks. While this is an improvement over the traditional waterfall, it only takes the concept of ‘parallel bidding’ so far.
In other scenarios, multiple SSPs run auctions independently, each submitting a bid to the publisher's header bidding solution that is $.01 above the second-highest bid. These SSP bids then go on to compete in a first-price bid within the publisher’s ad server. This two-part auction dampens yields for the publisher, since multiple second-price auctions can’t bid against one another.
Mobile app header bidding is also on the rise, which is ironic, given that mobile apps don’t even have a header. How exactly is it engineered? In-app header bidding cobbles together a few server-side integrations with desktop exchanges. Unfortunately, this approach eliminates all of the demand connected through SDKs on the client side, forcing the publisher to leave money on the table.
Clearly the industry needs a standard definition of header bidding, but how do we get there? We could start by defining what the industry aspires to when we talk about header bidding. I think we can all agree we want the highest possible prices for publishers for each and every desktop and mobile impression they offer. That, in turn, requires that all demand sources have the ability to compete fairly for the users they value. This necessitates a truly unified auction, in which mediated networks compete simultaneously with DSPs and trading desks. And, we want greater efficiency; why have multiple auctions that shut out demand and keep end users waiting for their pages to load?
We Need a Revolution in Header Bidding
Achieving our aspirations in header bidding will require a revolution, and that’s a good thing. In general, there is a disturbing amount of distrust in digital ad tech, and the confusion around header bidding only adds to it. With definitions of header bidding that essentially describe imperfect auctions, and no transparency into how header bidding selects its winners, both the buy and sell side may decide it’s better to revert back to the old way of manually signing direct deals. (If this sounds great to you, be careful what you wish for, direct deals are grossly inefficient.)
How do we achieve truly unified auctions in which all buyers, regardless of their buying mechanism, get a look at all impressions? I believe we need to transform header bidding and the traditional mediation from the hack it is today and replace it with a fair and transparent programmatic ecosystem. We need to give mediated networks the ability to compete with programmatic buyers in one flat, transparent and fair auction. That means the mediated networks can have the ability to see and bid for all ad requests, not just the ones that trickle down to their spot in the waterfall, while still keeping their SDK directly on the client side to allow benefits such as rendering specific rich media ad formats.
In the end, a truly unified auction is best for the industry. It will benefit buyers by enabling them to see all impressions and compete, in real-time, for the users they value most, and it won’t force publishers to choose between efficiency, transparency and yield. Bring on the revolution!