“The Sell Sider” is a column written by the sell side of the digital media community.
Today's column is written by Chris Bell, senior director of product, inventory, at TubeMogul.
Dr. Frankenstein meant well. We all know the story: The young doctor discovers a secret technique to animate the lifeless. However, the creature built by Frankenstein bears no resemblance to his original vision; horrified, the doctor shuns the creature, which in turn torments Frankenstein and ultimately ruins the man’s life.
Header bidding could very well be the advertising world’s newest such creature. Despite header bidding’s theoretical benefits, in practice it is creating unexpected dynamics with implications for advertisers, publishers and the technology companies in between.
Header bidding is still primarily a publisher phenomenon. The technology allows publishers to make their available inventory visible to all potential buyers simultaneously rather than the traditional waterfall, where inventory is visible on a prioritized basis previously established by the publisher.
By removing this pre-existing prioritization and allowing all the demand sources to compete on equal footing, header bidding maximizes the value of each individual impression. Subsequently, advertisers gain visibility and opportunity they otherwise would not have had, making each impression more fairly priced.
That’s the theory, at least. But the reality, like Frankenstein’s creature, is much more complicated.
A serious and unexpected dynamic of header bidding relates to auction mechanics, which are currently structured between the publisher’s ad server and the supply-side platform (SSP). Real-time bidding uses a second-price auction structure, where winning buyers do not pay their full bid price – they pay one penny more than the second-highest bid. Critically, the SSP submits the bid to the ad server – the ad server itself does not support auctions and will pick the highest price submitted.
With this in mind, imagine a header tag where five advertisers are competing for the same impression. Advertiser A has a direct deal at a fixed rate of $19; Advertiser B bids $20 and Advertiser C bids $18 through SSP No. 1; and Advertiser D bids $19.50 and Advertiser E bids $18 though SSP No. 2. Because the SSPs run second-price auctions, SSP No. 1 submits B’s bid at $18.01 and SSP No. 2 submits D’s bid at $18.01. Advertiser A wins the impression at $19.
The publisher entirely misses out on $0.51, and the fact that two advertisers are willing to pay more than $19 undercuts the very benefits header bidding provides. Likewise, there was nothing advertisers B and D could do to win that impression.
Some companies are already trying to combat this by shifting to a first-price auction model, where SSPs submit the highest bid price. But this is an incomplete solution. Even though Advertiser B would now win with a $20 bid, it did so inefficiently – by paying 49 cents more than it should have. First-price auctions drive buyers to manipulate bidding strategies, which is why RTB uses a second-price structure.
This example is not uncommon and clearly illustrates the need for closer collaboration between SSPs and publisher ad servers. Ideally, the SSPs would submit a first-price bids to the publisher ad server, which the ad server would evaluate against all other demand in a second-price auction. This mechanism would allow all demand to be evaluated fairly and result in the publisher earning $19.51 from Advertiser B.
This need for sell-side coordination is likely one of the reasons why Comcast’s ad server FreeWheel purchased SSP StickyAds – FreeWheel recognized it needed to bolster its own bidding technology. There is a healthy market for SSPs and I would not be surprised to see more deals and partnerships on the horizon.
So how can we make sure that header bidding doesn’t end up like Frankenstein’s monster? Both publishers and marketers should demand the ability to truly value each impression individually, without the unintended consequences of an incomplete solution, and accelerate the collaboration between publisher ad servers and SSPs.
In the long run, header bidding should lead to fewer fixed-price deals. In turn, advertisers may leverage their buying power to negotiate rebates with publishers. With fewer incentives to deal directly with publishers, advertisers will increasingly depend on buying platforms to streamline access to premium inventory. These dynamics will, in turn, require adaptations to the interactions throughout the ecosystem.
Regardless, the switch has been flipped and header bidding is here to stay. It is up to us to ensure that it lives up to its potential. As the poor Dr. Frankenstein would surely attest, good intentions alone are not enough.