Nothing stays static for long in digital advertising.
The use of header bidding, a method that lets buyers bid simultaneously in auctions, has been popular among publishers, for its ability to drive up programmatic revenue yields. But it can only scale so far, since the more demand partners a publisher plugs in, the greater the risk of page latencies. That’s why server-to-server integration is gaining steam. Publishers can offload the burden of the auction from their own headers onto the third-party server and therefore scale it.
That said, server-to-server is still fledgling and has its own drawbacks. One big remaining uncertainty is whether user matching is weaker with server-to-server, which can reduce the value of an impression. And that’s a worry for buyers and sellers alike. Here’s a look at the current state of play.
Publishers get more control
In theory, pushing what was previously the header-bidding auction from client-side (publisher) to server side puts the publisher more firmly in the driving seat, when it comes to control (and autonomy) of their inventory. Via server-side integration, they can have a direct view of the bidding process, across multiple exchanges. Previously, header-bidding vendors used proprietary tech, which is why they initially resisted collaborating on joining each other’s wrapper. But publisher demand for a more transparent view has prompted more collaboration among suppliers.
The Guardian is a fan of server-to-server’s potential. “Server-to-server technology will deliver huge benefit for publishers if it allows them to execute a true, unified second-price auction which they control, as this enhances transparency for buyer and seller,” said Danny Spears, programmatic director at the publisher.
Trinity Mirror’s programmatic chief Amir Malik is adamant server-to-server brings back more control to publishers, who have been on the back foot since the rise of programmatic advertising. He believes Trinity Mirror’s own integration of server-to-server across 20 percent of its inventory gives it vital control of the auction and its own media, and disrupts the typical chain of command in the digital ad supply chain. The publisher itself can now manipulate bids depending on who it has commercial partnerships with, and it can see who, how and when buyers are bidding, and has found ways to package up that capability to take to advertisers directly.
By using ad tech platform Switch, which doesn’t bid itself on inventory, Trinity Mirror can also use the same methods off its own sites, via its trading desk which currently grosses £500,000 ($624,000) a month, according to the publisher. In theory, any publisher with a trading desk can do the same.
Although Amazon has quietly been supplying header-bidding tools to publishers for the last four years, its place in the ad tech space will likely mushroom this year, thanks to its own cloud-based server-to-server play. It is already testing this with U.S. publishers, and the wealth of customer-intent data the e-commerce giant has, across multiple devices via a persistent ID, gives it a very interesting selling point for publishers. That, and the fact it pretty much powers the web, allows it to provide lightning-quick ad calls. That’s likely to be hugely popular among publishers concerned about page latencies (which is all of them.)
One of the drawbacks for publishers trying to scale header bidding was the politics among the vendors themselves. Typically no ad tech vendor wanted to be in a rival’s wrapper tag, for competitive reasons. With server-to-server, that friction seems to be dissipating, among the bigger players at least: AppNexus and Index Exchange have been quick off the mark to collaborate with a joined-up server-to-server offering. “We heard directly from the publishing community that they wanted open, transparent, unbiased mechanisms that would help them work with more partners,” said Tom Shields, AppNexus’ chief strategy officer.
Digiday Daily Newsletter
“Being able to see what’s going on and not having a back box is of huge benefit. With the proprietary software, you’re never quite sure who won the bid, for example, because it’s happening across different partners,” said Shields. “[Server-to-server] ensures you’re being treated fairly and getting good yields,” he added.
As much as everyone criticized the waterfall approach as wasteful, the different positions SSPs could get in the waterfall would differentiate them to buyers. “If a buyer wants first look, early session depth or higher audience match rate, ordinarily you would select the SSP with the best position in the publisher’s waterfall,” said independent ad tech consultant Paul Gubbins.
As header bidding came into favor, publishers adopted wrappers and header tags, and removed tags from their pages. “Many on the supply side that used to leverage their position in the waterfall as a unique-selling point have the same position and priority via their header tag as their competitors,” said Gubbins.
This is bad for SSPs. And confusing for the buy side. “It can be difficult for the buy side to establish who they should be pointing their demand at if four SSPs all have access to the same publisher and at the same priority via their header-tag implementations.” added Gubbins.
SSPs with weaker demand
Any SSP that joins the server-to-server party, however, won’t have their headaches disappear overnight. Server-to-server isn’t going to solve the commoditization issue in the long run because SSPs will still have simultaneous access on bids.
Sources say differentiators for those on the sell side will now increasingly be things like strength of their demand, commercials, ability to support publisher private marketplaces or programmatic guaranteed deals, service layer and analytics.
“Historically, SSPs serviced the sell side, but we’ll see a lot more investment in their demand development teams this year as the race to capture demand as their proxy for differentiation heats up and their service layer for publishers is decreased as publishers migrate from managed to self-service yield management tools,” added Gubbins.