Why server-to-server won’t displace header bidding

“I think everyone involved is still trying to feel out server-to-server and who to work with,” said Justin Festa, vp of revenue at LittleThings. “The mix is going to have to exist for awhile.”

For the header-bidding curious, here’s why this matters. The way publishers fill their inventory has rapidly evolved in recent years. The so-called waterfall technique was brutally efficient and straightforward: It let publishers move their inventory from one market to the next. But it was also clunky and inefficient at driving revenue. So many publishers adopted header bidding, which simultaneously offers inventory to multiple exchanges before making calls to their ad servers. But the bids occur on users’ browsers, which slows down page loads — at a time when mobile is making page speed more important than ever. To speed up load times, publishers are now moving to server-to-server connections, which move the bids from browsers to servers.

But here’s the thing: These methods of selling inventory aren’t mutually exclusive. Many publishers using header bidding also use waterfalling to sell of remnant impressions. And now publishers who are moving to server-to-server are still maintaining on-page header bidding.

Jason Fairchild, CRO of OpenX, said that when publishers use a mixed model, all bidders have simultaneous access to the same inventory regardless if the bid occurs on servers or not. Because publishers aren’t carving out certain pieces of inventory for each type of vendor, it is tough for publishers to predict how the revenue mix between on-page bidders and server bidders will play out each month. Throughout the industry, most revenue is generated through browser-based header bidding, he said.

“We still have not seen data that shows that a publisher will be able to increase their revenue by solely leveraging a server-side solution for their header-bidding needs,” said Michael Grosinger, product manager at bRealTime. “Therefore, it is important to offer flexibility to publishers as they experiment and compare results between server-side and client-side connections.”

Because “server-to-server bidding is still very much uncharted territory,” Thought Catalog will adopt a mixed model as it integrates server-side bidding in upcoming months, said Cristina Calderin, Thought Catalog’s director of programmatic. LitteThings had a similar outlook.

“Server-to-server is not widely adopted and I don’t think will be for some time,” Festa said. “Having some server-to-server and some traditional bidders allows publishers to try out and understand server-to-server, learning and making tweaks to their setup, instead of going all in from the outset.”

Sources suggested that some publishers moving server-to-server keep on-page bidding around because the code on page might perform better than bids running through a third-party server-side product. Paul Bannister, co-founder of CafeMedia, said that his company will use a mixed model because server-side bidding can have issues with ID synching and CafeMedia has private marketplace relationships that will likely require the publisher to keep their partners bidding on page.

A Media.net spokesperson pointed out that even when results tilt in favor of server-side bidding, publishers can have difficulty moving entirely to server-to-server because tech vendors continue to fight over integrating with each other’s technology.

Calderin noted, “It doesn’t work if not everyone wants to play nice.”