Server-to-server bidding doesn’t end up replacing header bidding

While going server-to-server speeds up page-load times since ad calls are hosted on publishers’ servers and not on people’s browsers, it is still in its infancy and has low user match rates and inventory fill rates. For these reasons, most publishers that are moving their ad calls to servers keep their vendors for on-page header bidding — where publishers simultaneously offer inventory to multiple exchanges before making calls to their ad servers — to make sure they don’t lose revenue when they switch to newer technology.

Among the top 1,000 sites on Alexa that sell inventory programmatically, 21 percent use both server-side and browser-side header bidding to sell their inventory, up from 14 percent in September, according to ServerBid. Publishers that use both strategies include Vox Media, Meredith Corp., Ranker, Chegg, Granite Media, Thought Catalog, Intermarkets, CafeMedia and Bauer Xcel Media.

By contrast, 6 percent of the top 1,000 programmatic sites solely rely on server-side bidding, which is the same number of sites that used the technique back in September. Purch — the publisher of sites such as Tom’s Guide and Top Ten Reviews — and the 25 publishers that license its ad tech are examples of sites that only use server-to-server connections to sell programmatic inventory.

A downside of server-to-server connections is it is tougher for the supply-side platforms and demand-side platforms to match their user IDs. With on-page header bidding, each SSP has access to the user’s browser, which means they gather and use their own matching data to sync with the DSP’s matching data.

But with server-to-server wrappers — where one vendor aggregates the bids from all the other vendors in a cloud-based product — only the vendor collecting the bids has access to the user’s browser, which means the other vendors have to sync their data to the aggregator’s data. This additional step can reduce matches.

For example, Ranker included the code of one of its SSPs on its webpage as well as inside a server-to-server wrapper. When Ranker removed the SSP’s code from its page, the revenue coming in from that SSP dropped by 25 percent, said Ranker CTO Premesh Purayil. This is because the SSP could not successfully match user IDs within the server-side wrapper without having code on the publisher’s page.

Likewise, CafeMedia ran a test for one week where it only sold its programmatic inventory through server-to-server connections for a few of its websites. This boosted page speed by about 40 percent on average, but the revenue coming in from programmatic auctions declined by nearly 30 percent, said CafeMedia evp Paul Bannister.

With the ad tech and digital publishing industries building ID consortiums, the revenue-generating ability of server-to-server connections should improve in the next few years, Bannister said. Theoretically, if more vendors and publishers start sharing user ID data, then the match rates will improve and programmatic platforms will fill more inventory.

But those days aren’t here yet. Danny Khatib, CEO of Granite Media, said his company’s sites “will likely stay on the hybrid model until server-to-server performance improves considerably.”