When it comes to programmatic advertising, milliseconds matter.
To maximize revenue, publishers have to configure their Prebid setups to account for how quickly SSP auctions can respond to bid requests. Response times can vary based on factors like site visitors’ geographic region, what devices they’re using or whether the SSP has a server-side or client-side integration with the publisher.
Typically, publishers have to make such adjustments to their Prebid wrappers manually. But automation is taking some of the pressure off.
Ranker, a website specializing in ranked lists voted on by its users, has been testing Prebid optimization for the past year via Demand Manager, a solution provided by its preferred SSP partner, Magnite. Demand Manager recently added a Prebid automation feature that is now available for broad adoption, Magnite announced today.
Previously, Ranker used Demand Manager for A/B testing different manual Prebid setups. But in April, Ranker began experimenting with the new feature, which fully automates Prebid wrapper optimization at the bid request level in real time. Between April and July, this automation produced a 5% increase in ad spend from Ranker’s demand partners, compared to a control group of unoptimized impressions, said Alex Mason, VP of programmatic sales, yield and strategy at Ranker.
Solving the optimization problem
“Prebid setup effectively is a massive optimization problem,” said Sunil Gupta, VP of product management at Magnite.
Consider the task of manually setting Prebid auction timeouts: Setting timeouts too low can cause a publisher to miss out on winning bids from SSPs that take longer to respond. But setting timeouts too high can increase the chance the ad won’t render properly, which also has a detrimental effect on revenue.
Publishers can make manual adjustments, but they can’t really account for the sheer variety of factors that can impact how quickly SSPs can respond to bid requests, such as how far certain geos are from data centers and whether the SSP prioritizes the type of ad inventory on auction, Gupta said.
In contrast, Demand Manager’s new automated optimization feature can tailor auction timeouts to the individual bid request at the moment the impression is served, Gupta said. Plus, it can make those adjustments outside of normal work hours.
The solution bases its optimization on past ad auctions run by the publisher and how well SSPs performed under different auction criteria. The solution silos this data on a per-publisher basis; optimizations that worked for other publishers do not inform how Ranker optimizes.
The potential impact of this automation on Ranker’s ad business could be substantial – the publisher currently earns about 50% of its digital ad revenue from Prebid, across both open exchange and private marketplace (PMP) deals, Mason said.
Ranker began testing Demand Manager’s new automation capabilities on a subset of its international traffic back in April. But after seeing positive results, Ranker has since expanded the tool across its entire Prebid display business, Mason added.
Freeing up resources
“Before Demand Manager, Ranker had its own in-house version of Prebid, which is great if you have a large team of developers that are only focusing on Prebid,” he said.
But with the web dev team spread across site development and Prebid, making small changes was a headache.
“It was like pulling teeth if we wanted to add a new bidder, remove a bidder [or adjust] timeouts,” Mason said. “Having to tell a dev team that you want to increase timeouts – I don’t wish that on my enemies.”
Instead, Ranker can now hand off Prebid optimization to Magnite, whose business is based on understanding the nuances of how Prebid setups affect auction dynamics, he said. Ranker’s web dev team can instead focus on webpage layouts and site updates, while its ad ops team can focus on optimizing direct campaigns, which bring in higher CPMs than programmatic.
Which SSP and when
The automation is also helping Ranker wrap its head around which SSPs are worth doing business with and which bid requests to send to which SSPs under which circumstances.
Ranker typically has between 10 and 12 active SSPs at a given time, Mason said. Some SSPs respond better to auctions based in certain regions, and some are only worth pinging for specific types of inventory. Demand Manager automatically surfaces those insights and adjusts which SSPs receive the bid requests accordingly.
For example, Magnite is Ranker’s primary demand partner for PMP deals. So, during times when Ranker has a lot of active PMPs in effect, Demand Manager can set Magnite as a high-priority bidder and give it first crack at PMP auctions, Mason said.
Ranker also works with Flipp, which serves native retail media placements through Prebid, but only on certain pages and under specific circumstances. Those Flipp ads monetize very well compared to other placements, so Ranker wants to prioritize serving them when it can, Mason said. Demand Manager can recognize when a Flipp ad is likely to win an auction and adjust the bid wrapper in real time to facilitate that, while avoiding sending bid requests Flipp won’t be interested in.
Going forward, Magnite anticipates updating Demand Manager to weigh the impact different alternative IDs have on demand, Gupta said.
Such a feature would be instrumental in helping publishers decide which ID partners and SSPs to work with, Mason said.
Mason has already used Demand Manager’s findings to cut SSPs from Ranker’s Prebid stack. And he’s also used it to test new demand partners before fully integrating them. Previously, Mason would evaluate demand partners across a quarter or multiple quarters.
“It helps us prioritize our partners that are consistently performing, pick out the ones that are consistently underperforming, and makes that decisioning easier,” he said. “It gives me the ability to give newer partners a chance, rather than having this drawn-out process of evaluating them.”
