The programmatic ecosystem is becoming a more agentic system, but that’s not to say the humans involved don’t have evolving expectations of their own.
In recent years, buyers have become “a lot more savvy,” OpenX CTO Joel Meyer told AdExchanger. Many such advertisers, he said, are seeking ways to sit closer to their ad inventory sources, which affords better access to data signals that influence buying decisions.
On Tuesday, relatedly, OpenX (an SSP that also caters to the buy side) announced OpenXBuild, a suite of API solutions for advertisers to better control ad performance.
On the origin of APIs
OpenX already offers curation tools and an audience identity graph, so OpenXBuild seemed like the natural “evolution of our offerings,” said Meyer.
The suite includes three APIs, each with a unique function: the Auction Insights API (which gives a historical view of bidstream events, like exposures and auction outcomes), the Identity Resolution API (an audience graph so data partners and advertisers can introduce audience segments to target or sell) and the Real-Time Bidstream API (where advertisers can introduce custom logic and models).
One pilot partner is the CTV ad-buying startup tvScientific, which was recently acquired by Pinterest. It has been testing uses for the Real-Time Bidstream API. By positioning tvScientific’s own AI model directly within OpenX infrastructure, the TV ad tech vendor is “close to the supply,” said Teddy Jawde, tvScientific’s SVP of product management.
The ability to see the entire inventory makes ad buying more efficient and enables more accurate bid decisioning, according to Jawde.
For instance, he said, tvScientific is faced with “enormous” numbers of impressions per second – which is both a blessing and a curse. Advertisers are, of course, hoping for a greater number of impressions, but when faced with so many, it’s hard to track who they reach and how effectively they do so.
Proximity to the programmatic supply means tvScientific sees the impressions up close and tags the ones that they want to bid on, based on criteria like content type, geography and channel, said Jawde. The tag “serves as a signal to our bidder that says, ‘buy this,’” he explained, which “creates a feedback loop” back to OpenX to prioritize similar impressions for that sort of buyer in the future.
Sign of the times
OpenXBuild is a suite of APIs. But make no mistake: It’s still very much a product of the new agentic era.
The IAB Tech Lab’s Agentic RTB Framework (ARTF) might not roll off the tongue (“we can debate the name later on,” Meyer quipped), but it partially inspired the development of OpenX’s RTB API. The ARTF advances the idea of containerization, meaning that one company can deploy its own agents or code directly in another’s infrastructure.
The ARTF and other similar open-source developments have led to an increase in ad tech companies wanting to “package up their logic and deploy it in our bidstream,” said Meyer. OpenX already provided clients with access to its identity graph, where data partners could sell their audiences. It also incorporated model training into its bidding intelligence data set, which lets buyers like tvScientific more easily see exactly what they were bidding on and how those ads were performing.
Those data sets were “the preexisting parts,” according to Meyer.
“When we added the ability to integrate with the bidstream, we said, ‘Hey, this is now a full suite of APIs.’”
Slow and steady
But … what about MCPs? Didn’t they usurp APIs?
The answer, in short, is that it’s not a question of one or the other.
As Jawde put it, an MCP allows ad buyers to integrate their own AI tools more easily within APIs.
And as of right now, “interacting with the bidstream through MCP would be incredibly inefficient,” said Meyer, since humans are still developing the best ways to determine which impressions make sense in a given moment.
The task isn’t ready to be outsourced to LLMs, he said, when “a human today can’t do [it] either.”
