Future of TV Briefing: WTF is IAB Tech Lab’s Concurrent Streams API?

In January, Index Exchange CEO Andrew Casale forecasted a “tsunami of supply” heading for programmatic ad market as more major sports and other live events become available to stream. That raised the question: How would the programmatic supply chain handle this inventory influx?

That’s a big question. And like many big questions, the best way to tackle it is to break it down into smaller problems and address those in turn. So what’s the most pressing part of the problem posed by this “tsunami of supply”?: How many impressions are actually for sale 

Pinning down that number can be tricky given how live viewership can fluctuate and how broadcasts can cut to an ad break seemingly on a whim. Which is where IAB Tech Lab’s Concurrent Streams API comes in.

WTF is the Concurrent Streams API?

The Concurrent Streams API is a specification that IAB Tech Lab introduced for streaming services to be able to report real-time viewership figures for live events. If a streaming service has adopted the API, ad tech entities – demand-side platforms, supply-side platforms, ad servers, etc. – are able to programmatically request how many people are streaming a live program, like an NFL game or awards show.

Why is the Concurrent Streams API necessary?

Because the ad tech ecosystem isn’t really set up for how live event viewership can fluctuate. 

For example, ad tech firms can set limits on how many queries they handle per second. Which is normally the responsible thing to do. But if Will Smith slaps Chris Rock and all of a sudden a ton more people start tuning into the Oscars, DSPs and SSPs won’t want to miss out on that additional inventory just because they didn’t update their QPS limits in time.

Similarly, an advertiser may have set up their programmatic buy to only snap up so many impressions per day. But if that Oscars influx means an opportunity to reach an even larger number of people concurrently – typically the primary aim for brand advertisers – no one wants to be the one to tell the brand’s CMO that they didn’t get that incremental audience because of the campaign caps.

How does the Concurrent Streams API work?

Think of it as the computer version of someone calling up Peacock and asking “How many people are watching ‘Sunday Night Football’ right this second?” Of course, it’s not so simple as that. 

For one thing, while IAB Tech Lab recommends the streaming service provide real-time viewership figures, it seems up to the streaming service to decide how frequently it will update those figures.

And then it’s not just a single number that the streaming service would be sharing. The streaming service would report the number of active streams by region and –

Region? To target ads based on geography?

Not really. It seems to be a way to minimize the amount of time it would take to respond to an impression and fetch the ad to slot in. As much as everything might seem proximate and instant on the internet, data does literally travel, which can take time. So declaring the region for various viewer segments would give ad tech firms a way to know whether to send advertiser bids and ad creative from their data center located on the West Coast or the East Coast or whichever is closest to the viewer segment and therefore would be quickest to handle the ad calls.

Zzz…

Still with me?

Uh, sure. Something about data proximity.

Yeah, so the API would break down active streams by region, and for each region, there could be a number for client-side ad insertion streams and a number for server-side ad insertion streams.

What’s the difference between client-side streams and server-side streams?

Client-side ad insertion streams refers to streams in which ads would be delivered to the viewer separately from the stream itself. This can give the advertiser more flexibility in deciding in the moment what ad gets served, but it can also make it take longer for the ad to get served, which could lead to it getting cut off or introduce some level of delay in that viewer’s stream.

Server-side ad insertion streams, by contrast, stitch the ads into the stream itself. That mitigates the latency issue but also limits the ability for ads to be controlled on the fly. You may be thinking thinking, What’s the point of breaking out these two stream types if SSAI ads have to be preloaded?

Was literally my next question.

Yep, so there’s actually a third, newer ad insertion type: Server-Guided Ad Insertion. This basically combines the low-latency of server-side ad insertion with the in-the-moment control of client-side ad insertion. And in the Concurrent Streams API – for some reason I don’t understand – SGAI streams would be counted within SSAI streams instead of reporting SGAI streams as their own property.

OK, this sounds nice and all. But don’t media companies usually try to keep these numbers close to the vest? And don’t they tend to rely on Nielsen and other measurement companies to keep count? And wouldn’t advertisers prefer independent measurement providers to keep those counts otherwise the media companies are doing the whole “grading their own homework” thing?

Yes, yes and yes. 

Right, so then isn’t having a way for any company to get a real-time count of live streaming viewership antithetical to all that?

Yes, in that the media companies are likely to be wary of making viewership figures broadly available. But they should be able to implement some controls over who can access that information, as is the typically the case with company APIs.

But it’s not like the media companies would be blasting these numbers out for public consumption, as mentioned above. And sure, they may prefer to not have to disclose these counts in such an accessible way, but they may be willing to make the trade-off if it means more ad revenue. 

As for the independent measurement, the numbers reported through the Concurrent Streams API aren’t gospel or even the whole truth. They are just what the media company is reporting as the available audience count for programmatic sales. A media company could deduct whatever viewership it would want to reserve for upfront advertisers, for example. The Concurrent Streams API counts are simply meant to be a way for a media company to tell the programmatic supply chain, “Hey, I’ve got a boatload of people watching this game right now, so do what you need to do if you want this inventory.”

Gotcha. So this is like step one in getting the programmatic supply chain set up to handle major live events. What else needs to happen?

Plenty. IAB Tech Lab has put together a whole Live Event Ad Playbook. The Concurrent Streams API is the first of four phases. Phase two is a Forecasting API, but I couldn’t find any details of how that API will work, let alone when it will be available.

What we’ve heard

“No matter what, it’s going to be less effective and likely more censored.”

— TikTok creator Trevor Barrett on how how the platform may change under U.S. ownership

Numbers to know

$150 million: How much Paramount is reportedly paying to acquire The Free Press and install Bari Weiss as editor-in-chief of CBS News.

$169.12: How much money the average person is spending on video, including pay-TV, each month.

7: Average number of subscription-based video services that people in the U.S. and Canada are paying for.

-13%: Percentage decline year over year in the number of movies and TV shows with $40+-million budgets that began shooting in the U.S. through the first three quarters of 2025.

What we’ve covered

Creators fear that cheap, scalable AI talent might soon come for their paychecks:

• AI-generated creators could cut into the audience attention and brand dollars going toward human creators.
• Creators think being a real person will protect their standing with audiences.

Read more about creators’ AI fears here.

Inside The Economist’s plan to grow revenues in a post-search, AI-driven future:

• The news publisher is rolling out a new twice-weekly show called Insider.
• The show will be available for free.

Read more about The Economist’s video plans here.

TikTok’s U.S. ownership shift raises creator concerns over algorithm changes:

• Creators are worried about TikTok’s U.S. ownership will change how their content is recommended.
• The app’s unpredictability can jeopardize creators’ brand deals.

Read more about creators’ TikTok concerns here.

CTV players hope importing social assets can lower brands’ barrier to entry:

• LG Ads has a tool that uses brand’s social videos as ads on its CTV platform.
• Dickies is among the brands that has used the social-to-CTV ad tool.

Read more about CTV’s social ads here.

What we’re reading

OpenAI’s Sora rev-share plans:

After launching its AI-generated social video app Sora last week, OpenAI CEO Sam Altman wrote in a blog post that the company plans to let copyright holders opt in to allowing their intellectual property on the platform — a reversal from its earlier opt-out stance — and teased plans to share revenue with rights holders (which could include creators).

Sora’s disinformation problem:

By enabling people to create AI versions of themselves that others can insert into realistic-looking AI-generated videos, Sora could create a new wave of disinformation despite the guardrails that OpenAI has introduced to the AI-generated social video app, according to The New York Times.

The AI-generated actor blowback:

Human actors are not taking kindly to an AI-generated actor named Tilly Norwood joining their ranks, with the likes of Emily Blunt and Natasha Lyonne (co-founder of AI film studio Asteria) decrying the AI creation who’s supposedly making the rounds with Hollywood talent agencies, according to BBC.

The Trade Desk’s CTV OS:

Almost a year after announcing plans to create a CTV operating system, the ad tech firm has secured a partner in DirecTV to create a customized version of TTD’s Venture OS in hopes of landing deals with smart TV manufacturers, according to Adweek.