Connected TV stopped being the emerging channel a while ago. Budgets have moved. Inventory has multiplied. Just about every advertiser with a performance mandate now has a CTV line item, and by most measures the channel has scaled faster than almost anything that came before it in digital advertising.
Ask a performance marketer how confident they feel about their CTV results compared to their search or social results, though, and you usually get some version of the same answer: we're spending more, we think it's working, but we're not totally sure why, or whether it could be working better.
That gap between scale and optimization is the real story in CTV right now. It isn't really a measurement problem and it isn't a quality problem either. It's an infrastructure problem.
CTV grew up fast because it inherited two things at once: the targeting and bidding logic of programmatic digital, plus the premium, brand safe inventory people associate with television. That combination made it an easy budget shift to justify. What it didn't inherit was the testing and feedback infrastructure that channels like paid search and social spent more than a decade building.
Search and social reached performance maturity through millions of small experiments: constant creative rotation, granular audience tests, bid adjustments that happen within hours of seeing the data. CTV largely still runs on a planning cadence borrowed from linear TV. A handful of creative versions, a flight plan locked in weeks ahead, a results readout that lands after the budget is already spent.
The channel modernized. The way teams operate it mostly didn't.
You can see this pattern in almost any CTV program if you look closely enough.
Creative tends to get frozen too early. Producing a CTV spot is expensive enough that most advertisers ship two or three versions and let them run for the entire campaign. There's rarely a real mechanism for learning which part of the creative, the pacing, the message order, where the call to action lands, is actually driving results, because nothing feeds performance data back into the next creative decision while the campaign is still live.
Attribution tends to lag behind the buy. A lot of CTV programs still get judged on completion rate and reach, not because advertisers actually care most about those numbers, but because connecting an impression on a streaming device to a purchase or signup has historically been hard to do cleanly. Teams end up making decisions based on the metrics that happen to be available rather than the ones that matter.
And optimization tends to happen between campaigns instead of inside them. In search or social, an underperforming ad gets adjusted the same day. In CTV, optimization more often means a debrief once the flight has ended and a slightly different plan for next quarter. That's not really a feedback loop. It's a postmortem.
None of this means CTV as a channel is broken. It's the predictable outcome of pouring digital scale budget into a medium that hasn't yet built digital speed infrastructure underneath it.
The advertisers pulling ahead here aren't necessarily spending more on CTV. They're spending differently.
They're running creative tests continuously, with several variants live at once and underperforming versions getting rotated out based on actual outcome data rather than a gut call after the fact. They're using attribution that ties a streaming impression to something real, a purchase, an install, a signup, instead of stopping at completion rate. And they're shrinking the time between what a campaign learns and what changes because of it, from a quarter down to a matter of days.
None of this requires new ad formats. It requires new infrastructure, the layer that sits between the buy and the outcome, constantly comparing what was expected to what actually happened and adjusting in response.
We built Starti around a simple assumption: CTV should be held to the same standard as any other performance channel, judged by results rather than impressions, and able to improve mid flight rather than only in hindsight. SmartReach™ AI evaluates creative and audience performance continuously while a campaign is live, and OmniTrack attribution connects that performance back to real outcomes instead of proxy metrics. Because our model is built around paying for results, the incentive to keep optimizing never really ends. It's the whole point of the system.
CTV's scale was never the problem. The missing learning loop underneath that scale is. Closing that gap is what decides whether your next dollar of CTV spend performs the same as the last one, or noticeably better.
Curious what a CTV campaign that keeps improving while it runs actually looks like?
Talk to the Starti team about turning your next flight into a learning system instead of a one shot bet.
