CTV is a performance channel now. What DV360, AdWeek and Amazon unBoxed showed, and how holdout markets prove CTV drove revenue. Our Performance Video Network (PVN) is built for where CTV is heading: biddable inventory, one buy across every screen, and measurement read off real sales.

This fall I sat in on Google's first DV360 Publisher Partnership Day, Amazon Ads unBoxed in San Francisco, and finally AdWeek in New York.

There was the same moment at all three: a speaker puts up a slide about CTV as a performance channel. The room nods, nobody argues anymore.

The problem is that most brands still can't prove it in their own sales data. If you want CTV to survive the next budget review, keep track of how search and social react to a CTV campaign, and set up holdout markets before it launches. That way CTV is graded on a level playing field from week one. It's the first thing we build into a new CTV program.

What changed in CTV buying this year?

Premium streaming inventory went biddable. At DV360's publisher day, Disney, Netflix, Tubi and Fox all pitched programmatic deal structures built for performance goals over pure reach.

If you've bought CTV before, you probably know the usual starting point: a premium rate card at $20-30 CPMs, bought one publisher at a time, with each buy capping frequency on its own.

Programmatic gives you a better way in. We start most new accounts on the open exchange at $10-14 CPMs, test audiences and creative there, then move money to what converts. And because one platform buys every screen, we can cap frequency at the household level.

Publishers are already packaging Q4 and Q1 tentpoles as curated private marketplace (PMP) deals with exclusive pricing. Plan in January and someone else has it.

Is AI-driven buying making CTV harder to measure?

AI makes CTV easier to run and harder to judge. At unBoxed, Amazon put AI at the center of its buying with a unified Campaign Manager and Performance+, which builds and optimizes ad groups across acquisition, remarketing and retention on its own.

If the reported numbers hold, Performance+ is an easy yes. But when the machine picks the audience, the bid and the placement, the platform is also grading its own work.

AdWeek pointed the same way. Nobody clicks a TV screen, so there are no fake clicks to clean out. It also means last-click attribution will never credit CTV, and search will keep taking the sale.

How do you prove CTV drove revenue?

Hold some markets out and compare sales to get to the number a CFO will sign off on. Pick markets that look alike, run CTV in some, leave the rest dark, and read the gap off your own sales data instead of platform conversions.

AI will keep taking over more of the buying. Proving what worked is still your job. This week, pull your CTV line and ask one question: if we turned it off in three markets, would we know what we lost? If the answer is no, that's your first test. What are you measuring CTV against right now? Tell me in the comments.

Being in these rooms gives me a head start on where the inventory and the pricing are going. That means our clients get new inventory, deal structures and measurement methods before they’re standard. Coming out of DV360's publisher day, we were already lining up Q4 and Q1 performance deals with publisher partners.

Managing our Performance Video Network (PVN) for clients is how we make sure they get the most out of every video dollar as buying changes. PVN runs CTV, YouTube, linear and social video as one buy, with one audience, one frequency cap and one measurement standard, so you can scale video and still prove what it drove.

Want to see where yours stands? We'll run a video gap analysis and send you a performance video snapshot: where your CTV, YouTube and social buys overlap, where frequency is running high, and what a holdout test could look like in your markets. Request one here.