The Measurement Problem Nobody Wants to Talk About
How CTV advertising became too big to measure accurately
The advertising industry has spent the last five years celebrating connected TV’s arrival as the future of video advertising. Budgets have shifted. Presentations have been made. Awards have been won. But there’s a problem that keeps getting glossed over in conference keynotes: we still don’t really know how to measure it properly.
The Trade Desk’s move to open up third-party measurement for CTV campaigns isn’t just a feature update. It’s an admission that the current state of measurement is inadequate. When platforms start emphasizing “independent verification” as a selling point, it usually means advertisers have been asking uncomfortable questions about the numbers they’re seeing.
The walled garden issue
Here’s what’s actually happening: most CTV measurement today relies on the platforms themselves to grade their own homework. Netflix tells you how your Netflix ads performed. Roku tells you about Roku. It’s not that these companies are being dishonest, but there’s an inherent conflict when the entity selling you media is also the only one measuring whether that media worked.
Traditional TV had Nielsen. It was flawed, but it was independent. Digital advertising had third-party verification from companies like IAS and DoubleVerify. CTV has mostly had... the platforms’ own reporting dashboards.
Why advertisers are finally pushing back
The pressure for change is coming from an unexpected place: the CFO’s office. As CTV budgets have grown from experimental line items to meaningful portions of media spend, finance teams want proof that the money is working. “Trust us, it’s performing great” doesn’t cut it anymore when you’re spending eight figures.
The Trade Desk’s partnership with measurement providers like EDO represents a tacit acknowledgment that marketers need someone other than the media seller confirming their ads drove business outcomes. EDO’s approach - tracking real-time consumer behavior like brand searches and website visits following ad exposure - provides the kind of tangible proof that finance teams actually care about.
The attribution mess
But measurement verification is only part of the problem. The bigger issue is attribution in a fragmented media landscape. A consumer might see your CTV ad on their smart TV, then later search for your brand on their phone, read a review on their laptop, and finally make a purchase in-store. Which channel gets credit?
Traditional attribution models weren’t built for this complexity. They certainly weren’t built for CTV, where a single “impression” might be viewed by multiple people in a household. The old assumption of “one device, one person” falls apart when you’re talking about a 65-inch screen in someone’s living room.
What actually works
The companies making progress on CTV measurement are the ones combining multiple data sources. They’re using panel-based data for reach measurement, pixel data for digital conversions, loyalty program data for offline impact, and survey data for brand lift. No single source tells the complete story.
Comscore’s Campaign Ratings tool, which The Trade Desk integrated for local CTV measurement, is instructive here. It combines set-top box data, smart TV data, and census-based projections to estimate reach and frequency. It’s not perfect, but it’s better than relying solely on impression logs from streaming platforms.
The incrementality question
Perhaps the most important question gets asked the least: would these customers have found you anyway? Incrementality is the third rail of advertising measurement. Nobody wants to run true holdout tests because they might reveal that a chunk of ad spend isn’t actually driving incremental sales.
But some advertisers are starting to demand it. They’re running geo-holdout tests, turning off CTV advertising in specific markets and measuring the impact on sales. The results are often sobering. CTV works, but it doesn’t work as well as the CPM-to-conversion reports suggest.
The future isn’t going to be easier
As CTV continues to fragment - with more streaming services, more devices, more viewing occasions - measurement will only get harder. Privacy regulations are tightening. Third-party cookies are disappearing (eventually). Device IDs are being restricted.
The solution isn’t going to come from a single measurement provider or a unified dashboard. It’s going to require marketers to get comfortable with imperfect data and probabilistic modeling. It will mean combining first-party data with modeled projections and accepting ranges rather than precise numbers.
What this means for marketers
If you’re spending meaningful money on CTV, you need to be asking your partners harder questions:
What data sources are being used for measurement? Is it just server-side impression counts, or are there actual consumption signals? Who is verifying the numbers? Is there independent third-party measurement, or are you trusting the platform’s internal reporting? What’s the methodology for attribution? How are co-viewing and cross-device behavior being handled?
The uncomfortable truth is that much of CTV advertising is being bought on faith. The creative looks good on a big screen. The CPMs seem reasonable compared to linear TV. The targeting feels more precise. But when you drill down into actual measurement methodology, it often doesn’t hold up to scrutiny.
That doesn’t mean you should stop advertising on CTV. It means you should stop pretending the measurement is better than it actually is. Acknowledge the limitations. Build in uncertainty. Use multiple measurement approaches. And push your partners - whether that’s The Trade Desk, the streaming platforms, or your agency - to provide independent verification.
The next stage of CTV’s growth depends on solving the measurement problem. The platforms that figure it out first will win. The ones that keep pretending everything is fine will eventually lose advertiser trust.
The question is: how long are advertisers willing to spend money on media they can’t properly measure?