CTV Ad Spending: Decoding the Black Box in 2026

The Connected TV (CTV) ad market is supposed to hit $30 billion in the US by 2026, but let’s be real, a huge challenge hasn’t gone away: we’re often flying blind about where the money goes and what it accomplishes. For a lot of marketers, this lack of real insight into CTV advertising performance turns the whole channel into a “black box,” which means you can’t confidently allocate your budget or plan your next move. So, how do you get inside that box and see if your investments are actually working?

The Opacity Problem: Why CTV Measurement Trails Other Channels

Compared to the measurement you get from search or social, tracking CTV ad performance feels like a total guessing game. The problem is fragmentation, you’re juggling dozens of streaming services, device manufacturers, and ad tech platforms, and every single one has its own way of collecting data and its own reporting standards. This inevitably creates huge discrepancies in the numbers. For example, your demand-side platform (DSP) might report 1 million impressions, but the publisher only counts 800,000, and your third-party verification partner says it’s just 700,000. That 30% gap isn’t a rounding error. It’s a massive slice of your budget that’s completely unverified.

Server-side ad insertion (SSAI) is a huge reason why things are so opaque. On one hand, it gives viewers a great experience by stitching ads directly into the content stream. On the other hand, it completely severs the direct line of communication between your ad creative and the user’s device. This means your traditional client-side measurement tools, the ones that depend on pixels and JavaScript tags, become almost useless because they simply can’t fire properly. The impression count happens on the publisher’s ad server, which creates a “trust me” situation where you have little choice but to take their word for it. Without any real industry-wide standards for SSAI measurement, you’re left staring at aggregated reports that don’t have the granular data, like which specific shows or apps are performing, that you need for any kind of real optimization. The Interactive Advertising Bureau (IAB) is making progress with its Open Measurement SDK (OM SDK) for CTV, but adoption is slow and the tech is still developing.

Demystifying Impression Discrepancies and Viewability

To understand your CTV performance, you first have to reconcile the impression counts. The discrepancies you see are usually because of three things: everyone has a different definition of a “viewable” impression, there are technical reporting issues, or there’s fraud. For instance, one platform might count an impression the moment the ad starts to load, while another needs a specific percentage of the ad to play for a minimum time. This difference in definition directly skews your campaign results. A recent DoubleVerify report shows that while CTV viewability rates are generally high, they can still swing wildly between publishers and devices, with some inventory not even meeting basic quality thresholds. You must demand clear definitions from your partners before you spend a dime.

And the whole idea of “viewability” is changing for CTV. An ad is usually full-screen and non-skippable, so you’d think it’s always seen. But what about co-viewing, where multiple people are watching one screen, or background playback where the TV is on but nobody’s in the room? These factors make it hard to know if you’re reaching a real person. Counting a served ad just isn’t enough. You should be asking for metrics like video completion rates (VCR) and, even better, attention metrics that try to measure if a person was actually engaged. For example, some approaches use panel data to determine if someone was physically in the room and looking at the screen. Tech from companies like TVision which does exactly that, gives you a look at the more advanced measurement that’s coming.

Attribution Challenges in a Fragmented Ecosystem

The single biggest challenge in CTV is attribution, connecting an ad someone saw on their TV to a later action like a website visit, an app download, or an actual purchase. Because there are no persistent identifiers (like cookies) on CTV devices and privacy rules are cracking down on cross-device tracking, the old last-click attribution models you might use elsewhere are completely ineffective. Think about the typical customer journey: someone sees an ad on their living room smart TV, researches it on a tablet, and then buys it on a desktop computer. How do you prove the TV ad started that chain of events?

To solve this, unified identity solutions are becoming more popular. Companies like LiveRamp and The Trade Desk (with its UID2.0 initiative) are building frameworks that create privacy-conscious identifiers that can be matched across a person’s different devices. Although they aren’t everywhere yet, these solutions are a promising way to get more accurate cross-device attribution. You should also be looking at probabilistic attribution models, which use signals like IP addresses, household device graphs, and viewing patterns to infer the connection between seeing a CTV ad and making a purchase. This approach gives you a much more realistic view of performance than trying to shoehorn a last-click model into CTV. From what I’ve seen, the brands that succeed with CTV attribution are the ones willing to run real experiments with incrementality testing, like using geo-based tests or holdout groups to isolate the actual impact of their campaigns.

Combating Fraud and Ensuring Brand Safety

As CTV has grown, it’s become a huge target for fraud. Sophisticated bots are out there mimicking legitimate streaming traffic, creating millions of fake impressions that drain ad budgets. Brand safety is also a major concern, you have to make sure your ads don’t run next to inappropriate or brand-damaging content. The “black box” of CTV just makes these problems worse. Because you often don’t get granular, real-time data, you might not realize that a bot is generating thousands of fake impressions from a single IP address until after your campaign has already spent a ton of money.

You have to fight this with a multi-layered defense. First, only work with reputable DSPs and publishers that use strong pre-bid fraud filtering. Second, you have to bring in independent third-party verification (TPV) solutions. Companies like Integral Ad Science (IAS) and DoubleVerify are specialists in detecting sophisticated invalid traffic (SIVT) in CTV environments, giving you post-impression reports that can spot suspicious activity that got through the initial filters. Because these TPV partners have direct integrations with the big CTV platforms, they can verify impressions and viewability far more accurately than your DSP can on its own. Finally, demand to know where your ads are running. Open programmatic exchanges are a breeding ground for fraud, while direct deals with premium publishers give you much more control and trust. Minimizing fraud’s damage requires this kind of vigilance and picking your partners carefully.

The Path Forward: Demanding Transparency and Control

Getting a real handle on your CTV performance means you have to be proactive and demand more from your partners. Advertiser pressure is what forces the industry to improve. Here are the steps I consider non-negotiable:

1. Demand Granular Data: Stop accepting aggregated reports. You need impression-level data that includes the device type, publisher, content genre, and placement details. This is the only way you can do real analysis and optimization, like shifting budget away from a low-performing streaming service.
2. Standardize Measurement: Push your partners to adopt standards like the IAB’s OM SDK for CTV and make them give you a crystal-clear definition of their viewability metric. If a partner can’t explain their measurement methodology clearly, it’s a huge warning sign.
3. Invest in Verification: Third-party verification is not a nice-to-have. It’s a must-have. It’s the only way to independently confirm that you got the impressions and viewability you paid for and to catch sophisticated fraud. Build this cost into your plan from day one.
4. Experiment with Attribution: Get away from last-click models that don’t work here. You need to be running incrementality tests and exploring multi-touch attribution with unified ID solutions to understand CTV’s real contribution.
5. Build Direct Relationships: Whenever you can, go straight to the premium CTV publishers. This almost always results in greater transparency and better inventory quality, plus it gives you more control over ad placement and data access.

The CTV space isn’t slowing down, and its measurement problems will only get more complex as new platforms and ad formats appear. If you actively dig into your data, hold your partners accountable for every dollar, and adopt new measurement tech as it comes, you can finally break open that “black box” and understand the real return on your CTV investments.

Getting real ROI from CTV comes down to being rigorous about transparency and measurement. The advertisers who are winning are the ones who push for granular data, insist on independent verification, and use advanced attribution models. They’re the ones who can show, for instance, that their CTV campaign drove a measurable lift in site traffic from a specific geographic region, proving its value beyond a simple impression count.