Connected TV as a Performance Channel in the US

Connected TV is sold as television you can finally measure. The honest position is that its measurement is weaker than digital and stronger than linear, and most of the disappointment in the channel comes from advertisers who were told otherwise. At $33.5 billion and growing 28% year on year, CTV is the fastest-growing premium format in US digital advertising, driven by cord-cutting reaching critical mass and every major streamer now offering an ad-supported tier. It is a genuine performance channel — but only when judged with methods that suit a screen with no click.

A capability page from Digital, Ecommerce & Performance Marketing in the United States. See also Meta performance marketing, which shares the delayed-conversion problem. Last reviewed August 2026.

1. Why CTV grew so fast

Connected TV has become the fastest-growing premium ad format as cord-cutting reached critical mass and every major streamer moved to offer an ad-supported tier. At $33.5 billion and 28% annual growth, it is outpacing every other established digital channel.

The supply side explains most of it. When streaming services that previously sold only subscriptions introduced advertising tiers, a large quantity of premium inventory arrived in the market over a short period. Advertisers did not suddenly demand more television; television became newly purchasable in digital ways.

CTV’s growth is a supply story first. A great deal of premium inventory became available in a few years because the streaming business model changed, not because advertisers discovered a new appetite.

2. Better than linear, weaker than digital

The pitch for CTV is addressability and measurement. Both claims are true relative to broadcast television and overstated relative to search or social.

Comparative assessment based on the structural characteristics of each channel. Specific capabilities vary by platform, publisher and buying route.

Placing CTV honestly in that middle position is the single most useful thing an advertiser can do before buying it, because expectations set at the digital end of the scale guarantee disappointment.

3. Household is not a person

CTV targeting generally operates at household level. A television in a living room is watched by whoever is in the room, which means the person you targeted may not be the person who saw the ad.

Illustrative. Household-level delivery is inherent to the format rather than a limitation of any particular platform. Co-viewing can mean an impression reaches more people than counted, or that it reaches the wrong person entirely.

4. How to actually measure it

There is no click on a television, so every CTV measurement approach is inferential. That is fine, provided the method is chosen deliberately rather than borrowed from search.

Measurement approaches for a channel without direct response. Household-to-device matching depends on identity resolution quality and applicable consent requirements.

Geo holdout is the most defensible. Run CTV in some markets and not others, hold everything else constant, and compare total sales rather than attributed conversions.

5. The frequency problem

CTV inventory is fragmented across many apps and publishers, and frequency capping generally works within a buying platform rather than across the whole ecosystem. The predictable result is the same viewer seeing the same advertisement far more often than the media plan intended.

This is worth monitoring actively rather than assuming, because excessive frequency wastes budget and damages the brand it is meant to build. Where possible, consolidating buying reduces the problem; where consolidation is not possible, survey or panel data is usually the only visibility available.

6. Creative built for a room, not a thumb

The most common CTV mistake is running social creative on a television. The two contexts share almost nothing.

Comparison of creative context. Operational judgement based on the physical differences between the two viewing environments.

The last row matters most for performance advertisers. Since there is no tap, the ad must leave something a viewer can act on later — a brand name they will search, a phrase they will remember, or an offer distinctive enough to look up.

7. Who CTV suits

Fit assessment based on the delayed-response characteristics of the format. Operational judgement.

8. Buying: platform, programmatic or direct

Buying route comparison. With 91.5% of US digital display bought programmatically, programmatic access to CTV is the default route for most advertisers.

9. What this page does not cover

Scope statement. Household-to-device matching in particular has consent implications that vary by state and require legal review.

10. The 90-day test plan

Indicative sequencing. Choosing control markets before spending is what makes the entire exercise measurable; retrofitting a control group afterwards rarely works.

11. Mistakes to avoid

Recurring errors when performance advertisers enter CTV; illustrative.

12. What changes in 2027

Inventory keeps expanding. With every major streamer now running an ad-supported tier and cord-cutting continuing, supply growth should moderate pricing pressure relative to more constrained channels.

Shoppable television advances. As commerce functionality spreads across formats, the gap between exposure and purchase narrows on TV as it has in social — which would materially improve the channel’s measurability.

Retail media and CTV converge. Retailer purchase data increasingly informs television targeting, which brings closed-loop measurement to a format that has never had it.

Key Takeaways

• CTV is $33.5bn growing 28% year on year, the fastest-growing premium format, driven by every major streamer adding an ad tier.
• Its measurement sits between linear and digital. Expectations set at the digital end guarantee disappointment.
• Targeting is household-level. One impression may reach several people, one person, or nobody.
• There is no click, so geo holdout testing against total sales is the most defensible measurement available.
• Frequency caps generally work within a platform, not across the ecosystem — monitor actual exposure rather than assuming the cap holds.
• Do not run social creative on television. Sound is on, the screen is far away, and there is nothing to tap.
• Leave something searchable. The response mechanism is memory and a second screen.

Frequently Asked Questions

Conclusion

Connected TV is the most honest test of whether a marketing team understands its own measurement. The channel does real work, at real scale, in a format people actually watch — and it will never produce the clean attribution that search does, because there is nothing to click.

Advertisers who accept that and design an experiment get a durable new channel with expanding inventory and improving targeting. Advertisers who buy it expecting a dashboard that looks like Google Ads conclude within a quarter that it does not work, cancel it, and never find out. The difference is not budget or creative. It is whether the measurement was chosen to suit the medium or borrowed from a different one.