CTV Attribution | How a TV Ad Is Tracked Back to a Sale

Published August 28, 2026· 7 min read

Key takeaways

• CTV attribution usually works by matching a household IP address, not by tracking a click.
• View-through conversions record exposure and timing, not proven cause.
• Platform-reported CTV conversions overlap heavily with search and social claims on the same order.
• Geo holdout tests are the most credible way to measure what CTV genuinely added.

What is CTV attribution?

CTV attribution is the set of methods used to connect a connected TV ad impression to a later action, usually a site visit, a lead form, or a purchase. It exists because connected TV has the reach and creative impact of television with none of television’s traditional measurement problem, and none of digital’s clean click trail either.

Nobody clicks a television. That single fact shapes every method below.

How does the household IP match work?

The dominant method is IP matching. The streaming device on the living room television reports an IP address when the ad serves. When a phone or laptop on that same home network later visits your site, the measurement platform sees a shared IP and connects the two events.

Step	What happens	What it proves
1	Ad serves to a streaming device, IP recorded	The household was reached
2	A device on the same IP visits the site	Same network, probably same household
3	That visit converts inside the lookback window	Timing correlation
4	Platform credits the CTV impression	Correlation, not causation

The method is reasonable and it is also loose. Shared networks, offices, apartment buildings and VPNs all blur the match, and a household of four people is treated as one identity.

What is the difference between a view-through and a click-through conversion?

A click-through conversion means somebody acted on the ad directly. A view-through conversion means somebody in an exposed household converted afterwards without clicking, inside the attribution window the advertiser set.

Almost all CTV conversions are view-through. That makes the reported number sensitive to a setting rather than to performance. Widen the lookback window from seven days to thirty and CTV conversions rise, without a single thing changing in the campaign.

We set the window before the campaign launches and we do not move it mid-flight. Moving it mid-flight produces a lift that only exists in the report.

Why does CTV double count with search and social?

Because the same order gets claimed more than once. A viewer sees a CTV spot, searches the brand two days later, clicks a search ad, and buys. CTV claims a view-through. Search claims a click. The business took one order.

Google documents several of these mechanisms in its own guidance on different ways to track conversions. This is the standard marketing attribution overlap problem, and CTV makes it worse because view-through crediting is passive. A channel that claims conversions without requiring any interaction will always look efficient in its own dashboard.

Reading total revenue against total media spend catches this. Summing platform-reported revenue does not.

How do you actually prove CTV worked?

Run a holdout. Split comparable geographies, run CTV in one set and not the other, keep everything else constant, and compare total business outcomes rather than platform-reported ones.

• Geo holdout: the most practical test for most advertisers, and the one we reach for first.
• Audience holdout: cleaner where the platform supports suppressing a randomised group.
• Time-based on and off: weakest, because seasonality and other channels move underneath it.

This is incrementality work, and it is the only category of measurement that answers what would have happened without the spend. A conversion lift study serves the same purpose where the platform offers one.

What we tell clients about CTV measurement

CTV attribution is directionally useful and precisely wrong. Treat the platform number as a signal that the channel is being delivered to the right households at the right frequency, then prove the business impact with a holdout.

We plan CTV as a demand-creation channel measured on total business lift, with the frequency and household reach figures used to confirm delivery quality. Judging CTV purely on its own view-through report tends to either flatter it or bury it, depending on one window setting.

What does a holdout actually show?

On the Goldback program we ran connected TV alongside search, paid social and out-of-home, and the measurement question was settled by a pause rather than by a report. When media spend was cut in February, paid sessions fell 88.1% and total site sessions fell 54.7%, even though organic search barely moved. Traffic that looked unpaid decayed with the paid layer, which is the pattern a view-through report can suggest but never prove.

The connected TV layer itself delivered 6.05 million impressions to 1.1 million households at a $1.97 cost per verified visit. Those are our own figures from a published campaign, and they describe delivery and measured visits rather than claimed conversions. Full detail is in the Goldback case study.

FAQs

• How is a CTV ad tracked to a sale? Most CTV attribution matches the IP address of the streaming device to the IP address of the device that later visits the site. If both share a household network, the platform credits the impression. Some measurement uses matched device graphs or panel data instead.
• Can CTV ads be clicked? Usually no. A viewer sees the ad on a television and acts later on a phone or laptop. That gap is why CTV attribution relies on household matching and lift testing rather than the click tracking used in search and social.
• What is a view-through conversion on CTV? A conversion recorded because someone in the household saw the ad and converted within a lookback window, without clicking. It proves exposure and timing. It does not prove the ad caused the purchase.
• How do you prove CTV actually worked? Hold out a comparable geography or audience, run CTV in one and not the other, then compare total conversions rather than platform-reported ones. Geo holdouts are the most credible test available to most advertisers.
• Why do CTV and search both claim the same sale? View-through crediting is passive, so CTV can claim a conversion without any interaction. Search claims the same order on a click. Reading total revenue against total spend catches the overlap.