How does CTV attribution work?

The basic mechanism

A TV screen has no click. To connect an ad on a television to a later action, a platform needs a way to recognise that the household that saw the ad is the same household that later visited a website or installed an app. Three building blocks do most of the work.

1. A pixel or tag on the advertiser's site or app records visits and conversions.
2. A household match links the TV that showed the ad to devices in the same home, usually through IP address and device graphs.
3. An attribution window sets how long after exposure an action still counts.

How platforms describe it

Vendors describe their methods in different levels of detail. On the pages we read:

• Vibe describes a pixel on the advertiser's site and an attribution window "from 1 hour to 7 days", reporting web visits, ROAS, app installs and in-app revenue.
• MNTN describes Verified Visits, built on a "proprietary identity graph" that links cross-device activity across "99% of available US households".
• tvScientific describes "patented attribution" using household-level deterministic matching, covering online and offline conversions.
• Tatari reports CPV, CPI, CAC and ROAS by network, creative, DMA and platform, and describes multi-touch attribution and DragFactor, which separates immediate from delayed response.
• StackAdapt lists "cross-device attribution" to site visits and conversions.

What attribution can and cannot tell you

Attribution answers "what did exposed households do afterwards?" It does not answer "what would they have done anyway?" A household that was already about to buy may see an ad and then buy, and attribution will count that sale. The longer the window and the broader the targeting toward existing customers, the more of that background activity gets counted.

This is not a flaw in any one vendor's method. It is a limit of attribution as a measure. That is why incrementality testing, covered in the next lesson, exists.

Settings that change the numbers

• Window length. A 7 day window counts more actions than a 1 day window. Compare platforms only on the same window.
• View-through versus click. CTV attribution is view-through by nature. Treat it differently from click-based search reporting.
• Retargeting and existing customers. Campaigns aimed at people who already know the brand show higher attributed ROAS, much of which may not be caused by the ad.
• Deduplication. If you also run search and social, the same sale can be claimed by several channels.

A practical checklist

1. Install the platform's pixel or connect your analytics before launch, not after.
2. Write down the attribution window and keep it fixed for the whole test.
3. Export results by day and by audience so you can compare them with your own analytics.
4. Treat attributed ROAS as a ceiling, not a final answer, until you have run a lift test.

How this shows up in our scores

Outcome measurement carries 20% of our weight. It rewards platforms whose pages describe how outcomes are tied to exposure, which metrics are reported, and over what window. It does not judge whether a method is accurate; we have not tested any of them.