CTV attribution: windows, household matching and what they can prove

A TV cannot be clicked. So CTV attribution works by matching: the platform knows which household's TV received an ad, and it links that household to phones, laptops and other devices through a household graph built from IP addresses, logins or partner data. When one of those devices later visits your site or buys, a tracking pixel or server-side connection records it and the platform credits the ad.

The attribution window is how long after an exposure an action still counts. A one-hour window credits only actions soon after the ad; a seven-day window credits actions over a week. Longer windows credit more actions, including many that would have happened anyway. When you compare platforms, use the same window, or you are comparing settings rather than results.

Match rates matter too. Not every household that sees an ad can be linked to devices, so attribution only ever covers part of the exposed audience. A report that shows strong results on a small matched share may say little about the rest. Ask what share of impressions could be matched and how unmatched exposures are handled.

Attribution shows that exposed households acted. It does not show that the ad caused them to act. People who already like your brand are more likely to be targeted and more likely to buy, so attributed results can look strong even when the ads changed little. That is why incrementality testing exists, covered in the next lesson.

• Which window the report uses, and whether you can change it.
• Whether view-through actions are counted and how they are labelled.
• How households are matched to devices, and what share of exposures could be matched.
• Whether results can be exported or read by your own analytics or a third-party partner.
• How the platform deduplicates actions also credited to other channels.

Set the rules once and apply them everywhere. Use one window across platforms, decide whether view-through actions count, and send every platform's exposure data to the same analytics tool or partner if you can. Keep a short written note of the settings next to every report, so that a result read in three months is read with the same assumptions. If platforms report on different windows or matching methods, do not compare their attributed ROAS directly; compare the lift from a test instead.

Finally, treat attribution as a daily steering tool and incrementality as the verdict. Attribution tells you which audiences and creatives respond while a campaign runs; a test tells you whether the campaign was worth running at all.

If a platform's pages do not describe attribution, ask how it reports outcomes before you buy.