One Audience, Several Evidence Layers
The question behind cross-screen video measurement is not whether television, OTT, and digital platforms can each produce a reach number. They can. The question is when those reports support one defensible estimate of an unduplicated audience, and when combining them would hide incompatible definitions or unknown overlap. This final synthesis connects the common exposure ruler, universe and identity evidence, model-free bounds, model-based estimates, privacy, quality control, and audit into a decision path. The conclusion is deliberately calibrated: one audience is achievable as a defined measurement construct, but no single total is stronger than the evidence layers beneath it.
The unified audience is an evidence stack
A credible cross-screen total begins before deduplication. It begins with a comparable exposure. Television, OTT, and digital video must use a common qualifying basis for the combined metric: the MRC Phase I standard specifies a full-on-screen video threshold for at least two continuous seconds, requires filtration for General and Sophisticated Invalid Traffic, and distinguishes the cross-media basis from platform-specific diagnostic metrics. The event should be measured close to delivery and after the ad or content has loaded or rendered. Dynamic ad models require discrete ad-occurrence measurement; a content audience cannot silently become an ad audience merely because both occurred in the same program.
Duration and time are part of the same stack. A report needs a defined reporting period, a consistent treatment of inactivity and session cutoffs, and enough measurement granularity to support the credited duration. A connected device continuing to send or record activity while a television is off can inflate an apparent audience if it is treated as viewing without adjustment or disclosure. Duration weighting can improve the comparability of exposure measures, but it does not establish attention or effectiveness. The first layer answers what kind of exposure is being counted, not what the exposure caused.
The second layer is the universe. The MRC framework defines the minimum cross-media universe as the deduplicated total of persons in the media universe for each medium, or more generally total persons. A component population—such as people reachable through a particular digital, mobile, or television dataset—may be valid for component measurement, but it cannot become the cross-media denominator without a supported projection to the intended population. “Census-like” collection is not automatically complete: unsupported devices, providers, browsers, platforms, and privacy-restricted environments can create systematic coverage gaps.
The third layer is the reporting unit. A device record, browser record, household, and person are different objects. One person can use several screens; several people can use one connected television; one household can contain multiple potential viewers. Person-level reporting therefore needs evidence that is at least partly traceable to people, and cross-device deduplication needs empirical support that reflects actual consumer duplication. An identity method that merely connects records is not the same as a validated claim that those records belong to one person.
The fourth layer is overlap. If the same person appears in more than one platform report, they should count once for unique reach during the defined period. But the overlap method must match the universe, media types, device mix, and privacy conditions. IP address alone is not a sufficiently stable foundation for audience uniques, and a model without a relevant truth or learning set cannot prove that its links reflect real duplication behavior.
Only after those layers are coherent does the statistical problem become well-posed.