A Creative ID Is Not an Impression

The subject here is the boundary between an identified CTV creative and a qualified impression. A persistent ID can tell several systems that they are discussing the same asset, but the Media Rating Council’s guidance asks a different question: what evidence shows that the ad was initiated, loaded, and began to render in a measurable viewing event? Keeping those questions separate helps a buyer read a cross-platform report without mistaking a clean join for proof that every attached event should count.

The Counting Boundary Happens at Playback

A creative identifier describes the asset. An impression is a claim about an event involving that asset. The first claim may be established through registration and transport: a registry gives the creative an identity, and the ad response carries that identity to the delivery environment. The second claim requires evidence about what happened after the request reached the viewing system.

MRC’s CTV guidance applies the core principles of digital video measurement where CTV video can be measured. The measurement flow must include client initiation, filtration for invalid traffic, and the requirement that the ad be loaded and at minimum begin to render. The render requirement applies after the stream has been initiated and post-buffering, when the ad itself begins to appear or play. These conditions draw a boundary around the event being counted.

Client initiation means that the measurement path cannot rely only on an ad being scheduled or an impression-like record being created on a server. The relevant viewing environment has to initiate the event in the measurement flow. Loading is a further step: the ad must reach the point where it is available to the player or delivery environment. Beginning to render goes further still. It asks for evidence that the ad itself started to appear or play, rather than evidence that a response was merely prepared.

The distinction is important because a single request can create several records. An ad server can return a response. A publisher can log a delivery attempt. A measurement vendor can receive a beacon. A player can then fail to load the media, stop before rendering, or render under conditions that the measurement system cannot verify. A persistent creative ID can connect those records to the right asset, but it cannot decide which state the viewer actually reached.

The counting boundary can be pictured as a sequence of states:

stateDiagram-v2 [*] --> IDENTIFIED IDENTIFIED --> INITIATED: Client starts ad request INITIATED --> LOADED: Ad loads after buffering LOADED --> RENDERED: Ad begins to render RENDERED --> ELIGIBLE: Required traffic treatment is satisfied RENDERED --> UNKNOWN: Traffic or evidence cannot be fully validated INITIATED --> EXCLUDED: No required load or render ELIGIBLE --> [*] UNKNOWN --> [*] EXCLUDED --> [*]

The diagram is deliberately stricter than a delivery funnel. “Identified” is the starting state for creative reconciliation, not the end state for measurement. “Rendered” is not automatically “eligible” if the traffic cannot be treated under the applicable invalid-traffic rules. “Unknown” is not the same as “invalid”: it represents activity that cannot be fully validated and therefore should not silently be included as valid.