Comparing CTV identity and household matching methods

Choose the unit before choosing the match: a device for device-level delivery, an account for authenticated activity, or a household for household reach and frequency. Roku RIDA, Amazon Fire TV ID, LiveRamp RampID, Lotame Panorama ID and Teads Omnichannel Graph illustrate different places to start; none alone proves who watched an ad on a shared television.

Choose the counting unit

For customer targeting, frequency control or deduplicated reach, define whether the decision needs a device, account, household or person. Then compare the planned inventory’s usable signals, permitted use and treatment of unlinked impressions.

| Method | Starting signal | Named CTV example | Limit to consider |\n| --- | --- | --- | --- |\n| Device advertising ID | Identifier supplied under platform rules | Roku RIDA; Amazon Fire TV ID; Apple device advertising identifier | One TV can serve several viewers; availability and permission vary. |\n| Signed-in account | Authenticated service relationship | — | Accounts can be shared, and not all viewing is signed in. |\n| Direct record match | Common eligible attribute or agreed key | LiveRamp RampID | A direct match can still describe an account or home. |\n| Inferred household link | Permitted device or connection signals | Lotame Panorama ID; Teads Omnichannel Graph | Wrong and missed links can distort reach or frequency. |

LiveRamp RampID links offline and online data into persistent user profiles and is used for people-based marketing, including CTV. It is a better fit to assess for people-based linking than to treat as proof of who watched a shared TV.

Lotame Panorama ID uses IP addresses and other signals for household-level targeting, including in CTV. Teads Omnichannel Graph is a household and user graph described as spanning more than 50 million CTV devices and over 100 million users across 25 countries; those totals do not establish coverage of Australian campaign inventory.

“Deterministic” and “probabilistic” describe how a link is made, not a universal accuracy grade. A provider may combine direct account matches with inferred device links, so assess the method mix and validation at the identity level being sold.

Key Statistics on CTV Identity Providers (Global Context)

Compare the error that matters

For household frequency control, a missed link can allow repeat exposure across devices treated as separate homes. A wrong link can group different homes and suppress eligible delivery.

For an existing-customer exclusion, a missed link can leave a customer in prospecting; a wrong link can exclude a prospect. Set the campaign’s tolerance for each error.

A match rate needs a numerator and denominator. It might describe eligible records, devices or households, but it does not show which impressions can be bought, delivered or observed.

Count the same shared-TV impression according to the chosen unit: as device activity, as account activity only when an account is resolved, or as household activity when a graph links the device to a home. Keep household counts separate from person reach.

Check the buying signal and permission

Apple requires permission through App Tracking Transparency on iOS, iPadOS and tvOS 14.5 or later for covered tracking or access to a device’s advertising identifier.

Include the OAIC’s guidance on tracking pixels and privacy obligations in an Australian privacy review. Record the source attribute, matching party, recipient, advertising purpose, retention and handling of changed choices; assess whether a hashed email or IP address could be linked to an individual in the CTV flow.

OpenRTB 2.6 includes ID Match Method Guidance. In representative CTV bid requests, check whether device or account identifiers and an ID-match-method signal are present, how they are populated, and what each field and value means; a signal’s presence alone does not establish a household or person match.

For a proposed comparison, line up eligible inputs, matches, activated identifiers, delivered impressions, and linked and unlinked counts for the same period and the Australian inventory in the plan. Record the validation method; an exposure-to-outcome match is not evidence that the ad caused the outcome.