Identity & Addressability
Part of Advertising identity and addressability
Checking the limits of cross-device identity claims
Audit cross-device identity claims by identity level, inventory coverage, graph quality, platform permissions and reporting gaps.
A cross-device identity claim is useful when it specifies who or what is linked, which inventory is covered and how link quality was assessed. “Recognises people across screens” might describe a person-level graph, a household link or identifiers associated with one account. Establish the unit before using the claim for reach, frequency or attribution.
Turn the claim into a specification
For the proposed Australian campaign, request:
| Claim to clarify | Evidence to request |
|---|---|
| Who is identified? | Definition of person, household, account, browser or device; treatment of shared use |
| Which screens are covered? | Eligible web, app and connected TV inventory by partner and buying route |
| How are links made? | Direct and inferred methods, and their shares in the activated graph where available |
| How good are the links? | Validation method, truth set, wrong-link and missed-link measures, dates and market relevance |
| How long do links last? | Refresh, expiry and handling of changed or deleted identifiers |
| What can be reported? | Delivered impressions, linked and unlinked populations, and unknown coverage |
IAB Tech Lab distinguishes consumer identifiers that may relate to individuals or groups within a household, and notes that they may be tied to devices or browsers depending on available data, such as logins. General channel support does not establish that a proposed Australian buying account can link the named inventory. Request evidence for the publishers, apps or deals in the media plan.
Cross-device identity claim verification checklist
- Claim to clarifyWho is identified?
- Evidence to requestDefinition of person, household, account, browser or device; treatment of shared use
- Claim to clarifyWhich screens are covered?
- Evidence to requestEligible web, app and connected TV inventory by partner and buying route
- Claim to clarifyHow are links made?
- Evidence to requestDirect and inferred methods, and their shares in the activated graph where available
- Claim to clarifyHow good are the links?
- Evidence to requestValidation method, truth set, wrong-link and missed-link measures, dates and market relevance
- Claim to clarifyHow long do links last?
- Evidence to requestRefresh, expiry and handling of changed or deleted identifiers
- Claim to clarifyWhat can be reported?
- Evidence to requestDelivered impressions, linked and unlinked populations, and unknown coverage
Check the unit and denominator
A household graph may combine devices at one address, but it cannot show which resident saw an ad. Shared tablets and accounts also complicate person-level claims. One person’s devices may remain unlinked.
If two people are wrongly joined, person reach may be undercounted. If one person’s devices are not joined, it may be overcounted. The effect depends on the report’s counting and modelling rules.
Ask how a unique entity is counted and what happens to impressions without a usable link. A cross-device rate calculated only for linked impressions describes that subset, not all delivery. Request linked, unlinked and total delivered counts for the same period and inventory, with unknowns identified.
Pros and cons of person-level vs. household-level cross-device graphs
- Pros: Person-level graphBetter targeting for individual users; supports accurate frequency capping per person; useful for attribution when tracking user actions across devices
- Cons: Person-level graphMay undercount reach if devices are not linked (e.g., shared tablets); risk of overcounting if two people are incorrectly joined; sensitive to privacy changes like Apple’s App Tracking Transparency
- Pros: Household-level graphAccounts for shared devices at a single address; aligns with Australian household-based advertising models; less affected by individual device changes
- Cons: Household-level graphCannot attribute ad exposure to specific individuals; may misrepresent reach if multiple residents use devices without login; limited utility for personalisation or conversion tracking
Separate the advertised uses
A graph that helps target an audience does not automatically observe every later impression. Deduplicating some impressions does not show that a sale or site action occurred, or that an ad caused it. For each use, identify the systems that pass a usable identifier and the events that can be joined.
For an incremental-reach claim, request the baseline, identity unit and deduplication method. For a frequency claim, establish whether the cap is enforced at person, household or device level, where it applies and how unmatched devices are treated. For attribution, distinguish an exposure-to-outcome match from evidence of a causal effect. A graph diagram alone establishes none of these results.
Check platform and privacy limits
Apple requires permission for covered tracking using data collected in an app across other companies’ properties. Without that permission, the app cannot obtain its advertising identifier for that use.
Platform rules differ. Check the relevant platform’s current documentation and ask how the proposed graph handles identifier changes and user choices.
A pilot can support conclusions only about the campaign configuration and inventory actually assessed. Missing validation or unobservable inventory limits the conclusion; it does not become a zero.
Key platform privacy requirements affecting cross-device identity
- Apple App Tracking Transparency (ATT)Requires explicit user permission to track across apps and websites; no permission = no access to IDFA
- Google Play Services (Android)Uses Advertising ID (AAID) but requires opt-in via Google Play policy; must comply with ATO guidelines on data handling
- Australian Privacy Principles (APPs)Organisations must obtain consent before collecting personal information via tracking pixels; disclose purpose and data sharing practices
- IAB Tech Lab GuidanceRecommends transparency in identity methods and validation of link quality using industry-standard truth sets
