Agentic era will stress test TV and streaming identity infrastructure
Analysis – The Future of TV: Global Series
The ‘agentic era’ will stress-test identity infrastructure across the TV and streaming industry, as the arrival of AI-driven, automated media systems increases the importance of this function.
That is according to a white paper from The Coalition for Innovative Media Measurement (CIMM).
It says: “In converged TV, identity is no longer merely a matching function. It is the connective tissue that allows AI agents to optimise responsibly, measure reliably, and personalise coherently.
“The agentic era does not eliminate identity complexity. It makes disciplined identity architecture indispensable.”
The paper finds that as media optimisation becomes more autonomous, identity quality directly impacts measurement accuracy, attribution, and decisioning.
“Incorrect linkages contaminate training data, distort attribution signals, and degrade future decision quality.
“In the agentic era, identity becomes a prerequisite for dependable automation rather than a supporting input,” the paper warns.
Identity signals as foundational inputs
“Automation systems ingest identity signals as foundational inputs,” it states.
“If identifiers do not align across channels, if records are incomplete, if household and person entities are conflated, or if linkage quality varies invisibly across environments, AI does not resolve those inconsistencies. It optimises against them.”
CIMM supports improvements and best practices in measurement and currency systems, the use and application of new metrics, and data collaboration. Focused on the US, its members include over 600 agencies.
In February, CIMM held an ‘Innovations in Identity Resolution Showcase’ with 100+ companies to review the latest developments. The white paper draws on the insights from that event.
You can download the paper (Identity Infrastructure 2.0: Innovation, Transparency, and Privacy-First Architecture in the US TV & CTV Ecosystem) here.
The 66-page report is about far more than just identity and AI, covering data quality issues, transparency, privacy-by-design, signal enrichment and much more.
Optimisation loops amplify errors
But on AI, it warns that when identity is fragmented, AI models learn from inconsistent entity representations, while optimisation loops amplify incorrect associations.
Measurement and personalisation also drift apart.
When identity is coherent but ‘stale’, AI optimises against outdated linkage, while campaign decisions reflect yesterday’s relationships.
“If identity refresh is overly aggressive, AI loses longitudinal continuity and model stability deteriorates,” it the report continues.
The paper provides a list of requirements for AI-ready identity, including persistence over time and consistency across environments (such as CTV, mobile, web, clean rooms, and analytics systems).
Quality becomes a board-level issue
The report predicts that identity quality will become a board-level issue in the agentic era.
“As automation scales, upstream weaknesses in identity resolution can propagate directly into financial outcomes, affecting pricing accuracy, media efficiency, and measurement credibility,” it says.
It also notes that when AI-assisted systems influence media investment decisions, organisations must be able to interpret the identity logic and signal quality underlying those outcomes in order to diagnose performance variance, manage risk, and maintain institutional trust.
The report section on identity accuracy and data quality says these remain foundational challenges.
Linkage accuracy across commonly used identifiers is materially lower than widely assumed, it warns.
This statement is based on recent studies undertaken by CIMM in partnership with Truthset and GoAddressable.
“These findings do not merely highlight isolated vendor variance. They challenge the prevailing assumption that large-scale identity graphs provide stable and measurement-grade linkages across households, devices, and demographic attributes,” it claims.
“If identity is to function as infrastructure, its accuracy must be understood, measurable, and defensible.
Higher uncertainty than acknowledged
“The empirical evidence suggests that the market has been operating with materially higher uncertainty than previously acknowledged.”
IP-to-postal and IP-to-email linkage are examples of where accuracy is lower than many in-market assumptions.
The paper warns that linkage inaccuracies are compounded when different links (e.g., device-to-person, person-to-household, household-to-demographic) are used together in a targeted campaign.
The paper discusses data quality as a catalyst for the structural redesign of the IDR (identity resolution) ecosystem, and outlines the necessity for deterministic grounding, particularly through authenticated subscriber relationships and ISP-level data where feasible.
It advocates transparent linkage scoring, enabling stakeholders to understand probabilistic confidence rather than relying on binary match flags.
Another measure to reduce identity resolution inaccuracies is standardised metadata, including harmonised timestamp definitions and freshness indicators.
Unlocking more TV inventory
This paper also outlines identity related opportunities to unlock more TV inventory for programmatic planning/buying, with linear TV a key beneficiary.
FAST would benefit from deterministic household anchoring – particularly through authenticated or subscriber relationships, it suggests.
“This can materially improve measurement-grade accountability relative to inference-based linkage.”
The white paper is positive about industry improvements, saying data and identity providers are responding to challenges.
“The ecosystem is shifting away from opaque, scale-first identity graphs optimised for match volume toward deterministic-first, privacy-grounded, enterprise-controlled systems optimised for measurement credibility and governance resilience.
“Those systems are increasingly designed to be interoperable across environments (cloud, clean rooms, direct integrations) and usable not only for activation and measurement, but as the identity substrate for AI‑driven workflows, as we enter the agentic era.”
Identity Infrastructure 2.0
CIMM defines the next phase of identity resolution as ‘Identity Infrastructure 2.0’, and this will include:
- Disciplined deterministic anchoring where feasible
- Explicit modelling of households, people, devices, and accounts
- Explainable linkage and documented provenance
- Interoperable signal translation across environments
- Enterprise-level governance and policy control
- Measurement-grade validation with clearly articulated failure modes.
The report says identity is moving towards being ‘industry infrastructure’ – similar to clearing and settlement rails in finance, numbering systems in telecom, or accreditation regimes in measurement.
This is because, “When identity is weak, the market does not simply perform worse. It becomes harder to price inventory, harder to compare outcomes, harder to audit claims, and harder to coordinate shared standards.”