At Advertising Week, a panel moderated by David Wells, Industry Principal, AdTech & Martech at Snowflake put a long-simmering industry tension front and center: identity is still core to powering activation and measurement — but the black-box model is holding teams back.
Across brands, publishers, agencies, and platforms, the message was consistent. The next phase of identity isn’t about improving single-vendor systems. It’s about owning the graph, composing it from multiple sources, and keeping control inside your data environment.
Composable identity = optionality, not abstraction
Uri Bushey, Head of Product at Narrative cut through the buzzwords by grounding composability in something buyers actually feel: optionality.
“For us, composability is optionality — on multiple vectors.”
That shows up in very practical decisions:
- Where identity data lives
- Whether identity is built from first-party, third-party, or both
- Which algorithms define “identity”
- How many providers you work with and when you switch
- Whether you want off-the-shelf algorithms, or you have a data science team with their own definition of identity
The takeaway: composable identity isn’t about flexibility for its own sake. It’s about adapting identity to the use case in front of you without rebuilding your stack each time.
Identity only works if you actually own it
Ian Maier, AdTech Lead at Hightouch emphasized a principle that’s becoming non-negotiable for sophisticated buyers: identity belongs with the data owner.
“Your data should be owned by you, your identity should be owned by you, in the place where it already exists—in your data warehouse.”
The problem with black-box identity isn’t just opacity. It’s also dependency. If you don’t build your own graph, it’s easy to get locked into a single provider. Or, if you push it into another platform, and now they kind of own your identity.
This framing shifts the role of third-party partners entirely. Providers become inputs, not gatekeepers — something you add, test, swap, or remove as needed, while the graph remains yours.
Publishers need interoperability that respects consent — not a “universal ID”
From the publisher side, Jenny Yurko, VP Data Product Strategy at Warner Bros. Discovery highlighted a reality often glossed over: identity fragmentation is structural, not accidental.
“Consumers go to different platforms in different ways, and different types of consents are applied. From an identity perspective that makes things difficult because combining those datasets is a difficult task. There’s not the same identity that comes from each platform.”
So, for publishers, composable identity isn’t about chasing a single ID standard. It’s about federation and translation, while staying compliant.
Composability enables an organization to anonymize that identity and work with PII in a compliant fashion while ensuring interoperability.
Advertisers want flexibility. Publishers need infrastructure that delivers it without breaking consent or governance.
The operational unlock: from one integration per year to hundreds
Jenny put the strongest business case for composability in simple operational terms. Under legacy models, each new ID or partner means months of work. With composable infrastructure:
“Instead of having to spend a year building one connection, I can have 100 connections right away. And I have the flexibility to move from one ID to another and fill the gaps.”
That’s the difference between identity as a bottleneck and identity as an enabler of growth, partnerships, and monetization.
Multi-vendor graphs exist for one reason: coverage and match rates
Jeremy Flynn, Head of Product at Horizon Media, grounded the conversation in agency reality: making first-party data usable in media while preserving privacy.
“We are responsible for making a brand’s first-party data addressable in media. At the same time we have to respect privacy and compliance…but that plays a role in match rates so the idea behind composable identity is how do you maximize match rates for activation purposes.”
Composable identity, in practice, is about filling real gaps. For example, in a retail environment maybe only 50% of purchases are tied to a loyalty program. That means the other 50% aren’t attached to a known person.
And the solution isn’t magic, it’s additive, testable identity inputs. Bring in incremental identifiers, increase the match rate, and increase reach.
Multi-vendor graphs aren’t about excess. They’re about closing coverage gaps without surrendering control.
AI needs normalization, not just more data
The panel closed on a forward-looking point that’s at the top of everyone’s minds. AI makes identity more powerful, but also more fragile if the data isn’t normalized.
Uri framed it as a semantic problem:
“You need a common understanding of the context behind the data — the semantic layer…It’s a column called ‘gender’ but what does that mean?”
Without consistent meaning across sources, AI can technically query data, but it can’t reliably understand it. You need to be able to take that data and turn it into something that the LLM is not only going to be able to query with SQL, but also understand the meaning behind consistently.
Jeremy added the buyer-facing requirement: transparency.
“We try to bring transparency — the breadcrumbs — into each step of that process.”
AI-driven activation only works if users can see what data was used, how audiences were built, and why decisions were made.
What you should take from this panel
If you’re evaluating identity today:
- Brands should demand ownership of the graph, with third-party identity as modular inputs — not permanent dependencies.
- Publishers and retail media networks need consent-aware interoperability that scales to advertiser demands without bespoke builds.
- Ad and martech platforms win by delivering orchestration and usability on top of identity — not by owning it.
Or, as the panel repeatedly reinforced: the future of identity isn’t a better black box. It’s a graph you own, with normalized data composed for the use case at hand, and built to change.
Sourced from Narrative.io [https://www.narrative.io/blog/breaking-free-from-black-box-identity]
February 13, 2026