What Will an Advertising Agent Be Allowed to Believe? Part 2

By Arian Senior, Director, Operations & Innovation at Utiq

An advertising agent can compare proposals at speed, but that comparison is only as reliable as the audience measures behind it. The challenge becomes especially clear when a buyer asks for incremental reach across several publishers.

The reach problem does not disappear

Consider a buyer asking an agent for incremental reach among a defined audience. Publisher A forecasts ten million people, Publisher B eight million, Publisher C six million. A human planner immediately asks: how much overlap is there?

If every publisher counts on a different identity system, the agent cannot add the numbers or reliably determine incremental reach. In direct deals, that overlap cannot be corrected later: once the volume is agreed, it is paid for in full. And the problem grows when the plan spans web, apps, CTV and audio. The audiences overlap in the real world. The systems measuring them may not.

Clean rooms, independent measurement and probabilistic techniques can all help. A common deterministic identifier offers something different: a common unit for counting the recognised population across sellers. It does not make the entire campaign deterministic. It does something more modest and more useful: it makes the part of the campaign covered by that identifier directly comparable.

Where Utiq fits

Utiq was built for this. The Utiq ID is derived from a verified telecom network signal and explicit user consent: no consent, no Utiq ID. It is not the third-party cookie under another name, but a durable, responsible, consent-based signal where users and publishers choose to take part.

It is already in market at scale: 90 million Utiq IDs across five European markets and 400 premium publishers. And because it is anchored in the network connection rather than in the browser, it already extends beyond the web – to mobile and tablet apps, and to CTV, where it is live on broadcasters’ own CTV properties, with FAST channels next.

For an agent, that changes what can be known:

• A common unit across sellers. Where sellers count audiences and forecast reach in Utiq IDs, their proposals count the same thing.
• Reach counted, not modelled. Wherever a campaign runs programmatically, the platform executing it sees the same identifier across every seller whose inventory carries it, so deduplicated reach and frequency are counted rather than estimated.
• Overlap before the buy. A neutral party holding a common key across publishers could compute the overlap between sellers’ audiences before the volume is committed, returning only aggregates and never user-level data. That is precisely what direct deals lack today.

For publishers, first-party audiences remain their own asset: a common identifier adds an interoperability layer on top, without exposing user-level data.

And it gives a buyer agent a question it can reason about:

How much of this proposal’s reach is measured against a common deterministic population, and how much depends on modelling?

Proposals should be measurement-readable

OpenProposal makes proposals machine-readable. The next step is to make them measurement-readable. Beyond inventory, price, audience and volume, an agent should be able to read:

• how the audience was defined;
• which identity methodology underpins the forecast;
• how reach was calculated, and whether it is deduplicated;
• what proportion is deterministic and what proportion is modelled;
• across which environments, and over what period;
• and what measurement limitations remain.

This is not about requiring one identity. It is about requiring disclosure of the unit behind the number. If two sellers both claim ten million people, the agent should not have to assume they mean the same thing. It should be able to ask.

Four layers, not one

In my view, the risk is not that agents will automate too little, but that they will optimise imperfect signals and move money faster than humans can intervene. The agentic stack needs more than protocols. It needs standards to communicate, identity to establish comparability, measurement to establish evidence and governance to establish trust.

No single company can provide all four, and no single identifier should be expected to. But identity is one of the foundations.

What matters is not which identifier wins. The future of agentic advertising is not machines processing more proposals. It is machines that can understand why one proposal deserves more confidence than another.

The next competitive advantage in agentic advertising may not be better optimisation. It may be better evidence.

That is where identity moves beyond the bid.

 

Read Part 1 of Arian’s essay here on Utiq.com