To ID or Not to ID: That is the Marketing Question

While a certain Danish prince agonized over existential questions of being, today's marketers face their own philosophical dilemma: to pursue deterministic ID-level consumer identity or to rely on probabilistic approaches.

In the fortnight since I wrote about Publicis' acquisition of Lotame and their "Connect or Die" strategy, I've had fascinating conversations with CMOs, agency heads, and martech leaders about what's become quite a fundamental divide in how the industry approaches consumer identity. The battle lines are being drawn, and how they're resolved will reshape marketing for years to come. So this week, since many of those conversations weren't the best informed on consumer identity approaches, I thought this week to demystify it a bit.

When Publicis made their $1.5 billion investment to secure the world's largest independent identity graph outside the walled gardens, they placed a strategic bet on deterministic identity—the ability to recognize and reach specific individuals across touchpoints. With Epsilon and Lotame combined, they now claim profiles on nearly 4 billion individuals globally.

The response from competitors has been swift and predictable. "ID-level data is overkill," said one holding company executive in a conversation last week. "Probabilistic targeting gets you 90% of the way there at a fraction of the complexity."

This isn't just corporate positioning—it's a fundamental divergence in how we approach consumer identity that will determine winners and losers in the AI-powered marketing landscape.

Let's demystify these approaches once and for all:

Deterministic Identity is about knowing with certainty that User A on Website 1 is the same as User A on App 2. It relies on authenticated signals like logins, email addresses, phone numbers, or other persistent identifiers that directly link touchpoints to a specific individual. It's the marketing equivalent of recognizing someone by their fingerprint. You have what's called Personally Identifiable Information (or "PII") data on a range of specific individuals. As a brand you may own these yourself through your loyalty and CRM efforts, and you can then match your knowledge of people with the PII-level IDs that Publicis for example owns, in a clean room to protect the assets of both parties, to make a higher resolution view of a consumer to find and engage them in the world.

Probabilistic Identity is about making educated guesses that User A on Website 1 is likely the same as User A on App 2 based on behavioral patterns, device information, and statistical modeling. It's the marketing equivalent of recognizing someone by their gait, clothing style, and the places they frequent. This is done by stitching together what is observable about people and their behaviours without their PII from a range of 3rd party sources that are available to practically anyone who cares to rent the knowledge.

The debate isn't new, but what's changed is the stakes. In a world of cookie deprecation, privacy regulations, and AI-powered marketing, how we resolve identity determines what's possible with personalization, measurement, and marketing effectiveness.

Identity isn't just a technical concern—it's the foundation of modern marketing:

This last point is crucial. All the generative AI capabilities in the world won't matter if you can't correctly identify whom you're talking to.

Based on my conversations across the industry, I see four distinct strategies emerging:

Example: Publicis with Epsilon + Lotame

This approach bets big on deterministic identity as the foundation for everything else. By controlling the largest possible identity graph, these players believe they can deliver superior personalisation, measurement, and media efficiency.

The Claim: "We know exactly who we're talking to, so we can deliver precisely the right message at the right time."

The Challenge: Scale, privacy concerns, and the cost of maintaining massive identity graphs.

Example: WPP and many independent agencies

This approach argues that probabilistic models deliver "good enough" identity resolution without the complexity and privacy concerns of deterministic approaches.

The Claim: "We can get you 90% of the benefits at 50% of the cost and risk."

The Challenge: Accuracy limitations, especially across channels and over time.

Example: Performance-focused agencies specializing in platform ecosystems (and the walled gardens themselves with their emerging self-service tools for brands like Advantage Plus and P-Max)

These players go all-in on leveraging the superior identity resolution within major platforms like Google, Meta, and Amazon.

The Claim: "Why build what already exists within the platforms where the media dollars go anyway?"

The Challenge: Dependency on platforms, inability to connect cross-platform journeys.

Example: Brands like Nike, Disney, and others investing heavily in direct relationships

These players focus on building their own first-party identity assets through direct consumer relationships, loyalty programs, and owned platforms.

The Claim: "We're building direct relationships that give us sustainable identity capabilities no one can take away."

The Challenge: Limited scale beyond their own customer base.

Here's where we need to cut through the marketing spin: in reality, no approach is purely deterministic or probabilistic. Even Publicis, with their massive identity graph, must use probabilistic methods to extend beyond their authenticated signals. And even the most ardent probabilistic players rely on some deterministic anchors.

Perhaps the most interesting aspect of this debate is the privacy dimension. Critics of deterministic approaches often cite privacy concerns—"it's creepy to track individuals at that level." Yet the technical reality is often the opposite.Deterministic identity, when properly implemented, can actually enable more privacy-friendly marketing. When you know exactly who you're talking to, you can honour their preferences precisely, minimize data collection to what's necessary and mplement consent and governance more effectively.

In contrast, probabilistic approaches often require casting a wider data collection net to make their statistical models work. They might end up knowing less about more people rather than more about specific individuals.

If you're a marketer trying to navigate this landscape, here are four practical considerations:

Not every brand needs the same level of identity resolution. A luxury automotive brand with long purchase cycles and high consideration might benefit tremendously from deterministic identity. A mass-market CPG brand might find probabilistic approaches perfectly adequate.

In some categories, competitive advantage comes from identification precision. In others, creative excellence or media scale matter more. Be honest about where identity fits in your competitive equation.

Most brands have more first-party identity signals than they realize. Before investing in external solutions, take stock of what you already own—subscriber databases, loyalty programs, app users, authenticated website visitors, etc.

The beauty of modern marketing is that you don't have to choose a single approach. Run controlled experiments comparing deterministic and probabilistic targeting for the same campaign and measure the difference in performance.

While the industry debates deterministic versus probabilistic as if they're mutually exclusive, the most sophisticated marketers I know are pursuing hybrid approaches. They're building deterministic cores from their most valuable first-party relationships and extending reach through probabilistic methods.

The real winning approach combines:

Shakespeare's Hamlet ultimately concluded that action, despite uncertainty, was preferable to paralysis. For marketers facing the identity question, the answer is similarly clear: you need both approaches, thoughtfully applied.

The real question isn't whether to pursue deterministic or probabilistic identity—it's how to combine them effectively for your specific brand needs while respecting consumer privacy and preference.

The marketers who thrive won't be those who pick a side in this false dichotomy. They'll be those who understand the strengths and limitations of each approach and orchestrate them together into a coherent identity strategy.

After all, as another Shakespeare character famously said, "The quality of mercy is not strained." Neither should be your approach to identity.

Brandflow is written by Justin Billingsley, who has spent his career on all three sides of the industry's table: senior client, global agency leader, technology founder. First published 27 March 2025 in the Brandflow newsletter on LinkedIn.