When One Person Is Not One Record

Why deduplication and unification are no longer the same problem

Article by:

For a long time, we treated identity as something that needed to be cleaned up.

Multiple records for the same person? That was a problem.
Different email addresses? Something to resolve.
Duplicate contacts? Merge them and move on.

This way of thinking made sense in a world where systems were simpler, data was smaller, and identity was expected to be singular. But that world is changing. What we’re seeing now is not just a technical shift. It’s a shift in how people exist across systems. And it’s forcing us to separate two ideas that have been incorrectly bundled together for years … Deduplication and unification.

Deduplication is about mistakes

Deduplication is familiar. It’s the process of identifying and resolving accidental duplicates inside a system. Two records that should be one. Two profiles that represent the same individual in the same context.

This is a data quality problem.

It belongs in the CRM.
It needs to be deterministic.
It needs to be governed by clear rules.

Same email? Merge.
Same person account? Update.
Conflicting records? Resolve.

The goal is simple: one person, one record, one source of truth.

And in many cases, that still holds. But only within a defined boundary.

Unification is about reality

Unification is something else entirely. It’s not about fixing mistakes. It’s about making sense of fragmentation that is often intentional.

A person might exist as:

  • A donor in one system
  • A volunteer in another
  • A parent in a third
  • An anonymous browser on your website
  • A subscriber under a different email

These are not always duplicates. They are partial identities.

Each one is valid in its own context. Unification doesn’t try to collapse them into one record. It tries to connect them. This is where platforms like Salesforce Data Cloud operate differently to traditional CRM systems. Instead of enforcing a single record, they use a model that is closer to a key ring:

  • Multiple identifiers
  • Multiple sources
  • Probabilistic matching
  • Context-aware linking

The goal is not a single record. The goal is a usable understanding of a person across contexts.

These are not the same thing

The problem is that many organisations still treat these as one. They try to use deduplication rules to solve unification problems. Or they use unification logic to justify poor data hygiene. Both approaches break.

Because they are solving different questions:

  • Deduplication asks: “Should these be the same record?”
  • Unification asks: “Could these belong to the same person?”

One is binary. The other is relational.
One is about control. The other is about interpretation.

The line is shifting from accidental to intentional

Here’s where it gets interesting.

Historically, multiple identities were often accidental. Now, they are increasingly deliberate. People are choosing to present themselves differently depending on context:

  • Work vs personal
  • Public vs private
  • Transactional vs relational
  • Anonymous vs known

They might:

  • Use different email addresses on purpose
  • Engage differently across channels
  • Avoid being stitched into a single profile

And in many cases, they are right to do so. This creates a new kind of pressure on systems.

Because now the question isn’t just: “How do we merge duplicates?”
It becomes: “When should we not merge?”

CRM holds the line

In this world, CRM still has a critical role. It remains the place where:

  • Legal identity is managed
  • Transactions are recorded
  • Permissions are enforced
  • Operational processes rely on certainty

This is where deduplication must stay strong. Because downstream systems depend on it being correct. CRM is not the place for ambiguity.

Data Cloud explores the edges

But outside of that core, we need a different way of thinking. This is where unification lives. Platforms like Salesforce Data Cloud allow organisations to:

  • Link identities without forcing a merge
  • Explore relationships between records
  • Activate segments based on patterns, not certainty
  • Maintain multiple truths without breaking the system

This is not about replacing CRM. It’s about extending it into a more human model of identity.

A more human system

If you zoom out, this is less about technology and more about people.

People are not single-threaded.
They are not static.
They are not always consistent.

They move between roles, needs and contexts. What we’re seeing now is systems finally catching up to that reality.

Deduplication still matters.
Unification now matters just as much.

But they need to be designed differently. Governed differently. Explained differently.

The real design challenge

The hardest part is not technical. It’s deciding:

  • When do we enforce a single identity?
  • When do we allow multiple identities to coexist?
  • When do we link them?
  • When do we keep them separate?

This is not a data problem. It’s a business logic and experience design problem. Because every decision changes:

  • How you communicate
  • What you personalise
  • What you assume about a person
  • And how much control they have over their own identity

Looking forward

As we move through 2026 and beyond, this distinction will only become more important. Organisations that get this right will:

  • Maintain clean, trusted operational systems
  • While still understanding people in a richer, more flexible way

Those that don’t will either:

  • Over-merge and lose nuance
  • Or under-connect and lose insight

The future is not one identity. It’s not unlimited fragmentation either. It’s something in between. A system that knows when to be certain,
… and when to stay open.