Deterministic matching

Learn how deterministic matching links records, what guarantees it provides, and when it’s sufficient for identity resolution.

What deterministic matching means

Deterministic matching links records only when identifier values match exactly.

There is no fuzzy logic, similarity scoring, or inference involved.
If two values are not equal, the records are not matched.

Examples of deterministic matches:

• email = email
• user_id = user_id
• anonymous_id = anonymous_id

If values differ—even slightly—the records remain separate.

What deterministic matching guarantees

Deterministic matching provides strong behavioral guarantees:

• Predictability
  The same data and rules produce the same results.

• Explainability
  Every merge can be traced to a specific identifier match.

• Control
  You explicitly define which identifiers are allowed to link records.

Because of these guarantees, deterministic matching is well suited for analytics, compliance-sensitive workflows, and operational systems.

How records are merged

When Identity Resolution evaluates a record:

1. Identifiers on the record are checked in priority order
2. If a valid match is found, the record is merged into the matching identity
3. If no match is found, a new identity is created

Identifier priority ensures higher-confidence identifiers (for example, user_id or verified email) are evaluated before weaker or more ambiguous identifiers.

Why identifier limits matter

Exact matching alone can still cause incorrect merges when identifiers are shared, reused, or low quality.

Identifier limits act as guardrails by restricting how many distinct values an identity can contain for a given identifier.

Examples:

• Limiting user_id to 1 prevents identities from merging across users
• Limiting email helps control reused or reassigned email addresses
• Limits on lower-priority identifiers reduce over-linking without blocking core matches

Limits work together with priority to balance coverage and accuracy as the graph grows.

→ See Identifier rules for details.

When deterministic matching is sufficient

Deterministic matching works best when:

• You have stable, reliable identifiers
• Identifiers are consistently formatted
• Exact matches reliably indicate the same entity
• Accuracy matters more than maximum coverage

Common use cases include:

• Logged-in user experiences
• CRM and lifecycle messaging
• Revenue, attribution, and funnel analysis
• Account- or person-level reporting

When you might need more than deterministic matching

Deterministic matching may be too strict when:

• Identifiers contain typos or formatting differences
• Users interact across systems without shared IDs
• Data is manually entered or incomplete

In these cases, you can optionally enable probabilistic matching (beta) to supplement deterministic rules.

Deterministic matching always remains in effect; probabilistic matching layers on top rather than replacing it.

Next steps

• Learn how matching behaves internally in How Identity Resolution works
• Configure exact-match behavior in Identifier rules
• Explore Probabilistic Identity Resolution (Beta) if available