Customer identity resolution guide
How a CDP creates a persistent unified customer record
Identity resolution links records that belong to the same person. A persistent unified customer record then maintains that relationship, history and current state so every audience does not have to rebuild the customer view from scratch.
Why one person becomes several records
A customer can appear in a CRM as a lead, in an operational system as a buyer, in GA4 as a browser and in a media platform as a matched identifier. Email changes, household sharing, duplicated CRM contacts and inconsistent source fields add further fragmentation.
Without identity resolution, the same person can enter conflicting audiences or remain in acquisition after becoming a customer.
Identity resolution is the set of matching rules
Identity resolution decides which records belong to the same person. Deterministic rules use agreed identifiers and relationships, such as a customer ID, CRM contact ID, normalised email address or another stable key.
The rule should be visible, testable and appropriate to the source. A match is not automatically safe simply because two fields look similar. False merges can be more damaging than leaving uncertain records separate.
A unified record is more than a matched email address
- A stable internal customer identifier.
- The source records and identifiers linked to that person.
- Current lifecycle, product, lead or customer state.
- Relevant behavioural and outcome fields.
- Timestamps showing when a state or source changed.
- Suppression, eligibility and governance fields required by the use case.
Persistent means the record is maintained over time
An audience-specific view assembles only the fields needed for one campaign. A persistent customer record survives that campaign and is updated as new source data arrives.
Fractyl creates and maintains persistent unified customer records. Audience definitions operate on that maintained customer layer, allowing the same person to move between prospect, customer, lapsed and other agreed states without rebuilding identity for each activation.
Source precedence resolves disagreement
Two systems may disagree about status, contact details or customer value. The implementation therefore needs explicit precedence rules: which source owns each field, how freshness affects the decision and what happens when the conflict cannot be resolved automatically.
This work is commercial as well as technical. The source that sales teams call authoritative may not be the source that confirms a completed transaction or current eligibility.
Hashing protects identifiers in movement, but it is not identity resolution
Hashing converts an identifier into a consistent non-plain-text value before it is used for matching or activation. It reduces unnecessary exposure of the original identifier during movement.
Hashing does not clean the source, decide whether two records belong together, create permission or make the underlying use lawful. Those decisions must happen before an audience reaches a destination.
Quality checks should be built into the customer layer
- Monitor duplicate and unmatched record rates.
- Test high-risk merge and split scenarios.
- Track source freshness and missing key fields.
- Review lifecycle states that conflict across systems.
- Confirm audience members enter and leave for the expected reason.
- Keep opted-out, invalid and otherwise ineligible records governed throughout the flow.
Questions to ask any CDP vendor
- Does the platform maintain a persistent customer record or assemble a view only when requested?
- Which identity rules are deterministic, probabilistic or vendor controlled?
- Can our team inspect why records were linked?
- How are source conflicts, late data and identity changes handled?
- Can a mistaken merge be investigated and corrected?
- Which identifiers leave the customer-data environment and how are they prepared?
How Fractyl uses the unified record
Fractyl uses the maintained customer layer to build visible suppression, lifecycle and acquisition audiences. One saved audience can then synchronise across Google Ads, Meta, TikTok, DV360 and supported third-party data clean rooms on the cadence the use case requires.
The result is not simply a larger database. It is a governed customer state that can change how media treats a person.
