A customer data platform (CDP) is often presented as a way to unify customer data, but “unify” can mean different things: resolve identities, create audiences, coordinate activation, or provide a governed profile. Define the specific job first so a vendor demo can be judged against the organization’s actual data and operating constraints.

Map the use cases and existing system roles
List the decisions or activations that are difficult today. Identify which systems already hold the source data, which team owns each definition, and where a CDP would sit. If the problem is inconsistent account matching or missing consent fields, adding another profile store may not solve it by itself.
State whether the required identity is a person, account, household, or product user. B2B relationships can include subsidiaries, buying groups, and multiple products; a single golden record assumption may be too simple.
- Draw the source-to-activation flow for each use case.
- Identify latency and freshness requirements.
- Separate identity resolution from audience orchestration.
Test matching, governance, and failure cases
Ask how the platform handles ambiguous records, source conflicts, merges, deletions, and consent changes. Test examples with shared domains, subsidiaries, missing identifiers, and duplicate contacts. A match rate should be accompanied by an understanding of false matches and unresolved records.
Review permissions, audit logs, retention controls, data residency, and export behavior with the appropriate technical, privacy, and security owners. Confirm how a failed destination sync is detected and corrected.
- Use sample records with known expected outcomes.
- Trace preference changes to every destination.
- Measure time and cost to resolve data quality exceptions.
Compare total operating effort
Estimate implementation and ongoing work: connector maintenance, identity rules, taxonomy stewardship, audience QA, and user training. A platform may reduce manual transfers while increasing the need for specialist administration. Include both in the operating model.
Score vendors against the same evidence-based criteria and note what remains custom. If a simpler warehouse or CRM change can solve the use case, include that option in the comparison rather than assuming a CDP is required.
- Separate licensed features from custom services.
- Request a test plan and exit/export example.
- Set a measurable adoption and value review after implementation.
How did this article land?
Choose one reaction. You can change it anytime.
