Customer Data & Identity
Unify fragmented customer signals into governed golden profiles.
Open interactive lesson and quizWhat you will learn
- Design deterministic match rules
- Explain confidence and source lineage
- Tune identity resolution safely
The same customer appears in sales, support and website data under slightly different identities. You must unify them without losing lineage.
Lesson 1
Deterministic identity rules
Normalized email and phone are strong identifiers. Name plus company can extend coverage, but should use an intentional confidence threshold.
Deterministic matching compares defined field values. In Customer 360, create a named rule, select email, phone, name and/or company, and choose exact or normalized comparison. All fields selected within a rule must be present and match. Separate enabled rules provide alternative ways to connect records. A name alone is not a dependable identifier.
For example, a Name + Company rule can distinguish two people with the same name at different businesses. Normalization helps with case and spacing but cannot prove that a shared phone or shared inbox belongs to one person. Save your rules, run resolution and inspect the resulting groups before relying on them.
Lesson 2
Golden profiles and lineage
Identity resolution should not destroy source records. A golden profile groups identities, records why they matched and retains source lineage for review.
A golden profile is a view of records believed to represent the same customer. It should let a reviewer trace the contributing contacts rather than erase their origin. Grouping contacts is different from physically deleting duplicates. Keep original identifiers when connecting another source so changes and mistakes can be traced later.
Inspect one resolved profile and compare names, emails and company relationships across its members. Ask whether each source really describes the same person. If an incorrect group appears, refine the rules and rerun resolution. Downstream sales and marketing decisions are only as trustworthy as the identities used to make them.
Lesson 3
Tune for the cost of a wrong match
Higher thresholds reduce false positives while lower thresholds improve match coverage. Use representative data and review duplicate groups before changing production rules.
A configured confidence score expresses your rule policy; it is not a measured statistical probability. Raising the minimum threshold excludes rules with lower configured scores. It does not make an ambiguous email or phone inherently more reliable. Also check which built-in rules remain enabled, because a broad built-in rule can still link records even when a new combined rule is strict.
Create a small evaluation set containing known duplicates and known different people with similar names. Compare groups before and after each rule change. Record false merges as well as missed duplicates. Prefer a cautious rule when an incorrect merge could expose information or distort customer communication.
Practice · Resolve a duplicate identity
- Import contacts with a shared normalized email or phone
- Run Customer 360 identity resolution
- Review confidence, source lineage and the golden profile