How customer deduplication works
Customer deduplication is the process of finding records that represent the same customer and grouping them under one resolved profile.
In DinMo, deduplication is handled by Profile Resolution. The result is a canonical profile spine keyed by dinmo_id.
Deduplication flow
DinMo deduplicates records in five steps:
Select profile models Choose source models that represent customers, users, contacts, accounts, or subscribers.
Define identifiers Decide which logical identifiers can be used to match records, such as email, phone, user ID, customer ID, or name attributes.
Map model fields Map physical source fields to those identifiers in each model.
Apply matching rules Use exact and, when appropriate, anchored fuzzy matching rules to decide which records merge.
Create the golden record Emit one golden record per resolved profile.
Exact and fuzzy matching
Exact matching is best for stable identifiers such as user ID, customer ID, standardized email, or normalized phone.
Fuzzy matching can help when secondary values contain small variations, such as names. Fuzzy criteria should be anchored by at least one exact criterion in the same rule.
Guardrails
Deduplication must stay conservative. DinMo uses guardrails such as:
identifier standardization
unresolved records
rule-level monitoring
audit tables
These guardrails make it possible to improve match coverage while reducing unsafe merges.
What to validate
Before consuming deduplicated profiles downstream, validate:
resolved profile count
match rate
unresolved records
conflict rate
rule contribution
identifier completeness
sample matched profiles
See Review and monitor for the validation workflow.
Last updated