Quick answer
Identity resolution works in five sequential stages: data collection from every source, cleansing to remove errors and duplicates, matching using deterministic and probabilistic rules, unification into a single customer profile, and activation for downstream use in marketing, sales, and analytics. Each stage has its own failure modes. Most identity resolution problems trace back to the cleansing or matching stages.
Stage 1: Data collection
Every source of customer data feeds into the system. Website behavior, email interactions, purchase history, support calls, mobile app usage, physical store visits, social media engagement. Each source contributes different identifiers and behavioral signals.
Common failure mode: not collecting from all relevant sources. If a system is missing data from a channel, the profile is permanently incomplete.
Stage 2: Data cleansing
Removes errors, duplicates, and inconsistencies. Normalizes formats (uppercase vs lowercase emails, phone number formatting, address standardization). Fills in gaps where possible from third-party sources.
Common failure mode: garbage in, garbage out. Poor cleansing produces poor matches downstream.
Stage 3: Matching
Applies deterministic and probabilistic matching rules to identify which records belong to the same person. Deterministic matching first (exact identifier matches). Probabilistic matching fills the remaining gaps using behavioral and device signals.
Common failure mode: aggressive probabilistic matching that creates false positives. Two different people get linked into one profile.
Stage 4: Unification
The system consolidates matched records into a single canonical profile with a persistent identifier. That profile becomes the golden record for the customer.
Common failure mode: keeping too many alternate identifiers active. Downstream systems get confused about which is canonical.
Stage 5: Activation
The unified profile flows to downstream systems: , CRM, analytics, personalization engines. Every system now sees one customer instead of many.
Common failure mode: broken sync. Downstream systems fall out of date with the identity graph and revert to fragmented views.
What makes identity resolution hard
Data quality varies wildly across sources. Customers use different identifiers in different contexts (personal email vs work email). Devices and cookies churn constantly. Privacy regulations (, CCPA) add constraints on what data can be collected and linked.
Getting to 85 to 95% resolution accuracy is achievable. Getting to 99%+ is very hard and often not worth the marginal cost.
Common follow-up questions
How long does identity resolution take to implement?
Three to six months for a mid-sized business starting from scratch. Longer if data quality is poor going in.
Can I do identity resolution without a dedicated tool?
Basic identity resolution can happen inside a or marketing automation platform. Sophisticated identity resolution across many channels usually requires purpose-built infrastructure.
What does identity resolution cost?
Wide range. Basic tooling starts around $500/mo. Enterprise identity resolution platforms run into six figures annually.
Related terms
- Identity Resolution
- Deterministic Matching
- Probabilistic Matching
- Golden Record
- Customer Data Platform