Identity Resolution at Scale — Patterns and Anti-Patterns
Identity resolution looks simple in slides. At scale, it's where most customer-data platforms stumble. Three production patterns.
May 9, 2026·2 min read·By admin
At a glance
Topics
IAM · Marketing Cloud · Salesforce
Published
May 2026
Customer-data-platform vendors love to demonstrate identity resolution. Three records, two attributes, perfect match. At scale — millions of identities, dozens of source systems, evolving consent — the picture changes dramatically.
Pattern 1 — Probabilistic + deterministic, in that order
Probabilistic matching for the long tail; deterministic confirmation for high-confidence claims. Many CDP rollouts use one or the other; the production pattern uses both, sequenced.
Pattern 2 — Identity-graph-as-data
Don’t encode identity resolution rules in code. Encode them as data your governance team can review and audit. The graph itself becomes auditable.
Pattern 3 — Consent-aware resolution
Resolution that ignores consent is illegal under GDPR + CCPA. Production patterns enforce consent at every join — not as a post-process filter.
Identity resolution at scale is hard. The vendors will sell you on the slides. The patterns above are what make it survive production.