Organise

Entity Resolution & Record Linkage

One identity per real-world entity across every source system.

Runs matching logic over incoming records to decide whether they describe something you already know about. When they do, it merges them into a golden record under a canonical id. When they don't, it mints a new entity. The result is a single authoritative version of each person, organisation, product, or place — regardless of how many source systems have their own id for it.

Shape

Operational dimensions

Requires approval

Each output waits on a human decision.

Event-triggered

Fires when an upstream condition occurs.

High data gravity

Owns a system-of-record; expensive to migrate.

Two-way integration

Reads from and writes to external systems.

Inputs

incoming records from one or more source systems
match rules or trained matching model
existing canonical entity table with golden records
blocking keys / similarity thresholds

Outputs

canonical entity id per incoming record
golden record per entity (merged attributes + provenance)
match decisions with confidence and evidence
unresolved / low-confidence cases for human review

Mechanism

Decides which incoming records refer to the same real-world entity, merges them under a canonical identifier, and maintains the linkage as new records arrive.

Why this is a primitive

Cannot be decomposed: the match-merge-maintain-canonical-id loop over arriving records is a single operation. It does not author a schema (vocabulary-authoring), apply a classification (classification-application), or build edges to other entities (graph-instantiation). It answers exactly one question: 'are these the same thing, and if so, what's the canonical id?'

Where it shows up

Financial services — resolves customer records across CRM, core banking, and card systems so the 360-degree customer view isn't triple-counting the same person

Healthcare provider — links patient records across registration, billing, and EHR to eliminate duplicate patient files and surface cross-system clinical history

B2B SaaS — deduplicates company and contact records arriving from web forms, enrichment APIs, and CRM imports before feeding the pipeline

Retailer — resolves supplier records across procurement, finance, and logistics systems so a single supplier isn't held as three distinct vendors

Related primitives

Composes with

Controlled Vocabulary Authoring →Schema Normalisation →Knowledge Graph Instantiation →Reference & Citation Resolution →Classification Application →Data Asset Inventory & Lineage →

Alternative to

Reference & Citation Resolution →