Definition

Entity resolution determines whether records from one or more data sources refer to the same real-world entity. In procurement, it helps reconcile supplier identities when names, addresses, identifiers, and system records differ.

The objective is not to make every similar name match. It is to establish defensible relationships while preserving meaningful differences between legal entities, locations, and corporate groups.

Distinguish identity from hierarchy

Two records can describe the same legal entity under different spellings. Two different legal entities can also belong to the same parent company.

Those are different relationships. Merging them indiscriminately can distort spend, risk, contract, and payment analysis.

VPP recommends maintaining explicit relationship types. “Same entity,” “same corporate group,” and “possible match” should not become one undifferentiated status.

Use more than name similarity

Names provide useful evidence but can be abbreviated, translated, misspelled, or shared by unrelated organizations. Additional attributes can improve the assessment when they are reliable and appropriate to use.

  • Identifiers: Relevant legal or business identifiers, with jurisdiction and validation status.
  • Addresses: Location evidence, recognizing shared sites and historical changes.
  • Names: Legal names, trading names, and documented variants.
  • Context: Source systems, contracts, products, and relationship history.
  • Ownership: Verified parent-child relationships kept separate from identity.
  • Review: Human assessment for ambiguous or high-consequence decisions.

Do not assume that one matching attribute is decisive in every context. The appropriate rules depend on data quality and the consequences of a false match.

A hypothetical supplier match

Assume three systems contain “Northstar Components LLC,” “North Star Components,” and “Northstar Logistics Inc.” The first two may represent the same supplier, while the third may be unrelated despite sharing a name element.

The resolution process compares supporting attributes and records the evidence. It should not group all three merely because a text-matching score is high.

If the first two are confirmed as one entity, their spend can be analyzed together under the approved identity. The third remains separate unless evidence establishes a different relationship.

The analytical result should not automatically change payment instructions or merge operational vendor records. Those changes require their own controls and authorization.

Manage false positives and false negatives

A false positive combines records that should remain separate. A false negative leaves records separate when they represent the same entity.

The costs differ by use case. A tentative analytical grouping may be reversible, while an unauthorized master-data merge can affect contracts, tax records, and payments.

Use confidence bands, review queues, and reversible mappings appropriate to the consequence. Evaluate performance on representative reviewed records rather than reporting only the percentage of records matched.

Preserve traceability as the data changes

Store source record IDs, match rules, evidence, reviewer decisions, and effective dates. Retain the ability to reverse an incorrect mapping without losing transaction history.

New acquisitions, name changes, and supplier reorganizations can change relationships. A match approved once should not be treated as permanent regardless of later evidence.

Use unresolved records as a data-quality worklist. Forcing every record into a resolved entity can hide the uncertainty that decision makers most need to understand.

Frequently asked questions

  • Is entity resolution the same as deduplication? Deduplication is a related use case; resolution can also connect records across systems and preserve distinct relationship types.
  • Can AI replace review? It can assist matching, but material or ambiguous decisions need appropriate validation.
  • Does a supplier-group match authorize payment changes? No. Analytical identity and payment authorization are separate controls.

Make the supplier view defensible

VPP uses AI-enabled intelligence to support clearer supplier and spend analysis. The value comes from evidence-backed identities that decision makers can inspect and correct.