Comparison

Exact match, variant and candidate substitute: three relationships to keep apart

How to type relationships between bearing records so that identity matching never silently merges variants or declares interchangeability.

Most catalog matching problems come from one word doing three jobs. “Match” is used to mean these two records describe the same product, these two products are closely related versions of each other, and this product could be used instead of that one. Those are different claims with different evidence requirements and different consequences when they are wrong.

This comparison sets out the three relationships, what each requires and the common mistakes that collapse them.

The three relationships

Exact identityVariantCandidate substitute
ClaimTwo records refer to the same manufacturer productTwo products share a base design but differ in a meaningful configurationOne product may serve in place of another for a stated purpose
Identity basisSame manufacturer and same full designationSame manufacturer, shared base, different suffix or configurationUsually different manufacturers or designs
Typical evidenceNormalized designation, manufacturer identity, matching source recordsSuffix or configuration attributes that differCompared attributes, scope of comparison, applicable technical review
Safe actionLink records to one product identityKeep separate products, link as relatedKeep separate products, show a scoped comparison
Risk if confusedDuplicate products, split stockWrong configuration shippedUnsuitable part used in an application

In the MyBearings draft record envelope these map to relationship types same_identity_candidate, variant and candidate_substitute, each with a status of proposed, reviewed or rejected and a list of source IDs.

Exact identity

Identity is manufacturer plus full designation. The full designation includes prefixes and suffixes that change the product: closure, clearance, cage, precision class, lubrication and similar configuration markers. What each suffix means differs between manufacturers, so a suffix should be interpreted with its manufacturer’s documentation, not by a cross-brand lookup.

Two records are an exact identity candidate when, after normalization of formatting only, manufacturer and full designation agree. Formatting normalization means things like case, whitespace, and separator characters, and only where the manufacturer’s convention says they are not meaningful.

Illustrative, synthetic values:

RecordManufacturer (raw)Designation (raw)NormalizedResult
Supplier A row 12Example Maker Ademo 101 sealDEMO-101-SEALSame identity candidate
Supplier B row 40EXAMPLE MAKER ADEMO-101-SEALDEMO-101-SEALSame identity candidate
Supplier C row 7Example Maker ADEMO-101DEMO-101Not the same: suffix missing

The third row is not an exact match even though it shares a base. It may be the open version, or it may be a supplier that truncated the designation. The record must stay unresolved until evidence decides which.

Variant

A variant shares a base design with another product but differs in a configuration that matters to the buyer. Open versus sealed, different internal clearance, different cage material: each produces a different product with its own identity.

Variants should be linked so users can navigate between them, and kept separate so stock, pricing and specifications never merge. A good test: if shipping one instead of the other would generate a return or a complaint, they are variants, not duplicates.

Candidate substitute

A candidate substitute is a product from the same or another manufacturer that has been compared against a reference product on a stated set of attributes. The word candidate is deliberate. Matching published dimensions and configuration supports a shortlist; it does not establish that one product can replace another in a specific application. Load ratings, speed limits, materials, tolerances, lubrication and application conditions can all matter, and some of them may not be published consistently.

A responsible substitute record lists:

  • the attributes compared and their values on both sides,
  • the attributes that differ,
  • the attributes that could not be compared because evidence was missing,
  • the source and date for each value,
  • who or what proposed it, and whether anyone qualified has reviewed it.

Negative examples

These are the patterns that cause most false merges. Illustrative, synthetic values.

PairWhy it looks like a matchWhy it is not an exact identity
DEMO-101-OPEN / DEMO-101-SEALSame base, same dimensionsDifferent closure; variants
DEMO-101-SEAL (Example Maker A) / DEMO-101-SEAL (Example Maker B)Identical designation stringDifferent manufacturers; designation spaces are not shared
DEMO-101-SEAL / SYN-77-S (same dimensions)Bore, outside diameter and width agreeDifferent products; at most a candidate substitute
DEMO-101-SEAL / DEMO-101-SEAL-CPrefix of one anotherExtra suffix changes configuration until shown otherwise
SUP-4410 / SUP-4410Same supplier SKU in two filesSKUs can be reused or mean different packs; check the manufacturer link

Identification is not interchangeability

Part identification connects a record to a manufacturer product and documents the evidence and uncertainty. It answers “what is this?” It does not answer “can I use this instead?” Systems and interfaces should keep the two questions visibly separate. A resolver that returns a substitute when asked for an identity, or that hides missing comparison fields, creates risk for everyone downstream.

How to store decisions

  • Never merge records destructively on a proposed match. Link them, and keep the originals.
  • Record relationship type, status, evidence and reviewer for every accepted link.
  • Allow rejection. A rejected candidate is useful information that prevents the same false match from being proposed again.
  • Re-evaluate when source data changes, such as a designation being superseded.

Apply it

Methodology article using synthetic examples only; it states no manufacturer specifications. Independent technical review: not yet assigned. Found an error? Tell us.

Apply it

Related resources

Bring one messy category.
Leave with a clear next step.

Share what your catalog looks like today. We will tell you what a scoped assessment would cover, what it would not, and what we would need from you.