What Is Product Identity?
Product identity is the resolved entity behind a product's many SKUs, MPNs, GTINs and supplier records. Not a SKU, not a GTIN, not a match score, and the prerequisite for write-back and trustworthy AI agents.
Product identity is the resolved representation of the real-world product behind its many IDs, names, supplier records, variants and documents. It is distinct from any single SKU, MPN, GTIN or match score: those are evidence for the identity, not the identity itself.
The same contactor arrives in four supplier feeds under four strings, three spellings of the manufacturer and two coil voltages. Product identity is the single resolved entity all of those records point at. Keeping track of it is the hard part of running a multi-supplier catalog.
Definition
Product identity is the resolved, persistent entity behind a product’s many identifiers, names, supplier records, variants and documents.
A catalog never has a shortage of identifiers. It has too many. One deep-groove ball bearing shows up as 6204-2RS in one feed, 6204 2RS in another, SKF6204-2RS in a third and as free text on a purchase order. None of those strings is the product. They are labels pointing, with varying reliability, at the same physical thing. Identity is that thing, made explicit and persistent so your systems stop treating four labels as four products.
Identity is not an identifier
Each common identifier is partial.
| Identifier | What it is | Why it is not identity |
|---|---|---|
| SKU | A supplier's or distributor's internal code | Two suppliers give the same product two SKUs; one supplier may reuse a SKU over time |
| MPN | The manufacturer's part number | Closer to identity, but written many ways and silently revised |
| GTIN | A trade-item barcode number | Often missing on industrial parts, and different for a pack versus a single unit |
| Description | Free text | Ambiguous and supplier-specific |
Identity sits above all of them. The SKU, MPN and GTIN are evidence for the entity. For the detail on each, see SKU vs MPN vs GTIN and one product, five part numbers.
Identity is not a match score
Product matching asks whether two records are the same and answers with a probability. That is useful, but a 94% is a one-time judgement about a pair that is discarded at the next ingestion. Identity is the durable result: a canonical entity that both records now resolve to, so the next time either string appears it lands on the same product without being re-adjudicated. A score compares two rows; identity remembers.
What a resolved identity carries
A product identity is not a winning row. It holds the attributes agreed across sources, the relationships to variants, packs, replacements and documents, and, on every value, the source and confidence behind it.
| Field | Value | Source and confidence |
|---|---|---|
| canonical | ABB AF16-30-10-13 | resolved, 0.98 |
| MPN | 1SBL177001R1310 | manufacturer, 0.99 |
| coil voltage | 100–250 V AC/DC | datasheet, 0.98 |
| variant of | AF range | manufacturer, 0.99 |
| supplied by | Supplier A, Supplier B | feeds, 0.97 |
Why identity comes first
You cannot write back to an ERP or PIM a value you cannot attach to a stable product; you would overwrite the wrong record half the time. You cannot let an AI agent answer “is this the right part?” when it cannot tell that two of its three candidates are the same product. Identity is the floor under write-back, deduplication, compatibility and agent reasoning. Resolve it once, with evidence, and everything above it gets more reliable instead of accumulating more guesses.
How identity is established
- Normalize the identifiers: strip separators, fix manufacturer aliases, standardize units.
- Match candidates with deterministic rules and probabilistic scoring (deterministic vs probabilistic matching).
- Decide using confidence thresholds: auto-accept, send to review, or reject.
- Persist the decision as edges in a product identity graph, including the rejections.
- Publish the result as a canonical product record.
Steps 2 and 3 together are entity resolution.
Common mistakes
- Using the GTIN as the key. Pack and single-unit GTINs differ; many industrial parts have none.
- Merging on a fuzzy match alone. A one-character difference can be a different coil voltage.
- Deleting the losing record. Keep it, linked to the identity, with its source.
- Not recording near-misses. Without a
NOT_SAME_ASedge, someone re-investigates next month.
Where Claro fits
Identity is the first building block of Claro’s Product Identity & Evidence Graph. Claro resolves supplier records to one identity, attaches provenance and confidence to each value, and writes the canonical record back to your ERP and PIM.
Related
Glossary
Product Identity Graph
How identity is stored as scored, persistent edges.
Glossary
Entity Resolution
The process that resolves records to an identity.
Glossary
Canonical Product Record
The operational record an identity is published as.
Glossary
SKU vs MPN vs GTIN
What each identifier is and is not.
Article
One Product, Five Part Numbers
Why industrial identity is hard.
Guide
Product Identity & Evidence Graph
The full model.
FAQ
What is product identity?
Product identity is the resolved representation of the real-world product behind its many IDs, names, supplier records, variants and documents. It is the entity that SKUs, MPNs and GTINs are evidence for, not any one of those identifiers.
Is product identity the same as a SKU or GTIN?
No. A SKU is an internal code that differs by supplier, and a GTIN is often missing on industrial parts or shared between a pack and its single unit. Product identity sits above these identifiers and is the entity they all point to.
What is the difference between product identity and product matching?
Product matching compares two records and returns a probability. Product identity is the durable result: a canonical entity both records resolve to, kept so future records land on it without being matched from scratch.
Why does product identity matter for AI agents?
An agent comparing offers or answering whether a part is the right one needs to know when two listings are the same product and when they only look alike. Resolved identity, with its evidence and near-miss history, gives it that.
Claro
See how Claro handles this in production
This concept is one piece of keeping a catalog trusted. See how Claro resolves identity, enriches missing attributes, and validates every update before it reaches your PIM or ERP.
Learn more