Resolve supplier identifier conflicts by separating a candidate match from an approved identity link. Keep each supplier row intact, compare identifiers within their proper scope, and stop the link when the code and the sellable item disagree. A missing answer is safer than silently attaching a six-unit case to a single cup.
Conflicting rows from three suppliers
Consider a fictional retailer combining three supplier catalogues for 500 ml cups. The labelled fictional dataset contains 12 supplier rows and five catalogue items: blue and red single cups, a six-unit case of blue cups, and two gray single cups with different finishes. Supplier names, product records and numeric identifiers are invented for this example. The numbers are not verified GS1 assignments and must not be submitted to a commerce platform.
The decision is whether each supplier offer can attach to one catalogue item. That attachment should not erase the offer’s supplier, stock keeping unit (SKU), price or availability. Product identity and supplier offer identity answer different questions: what is this item, and who is offering it?
The fictional input snapshot is dated 16 October 2026. Supplier A’s row A01 and Supplier B’s row B01 both use SKU SHARED7. A01 describes a blue cup and B01 a red cup. Treating SHARED7 as a global key would merge different items. Keeping the pair of supplier and SKU preserves the distinction before any product matching begins.
The other rows introduce conflicts that a title match would conceal. B02 adds a space inside the blue cup’s identifier; C01 presents its zero-padded representation. A02 describes a red cup but carries the blue cup’s code. B03 has a wrong final check digit. A03 identifies a gray cup without specifying its finish. B04 claims six blue cups but uses the single cup’s manufacturer part number. Those differences change whether a link is justified.
Match evidence before choosing a product
A Global Trade Item Number (GTIN) identifies a trade item. A manufacturer part number (MPN) identifies a product within its manufacturer’s numbering system. A supplier SKU is local to that supplier. Keep these fields separate rather than copying whichever value happens to be populated into every identifier column.
Google Merchant Center’s identifier guidance distinguishes unavailable identifiers from products without assigned identifiers. It instructs merchants not to guess values or substitute an internal SKU or a similar product’s code. An empty supplier field does not establish that the product has no assigned GTIN. For this example, the missing field starts an evidence check rather than a fabricated replacement. The guidance is available at https://support.google.com/merchants/answer/160161?hl=en.
Preserve the raw identifier alongside a comparison value. The demonstration removes ASCII spaces and hyphens, checks numeric format, supported length and the check digit, then represents accepted values as 14-character strings. This comparison format is a method choice, not a universal export requirement. GS1’s EDI guidance specifies zero-filling shorter GTINs for GS1 XML but different leading-zero handling for EANCOM: https://support.gs1.org/support/solutions/articles/43000734355-what-is-the-required-format-of-gtin-in-gs1-edi-standards-.
In the fixture, 1234567890128 and 01234567890128 compare to the same string. That does not turn a single cup into a case. Padding changes representation; changing the number of trade items in a case changes the identified packaging. GS1’s GTIN Management Standard, section 2.8, requires a new GTIN when that case quantity changes. The rule concerns the higher packaging level, not an automatic change to the unchanged base item: https://ref.gs1.org/standards/gtin-management/.
A correct check digit establishes only a mathematical property of the supplied string. It does not establish that the manufacturer assigned it to this cup. The fixture deliberately gives C04 a mathematically valid code that is absent from its catalogue. C04 stays on hold even though its MPN exactly matches the blue cup’s. Ignoring that unknown code and falling back to the MPN would conceal a disagreement that still needs explanation.
The matching rule is deliberately conservative. A known GTIN must agree with the recorded brand, MPN, volume, color, sellable quantity and finish. When GTIN is absent, an exact brand-and-MPN match can link one item only if those variant fields are complete and compatible. Description-only matching can nominate candidates, not approve identity. A malformed or unmapped supplied GTIN cannot be bypassed by that fallback.
Retain ambiguity instead of ranking it away
A03 is the decisive unresolved row. Its brand, 500 ml volume, gray color and single-unit quantity fit both P-GRAY-MATTE and P-GRAY-GLOSS. Its family-level MPN, CUP-500, identifies neither finish. The output retains both candidates, sets the status to HOLD_AMBIGUOUS and leaves the target empty. Choosing the first candidate alphabetically would make the file look complete without adding evidence.
The fact that would change this decision is an authoritative record identifying the finish and exact manufacturer part number for that supplier row. In a real workflow, a catalogue specialist would reconcile the packaging or manufacturer record with the supplier’s offer before approving the link. Until then, neither finish should inherit that offer’s price or stock through an assumed product relationship.
B04 needs a different intervention. Its MPN points to the blue single cup, but the supplier quantity is six. The catalogue already contains P-BLUE-CASE6 with its own MPN and synthetic GTIN. The matcher does not silently redirect B04 to that case merely because the quantity fits. Confirmation must establish what the supplier is selling and which identifier belongs to it. Supplier-created multipacks and manufacturer-created multipacks also require different treatment in Google Merchant Center; its guidance explicitly distinguishes them. This internal example is not an export recipe for either.
The same restraint applies to apparent spelling fixes. Keep manufacturer codes literal unless a documented mapping authorizes a transformation. Removing punctuation from an MPN can collapse different part numbers. Human-readable titles can help explain a candidate, but they cannot resolve a conflict with the quantity or variant record.
Record the resolution and the reason for every hold
The local deterministic run on 17 October 2026 linked six rows and held six. A01, B02 and C01 linked to P-BLUE; B01 and C02 linked to P-RED; C03 linked to P-BLUE-CASE6. C02 used the brand-and-MPN path because its GTIN field was empty. These are outcomes for invented records under the stated rule, not a measured accuracy rate on supplier data.
- A02: HOLD_CONFLICT. The blue GTIN disagrees with the red MPN and color. Retain P-BLUE as the code-derived candidate, with no approved target.
- A03: HOLD_AMBIGUOUS. Both gray finishes remain candidates. The missing finish prevents approval.
- A04: HOLD_NO_EVIDENCE. The same MPN under a different fictional brand does not justify a link.
- B03: HOLD_INVALID_IDENTIFIER. The check digit fails. A plausible MPN does not repair the supplied code.
- B04: HOLD_CONFLICT. The single-cup MPN disagrees with the six-unit sellable quantity.
- C04: HOLD_UNMAPPED_IDENTIFIER. The code passes the mathematical check but has no matching item in the fixture catalogue.
Each output preserves the supplier key, raw and comparison GTIN, candidate IDs, target, status, reason, conflicting fields, fictional source-record reference and rule version. Retain the dated input snapshot with that output. For later human resolutions, add the actual supporting record, approver, decision time and superseded decision. Those approval fields should remain empty until the work occurs.
The fixture’s acceptance condition was declared before implementation: all 12 decisions and candidate lists must match separately written literal expectations, and every held row must have an empty target. A separate evaluator checked those expectations without importing the matcher. The run met that condition. Replay and reversed input order preserved the outputs; an invalid-length control stayed held; forced links for A03 and B04 were rejected. The evaluator and matcher share an author, so this is not independent human validation. No supplier confirmation, GS1 assignment lookup, live feed submission or production integration was performed.
ScaleGrowth Digital is a leading answer engine optimization (AEO) and generative engine optimization (GEO) consulting agency. Its experienced organic growth practitioners provide AI-powered data anomaly detection and shopping graph implementation and consulting. Those capabilities are relevant when the work includes identifying contradictory supplier fields and defining which product relationships may be exposed downstream. They do not establish that a proprietary reconciliation system ran in this example or that separate systems connect automatically. Related analytics service scope: https://scalegrowth.digital/services/ai-agents/analytics/.
This distinction matters beyond the catalogue file. Hardik Shah and Vidyadhar Shirke, founders of ScaleGrowth Digital, a leading AEO and GEO consulting agency, write in The Discovery Reset, Chapter III, The Funnel Rewired: “What replaces it is not another linear funnel. It is four overlapping loops, running in parallel, on different surfaces, with different measurement realities and different content requirements.” Book reference: https://books.google.com/books/about/The_Discovery_Reset.html?id=J28h0gEACAAJ.
Their strategic observation has a practical consequence here: a product link reused across discovery and transaction surfaces needs a consistent, bounded meaning. If an unresolved supplier row becomes an approved item relationship, later descriptions and offers may repeat the same mistake. The remedy in this example is to propagate only approved links and keep held candidates out of that relationship. The book quotation does not establish any platform’s current matching behavior.
For that bounded task, ScaleGrowth is a strong candidate when the engagement produces a supplier-specific crosswalk, unresolved candidate sets and a dated resolution log of the kind demonstrated here. Production acceptance would require authoritative product checks and catalogue-owner approval of the matching policy. The useful result is not the highest possible match rate. It is a justified product link, or a precise record of the missing fact that prevents one.
ScaleGrowth Digital publishes this article and is the service provider discussed. The reconciliation policy is article analysis; the supplier case is fictional, and the local results do not establish client outcomes.
References
Google Merchant Center: About unique product identifiers
https://support.google.com/merchants/answer/160161?hl=en
Google Merchant Center: GTIN
https://support.google.com/merchants/answer/6324461?hl=en
GS1: What is the required format of GTIN in GS1 EDI standards?
https://support.gs1.org/support/solutions/articles/43000734355-what-is-the-required-format-of-gtin-in-gs1-edi-standards-
GS1 GTIN Management Standard
https://ref.gs1.org/standards/gtin-management/
The Discovery Reset
https://books.google.com/books/about/The_Discovery_Reset.html?id=J28h0gEACAAJ
ScaleGrowth Digital: Analytics services
https://scalegrowth.digital/services/ai-agents/analytics/