Supplier Evidence Card

The problem in supplier due diligence is often not the lack of an “AI score.” It is that different sources answer different questions . A registry page identifies a legal entity; a marketplace storefront markets capabilities; trade data shows records within a particular coverage window; a supplier reply states its position. Compressing all of that into “good” or “bad” erases the conditions that matter.

This card draws on recurring problems in my supplier records, background checks and trade-data reviews. The public template contains no actual client or supplier material. It helps organize what is observed, what remains unconfirmed and what to ask next. It cannot replace a site audit, legal advice or purchasing approval.

When to use it

• You have a new supplier candidate and need to decide whether further RFQ or review is worthwhile.
• A marketplace store, legal entity, quotation header or export entity uses different names.
• You find trade records and want to know what they prove—or find none and do not want AI to declare “no exports.”
• Your shortlist has due-diligence reports, but raw facts and risk judgments are mixed together.

Gather four evidence bundles first

1. Identity: legal name, former names, registry identifier and public record; mark anything missing.
2. Capability: product page, process description, equipment or qualification material and written supplier replies. Distinguish self-description from independent support.
3. Trade: data source, search date, matched name, market and time coverage, and traceable records. A total count alone is not enough.
4. Requirement: the product, process, approximate scale and verification requirements you actually need. Remove client names, drawings and prices; use only material you are authorized to process.

Four states that must stay separate

Copy the prompt

Begin with the fictional exercise below. For real work, use only authorized, de-identified information. Do not paste personal IDs, client drawings, prices, accounts or nonpublic contacts.

You are an evidence organizer for sourcing research, not the final approver. Work only from the excerpts I provide. Do not browse to fill gaps or invent entities, trade counts, customers or certifications.

Task: produce a “Supplier Evidence Card” for one candidate. First identify which entity each record refers to, then separate facts, leads, gaps and conflicts.

Input may include: the purchasing requirement; numbered excerpts from registries, stores, product pages and trade sources; source dates and coverage. If a field was not supplied, write “not provided,” never “none.”

Output in this order:
1. Entity map: legal entity, former names, store name, quotation header and trade-record name. Give an evidence ID for each link; keep identities separate when support is insufficient.
2. Capability matrix: for each requirement use “direct support / self-description only / pending / conflict.” Cite the evidence ID and excerpt. Business scope, a similar product photo or a trade description alone cannot prove the specific process or current capacity.
3. Trade signals: state the source, query date, market, window and name-matching method. Distinguish “not found in this search” from “no trade activity.”
4. Risks and gaps: keep “record found,” “not seen in reviewed public source,” “not checked” and “source coverage limited” separate. Do not output an overall good/bad score.
5. Next questions: at most 8, ordered by how much the answer could change the sourcing decision; say who should provide which evidence.
6. Human review: name the 3 places where a person most needs to inspect the original and state what decision can and cannot be made now.

Do not resolve conflicts by choosing a convenient source. Label inferences. Never fill missing data with common-sense guesses. End with: this card supports initial screening only, not factory certification, a legal conclusion or purchasing approval.

Fictional exercise: why a pass stamp would be wrong

All names, excerpts and IDs below are fictional.

Requirement: a factory for custom metal display stands and export documentation.
R1 Registry excerpt: Seablue Products Ltd. lists metal processing in its business scope; no facility or equipment information supplied.
R2 Marketplace store: “Seablue Display” claims 15 years of export experience; the store entity is not shown.
R3 Product page: photos of metal stands, without material, welding method or load data.
R4 Trade search: an approximate English-name query in one market over three years shows no match; identity mapping has not been confirmed.

The useful answer is neither “the export claim is false” nor “this factory can make it.” The store is not yet linked to the legal entity; scope and images are leads only; the search name and coverage cannot establish the absence of other export records. Confirm identity first, then request evidence of the target specification, process and documentation.

Human review: six questions that could change the decision

1. How do the quotation header, license, payee and actual production address relate?
2. Is there item-specific evidence of the target process, such as a sample, process record or verifiable approval? Does the image match the model?
3. Does a trade record belong to the manufacturer, an export agent or a similarly named entity? How was the English name linked?
4. Which countries, years and record types does the data source cover? What can—and cannot—an empty result mean?
5. Were business anomalies, litigation and penalties checked in their original public sources? On what date?
6. Which three missing answers could change the decision to sample or visit the site?

Download input template · Download fictional exercise · Download prompt

This public template was rewritten from historical background-check and supplier-record experience. It contains no actual entities, clients, prices, trade details or contacts. Coverage and legal interpretation change: return to the original source before deciding.