How to Find New Suppliers Without Just Reusing Your Incumbent List

What is supplier discovery, and how do you find new suppliers?

Supplier discovery is a controlled search that expands and documents the candidate set for one category. It answers two questions: which organizations might plausibly meet the need, and why did each one enter the candidate set? It stops before the buyer decides that a supplier is qualified, invites it to compete, scores a response, or awards business.

That boundary mirrors the NIST MEP sequence: a request is distributed for a capability search and the results are summarized for the requester. The NIST playbook leaves the requesting organization responsible for vetting identified manufacturers. The same division of labor is a sound default for enterprise discovery: produce a traceable longlist with evidence gaps, not an automatic shortlist.

Why do incumbent supplier lists persist?

Familiar suppliers carry observed performance, known contacts, approved specifications, and established operating routines. Alternatives carry uncertainty. Wagner and Friedl model how incomplete information about an alternative can make switching inert, while Monteverde and Teece show why transaction-specific skills can be costly to transfer.

Those mechanisms make an incumbent preference potentially rational, especially where quality failure, requalification, tooling, or continuity risk is material. They do not prove that the existing roster is complete. Discovery is useful when it tests the market evidence while keeping the incumbent and the alternative under the same written criteria.

What are the seven steps of supplier discovery?

1. Set the decision boundary. Name one category, geography, time window, buying entity, and intended handoff. State that discovery ends before qualification and award.
2. Freeze the starting roster. Export incumbents, previously qualified suppliers, referrals, and known exclusions. Record the source and date so later additions are measurable.
3. Write criteria before searching. Separate must-have capability, capacity, certification, delivery, geography, and risk requirements from preferences. The NIST playbook uses a structured form to collect detailed technical information; the same discipline prevents criteria from drifting toward a favored name.
4. Search distinct channels. Draw candidates from more than one channel so coverage never depends on a single list: internal ERP and accounts-payable vendor masters; public programs such as the NIST MEP Supplier Scouting network, SAM.gov, and state manufacturing-extension partners; and commercial directories such as ThomasNet, Kompass, and Dun & Bradstreet. Add industry accreditation directories, trade associations, trade shows, and referrals from current suppliers and peers. Log the channel, query, date, and candidate source for each.
5. Normalize and de-duplicate. Resolve legal names, domains, parent companies, locations, and duplicate referrals. Keep the raw record so no discovery is silently erased.
6. Desk-screen the candidate set. Apply the same must-have screen to incumbents and new candidates. Mark pass, fail, and unknown separately; do not call a desk-screen pass qualified.
7. Produce and archive the handoff. Deliver the desk-screened longlist, evidence gaps, and candidate origins to the separate qualification process. Preserve the incumbent roster, raw leads, exclusions, criteria version, and discovery decisions for the next refresh.

A new category search produces large volumes of supplier data—capabilities, certifications, locations, duplicates. Guida, Caniato, Moretto, and Ronchi studied twelve provider cases involving AI-based scouting solutions and frame supplier scouting as a high information-processing need. That supports structured data capture and tool assistance; it does not show that automation makes a candidate qualified or produces a commercial outcome. Keep the criteria, the evidence, and the human decision visible even when parts of the search run through software-assisted scouting.

Worked example: read the candidate funnel

The table is useful because each rate diagnoses a different constraint. A low distinct-candidate count may point to narrow search coverage; a steep desk-screen drop may mean the search ignored must-have requirements; a qualification drop may expose missing evidence or real risk; and weak response coverage may require a better invitation or event design. None of those rates is inherently good or bad. Compare categories only when scope, criteria, channels, and denominators are defined the same way, then decide whether a competitive-bidding process is warranted.

Which measures reveal a narrow candidate set?

• Source concentration: candidates from the largest single channel divided by distinct candidates. A high share is a prompt to inspect coverage, not a universal failure threshold.
• Overlap rate: duplicate or already-known leads divided by raw leads. This shows whether additional channels add new names.
• Desk-screen yield: candidates passing must-have criteria divided by distinct candidates. Keep failed and unknown evidence separate.
• Qualification yield: qualified suppliers divided by desk-screen passes. This locates the gap between plausible capability and verified readiness.
• Response coverage: valid responses divided by invited suppliers. Interpret it with invitation timing, requirements, and supplier feedback.
• Decision traceability: invitation decisions with a recorded reason divided by all invitation decisions. Missing rationale is an audit problem even when the final list is sensible.

When is keeping the incumbent the right choice?

Discovery should be able to end with the incumbent retained. The 2007 model allows no, partial, and complete switching under different information, competitive-reaction, and scale conditions. The older automotive evidence adds the concrete mechanism: transaction-specific know-how and the difficulty of transferring skills can make switching to an alternative supplier costly.

If alternatives cannot clear qualification, switching would disrupt operations, or the cost of learning outweighs the decision value, retention may be the defensible result. The audit question is not whether a new supplier won; it is whether alternatives were tested under predeclared criteria and the final decision has evidence.

Why cite switching research from 1982?

The 1982 study remains in the evidence record because it explains how transaction-specific know-how can raise switching cost, while the 2007 model examines information asymmetry and no, partial, or complete switching. The 2023 study adds a current view of AI-supported scouting through twelve provider cases.

These sources answer different questions with different methods. They do not form a time series, and the newer study does not invalidate the older mechanisms. The older work supplies the switching-cost mechanism; the newer work supplies the current tooling context.

The auditable handoff

The final discovery record should contain the frozen roster, criteria version, search log, raw and normalized candidates, source for each lead, desk-screen result, evidence gaps, and handoff status. Give suppliers clear requirements and next steps using the supplier participation guide. Keep discovery evidence separate from RFx scoring so a broad search cannot quietly become an award recommendation.

After handoff, append qualification and invitation outcomes as separately governed fields without rewriting the discovery record. On refresh, add new searches and sources, preserve prior exclusions, and explain every changed disposition.

How do AI agents change supplier discovery?