The wrong answer to whether you should run an RFP shows up as wasted time. A full RFP for a $40K project is over engineering; a sole source for a $400K project is under engineering. This framework is five tests scored simply; if three or more tests come back yes, run the RFP. If two or fewer, run a lighter procurement. The tests work for private sector procurement and for public sector where policy thresholds set the floor.
Total contract value above the spend threshold? Federal: simplified acquisition threshold ($250K in 2026 for routine procurements; lower for some categories). State / local: typically $50K to $150K. Private sector: typically $100K annual or $250K one-time. Above the threshold this test scores yes. Below it, lighter procurement is appropriate unless one of the other tests overrides.
Does the procurement require multiple valid solution approaches, deep technical evaluation, or structured comparison across more than 3 vendors? Procurements with complex scope (integration with 3+ systems, multiple service lines, multi-year delivery) score yes. Procurements with a fixed specification (commodity software licence, standard maintenance contract) score no.
Is there genuine variability in how vendors will propose to solve the problem? Build vs buy, on-premise vs cloud, in-house vs outsourced, fixed-fee vs T&M. When the right answer is unclear, the RFP is the structured forum to compare approaches. When the approach is settled, an RFQ on the specified approach is lighter and sufficient.
Does the procurement need to demonstrate fair process to multiple stakeholders? Board oversight, audit committee, donor or grant reporting, public-sector transparency. When the process needs to be auditable beyond the procurement team, the RFP is the documented audit trail. When the procurement is straightforward internal sourcing, the audit-trail value is lower.
Is the procurement high-risk on dimensions other than cost? Vendor lock-in for a multi-year period, security or compliance criticality, large-scale change management, public reputation exposure. High-risk procurements benefit from the structured discovery the RFP provides; low-risk procurements do not justify the overhead.
Score each test as yes or no. Three or more yes scores = run the RFP. Two or fewer = run a lighter procurement (head-to-head bake-off, sole source with cost discipline, or vendor questionnaire). The exception is the spend test in public-sector contexts where the threshold is policy-mandated; above the threshold, the RFP is required regardless of the other tests.
Decision frameworks fail when applied mechanically. Each of the five tests has a counter example where the score points to the wrong answer.
Large spend on a commodity purchase (e.g., $500K bulk-licence renewal for Microsoft 365). The spend test says run RFP; the procurement reality is the licence rate is published, the contract is templated, and an RFP produces no incremental discount. Use the published-discount channel (CSP / EA route) and document the cost-reasonableness analysis instead.
Complex procurement where only one vendor can perform (proprietary technology, unique expertise). The complexity test says run RFP; the market reality is there is nothing to compete. Use sole source with a defensible justification (see the deep dive on sole source). Running an RFP that you know only one vendor can win wastes everyone's time.
The approach is settled; the buyer has decided cloud + SaaS + per-user pricing. The multiple-approaches test says no, run lighter procurement. But if the spend, fairness, and risk tests all score yes, the buyer should still run an RFP. The approach being settled means the RFP scope can be tighter, not that the RFP is unnecessary.
Internal procurement with no external stakeholder oversight, but a powerful internal stakeholder (a function head, a board member) has a preferred vendor. The fairness test scores no on external oversight, but the internal fairness consideration may still warrant the documented competitive process to defuse the politics. Procurement governance is about internal trust as much as external compliance.
Genuinely low-risk procurement (replacing one commodity vendor with another commodity vendor; existing infrastructure unchanged; low business impact). All five tests may individually score no but the procurement policy may still require an RFP above the spend threshold. Policy compliance overrides the test framework in regulated industries and public-sector contexts.
When the decision is to skip the RFP, three lighter procurement modes are available. Each has its own discipline. Skipping the RFP does not mean skipping procurement rigour; it means matching the rigour to the procurement size and risk.
Informal pitch with 2 to 3 known vendors. Each delivers a 30 to 60 minute pitch; provides a one-page proposal; supplies 2 to 3 references. Decision in 1 to 2 weeks. Documented in a short award memo (one page). Works for procurements where the vendor universe is small and the buyer already knows the candidates.
Single vendor selected without competition. Cost discipline applied via independent cost estimate, published catalogue pricing, or should-cost analysis. Documented in a sole-source justification (see the deep dive). Works when there is genuinely only one viable vendor and the procurement governance accepts the justification.
Single vendor completes a structured questionnaire covering technical approach, pricing, references, terms. No competitive scoring but the structured response surfaces information that a sole-source conversation alone would miss. Works when the vendor is selected but the buyer wants the documented structure for governance purposes.
Spend executed against an existing master services agreement, GSA schedule, framework agreement, or cooperative purchasing contract. No new procurement; the pre-negotiated agreement covers the terms and pricing. Works when the relevant master agreement exists; saves substantial procurement time.
For the sole-source deep dive, see sole-source vs competitive RFP; for the comparison, see RFP vs RFI and RFP vs RFQ vs RFI.