An RFP evidence hierarchy is a claim-specific rule for preferring stronger support over weaker support. It ranks candidate evidence by authority, directness, scope match, recency, independence, reliability and approval, then sets the minimum acceptable level according to the consequence of being wrong. It is not a fixed list in which a certificate always beats an internal record. The strongest item is the one that credibly proves the exact proposition in the offered scope.

Bid teams often use “evidence” to mean any file that resembles support. A global certificate is attached to a claim about a service outside its scope. A dashboard screenshot appears without the period, population or calculation behind it. A sales deck repeats a capability from an old expert conversation. Each artifact may look persuasive while proving little. Under deadline pressure, the newest or most polished file wins, and teams apply the same burden of proof to a minor process description and a contractual availability promise.

Start with the claim, not the evidence folder. Define what an evaluator is asked to believe, which entity, product, geography, period and conditions it covers, and what happens if it is wrong. Set a required assurance level. Compare sources on explicit dimensions and use the strongest applicable combination. If proof falls below the threshold, narrow the claim, obtain approval for a qualified statement or remove it. Adding more weak sources does not create strong evidence.

Define what must be believed before ranking the proof

Break a broad sentence into atomic propositions. “Our certified platform reduces processing time by 40 percent and will meet 99.99 percent availability” contains at least four claims: certification, applicability of the certificate to the offered platform, a measured outcome and a future service commitment. Each has a different source, boundary and approval path. One logo or case study cannot inheritably prove the whole sentence.

Write a claim card with proposition, type, subject, entity, product or service, configuration, geography, population, measurement period, conditions and intended tender location. Record whether the wording describes current fact, past outcome, forecast, target or binding offer. Qualifiers such as “for the hosted service,” “during the 2025 reporting period” and “subject to the proposed dependency” are not decoration. They determine alignment.

Rate consequence before searching. A minor statement about meeting cadence may need an approved operating description. A data-residency statement, minimum staffing promise, regulatory assertion or service-level target can affect eligibility, contract exposure and delivery. Consequence is not rhetorical importance. A dramatic differentiator may create little obligation, while one annex line can create a serious commitment.

Claim types require different proof
Claim typeWhat must be establishedTypical weakness
Current capabilityAvailable function in offered scopeRoadmap treated as live
Quantitative resultPopulation, method, period and valueNumber without denominator
Causal outcomeCredible link between action and resultCorrelation presented as cause
CertificationEntity, scope, standard, issuer and validityLogo used beyond scope
Future commitmentAuthority, feasibility, dependency and wordingExpert optimism treated as approval
ComparisonComparable basis and current alternativesSuperlative without a defined market

Rank applicability and control, not document prestige

Evaluate each source on seven dimensions. Authority asks whether its creator owns or can attest to the fact. Directness asks whether it proves the proposition rather than a nearby fact. Scope match covers entity, product, geography, population and conditions. Recency checks whether it remains valid. Reliability examines accuracy, completeness and method. Independence considers whether a separate party tested the fact. Approval confirms permission to use the source and wording in this bid.

The dimensions prevent rigid source worship. An external audit report can provide high independent assurance but does not prove a control outside its examination boundary. An internal production record may be the most direct proof of a service metric if its data lineage and calculation are controlled. A signed expert attestation can resolve a precise architecture fact, yet it is weak support for a market-wide comparison or an outcome the expert did not measure.

GAO guidance describes data reliability through accuracy, completeness and applicability for the intended purpose, using a risk-based assessment. Applicability matters in bids: information can be accurate in its original setting and still be unsuitable for the proposed sentence. NIST control-assessment guidance also tailors assessment procedures to risk tolerance. Use these as design principles, not claims that a bid review is an audit or security assessment.

  • Authority: the source creator is competent and accountable for the fact.
  • Directness: the artifact proves this proposition rather than an adjacent one.
  • Scope: entity, service, configuration and population match the offer.
  • Recency: the fact remains valid for the stated period or future use.
  • Reliability: method, completeness and data lineage can be examined.
  • Independence: corroboration does not merely repeat the same origin.
  • Approval: disclosure and offer wording are authorized.

Raise the threshold as the cost of being wrong rises

Define a few consequence classes with minimum rules. For low-consequence descriptions, one current approved source owned by the relevant function may suffice. For scored claims about method, benefit or experience, require a direct source with clear scope and owner confirmation. For claims affecting eligibility, regulation, safety, price, contractual performance or material buyer risk, require controlled primary evidence, independent verification where available and explicit offer authority.

Do not confuse a hierarchy with an automatic pass. Level A evidence for one proposition may be Level C for another. A certificate is primary evidence that a named entity holds a certification within its stated scope and dates. It is only indirect support for the claim that every proposed service control is effective. A customer report can directly prove what that customer measured, but not what every future customer will achieve.

Complementary evidence is useful when items close different gaps. A control description establishes design, a test report establishes operation during a period, and an approved solution statement maps that control to the offered service. This is a chain, not a vote. Three brochures derived from one product page do not create independent sources. Record common origins so repetition is not mistaken for corroboration.

Example consequence-based thresholds
ClassExamplesMinimum treatment
RoutineProcess description or nonmaterial contextCurrent approved owner source
EvaluatedMethod, outcome, experience or differentiatorDirect scoped evidence and owner confirmation
MaterialCentral quantitative or comparative claimControlled primary evidence plus method review
High consequenceEligibility, regulation, security, SLA or promisePrimary record, independent check and named approval
UnsupportedSource misses the required levelNarrow, qualify, escalate or remove

Change the claim when the evidence cannot carry it

For every material claim, record the selected evidence, owner, date, version, relevant extract, applicability boundary, approval and reviewer decision. A source is not stronger because the full file sits in a folder. The exact passage and its relation to the sentence must be inspectable. If confidentiality prevents attaching the record, decide what permissible representation, attestation or controlled access preserves evaluative value without implying the buyer has seen unavailable proof.

When evidence falls short, first seek a better primary record or responsible confirmation. If neither exists before the deadline, narrow the proposition. Replace “all implementations reduce time by 40 percent” with an approved, contextual observation from the measured project, if disclosure is permitted. Label a forecast as a forecast and a target as a target. Route any exception to an authority who sees both the proposed wording and the missing assurance.

Review the hierarchy after major amendments and before submission. Changes to solution scope, legal entity, subcontractors, hosting, service level or price can break applicability without changing the evidence file. FAR treatment of past performance illustrates the discipline: currency, relevance, source and context all matter. The hierarchy does not exist to make a library look orderly. It stops the final offer from claiming more than the organization can responsibly prove.

  • Link the exact supporting extract to the exact claim.
  • Record scope, date, owner, version and permission to use.
  • Seek stronger primary proof before adding secondary material.
  • Narrow fact, period or population when support is limited.
  • Revalidate applicability whenever the offered solution changes.

Useful outcomes from evidence hierarchy for RFP claims

  • Each material claim has an evidence threshold before drafting is finalized.
  • Source prestige cannot hide a mismatch of entity, service, period or proposition.
  • High-consequence commitments receive more independent and current verification.
  • Expert confirmation is recorded with scope and authority instead of becoming folklore.
  • Authors narrow language rather than stretch the available support.
  • Reviewers can explain why a source is sufficient, conditional or unacceptable.

How to run the work

  1. 01

    Atomize the claim

    Separate capability, scope, quantity, timeframe, causality, commitment and comparison so each proposition can be proved independently.

  2. 02

    Rate the consequence

    Assess contractual, eligibility, safety, regulatory, financial, delivery and reputational harm if the proposition is inaccurate or misunderstood.

  3. 03

    Set the evidence threshold

    Define the required authority, directness, currentness, approval and independent corroboration for that consequence class.

  4. 04

    Test candidate sources

    Compare every source to the exact claim and combine items only where each closes a different known gap.

  5. 05

    Resolve the shortfall

    Obtain stronger proof, narrow or qualify the claim, route an explicit exception approval, or delete it before release.

Questions that change the decision

  • What exact proposition would an evaluator reasonably take from this sentence?
  • Which entity, service, configuration, geography, population and period does it cover?
  • Is the claim descriptive, quantitative, causal, comparative, predictive or contractual?
  • What is the worst credible consequence if the claim is false, stale or overbroad?
  • Who owns the fact and who may approve its use in an offer?
  • Does the source prove the proposition directly or only make it plausible?
  • Would independent corroboration materially reduce the risk?
  • Can the wording be narrowed to the part the evidence actually supports?

Where teams lose control

01

A prestigious source may be accepted despite covering the wrong entity or control boundary.

02

A precise number may be copied without denominator, period, exclusions or method.

03

Multiple weak sources may share one origin while appearing to corroborate one another.

04

A subject expert may confirm a fact outside their decision authority.

05

A current webpage may contain less controlled information than an older approved record.

06

A customer example may be accurate but unauthorized for external use.

07

A low-risk evidence rule may be reused for a regulated or contractual commitment.

08

A reviewer may reject useful evidence because it is not in a favored document category.

Measure the finished job

Measure the completed workflow, including review effort and exceptions. Output volume on its own is not evidence of a better process.

  • material claims with an assigned consequence class
  • claims supported by sources matching entity and offered scope
  • evidence items with owner, date and use approval
  • high-consequence claims independently corroborated
  • claims narrowed after evidence review
  • exceptions accepted below the normal threshold
  • post-review claims with missing provenance

Common questions

What is the strongest type of RFP evidence?

There is no universally strongest document type. The strongest evidence is authoritative, direct, current, reliable, appropriately independent, approved and matched to the exact entity, scope and proposition.

Can a subject-matter expert’s statement count as evidence?

Yes, when the expert owns the fact, the statement is recorded with date and scope, and its consequence fits that assurance. It should not replace a controlled record that ought to exist for a high-risk claim.

Does more evidence always make a claim stronger?

No. Repeated sources with one origin do not add independence, and irrelevant attachments obscure useful proof. Add an item only when it closes a known gap in authority, scope, reliability or corroboration.

What if the deadline arrives before the evidence?

Narrow or qualify the wording to the support that exists, seek an explicit risk decision where permitted, or remove the claim. Do not present expected future evidence as already available.

Primary references

Tony Kim

Tony Kim

Founder and CEO

Tony writes about applied AI, dependable product engineering and the systems that turn complex response work into controlled delivery.

Proposal software for source-grounded RFP, RFI, DDQ and questionnaire response work.

Bid, proposal, presales, security and compliance teams. Start with the workflow, constraints and evidence you already have.