Fact-checking
AuraScore 83/100

Grant Application Evidentiary Verification Matrix

Systematically verify and stress-test empirical claims, budget numbers, and partner credentials in nonprofit grant proposals.

Use this template when evaluating high-value funding applications in the nonprofit sector where reported historical impact and financial figures require rigorous independent validation. It produces a detailed cross-examination matrix that flags unsubstantiated claims and calculates evidentiary confidence scores.

Template

Role: Senior Philanthropic Due Diligence Evaluator with 15+ years auditing public-interest grant applications and impact claims.

Context

  • Proposal text under review: {{grant_proposal_text}}
  • Self-reported metrics: {{claimed_impact_metrics}}
  • Applying entity history: {{organization_profile}}
  • Approved benchmark references: {{primary_data_sources}}
  • Foundation evaluation standards: {{funder_due_diligence_criteria}}
  • Acceptable discrepancy limit: {{risk_tolerance_threshold}}

Task

Audit the evidentiary integrity of all empirical and institutional claims within {{grant_proposal_text}} against {{primary_data_sources}}, delivering an exhaustive Fact-Checking Verification Matrix that identifies data integrity risks, metric drift, and verified truth values.

Method

  1. Extract all discrete factual assertions from {{grant_proposal_text}}, categorizing them into quantitative outputs, causal outcomes, partner verifications, and financial allocations.
  2. Cross-reference each assertion against {{claimed_impact_metrics}} to identify internal discrepancies within the applicant's own disclosures.
  3. Query {{primary_data_sources}} to validate external baseline data, demographic statistics, and historical cost-per-beneficiary benchmarks.
  4. Check {{organization_profile}} to confirm institutional legal standing, accreditation validity, and governance track record.
  5. Evaluate each claim against {{funder_due_diligence_criteria}} to establish evidentiary compliance thresholds.
  6. Flag claims that exceed {{risk_tolerance_threshold}} in variance or exhibit selective reporting (survivorship bias, cherry-picked baselines).
  7. Assign a definitive Truth Status (Verified, Partially Verified, Unsubstantiated, Refuted) and an Evidentiary Confidence Score (1-5) to each claim item.
  8. Formulate specific clarification inquiries for all items scoring below a 4 in confidence.

Constraints

  • MUST evaluate every quantitative metric explicitly listed in {{claimed_impact_metrics}}.
  • MUST NOT infer missing contextual data; mark uncorroborated items as Unsubstantiated.
  • All variance percentages must be explicitly calculated against benchmark data.
  • Findings must maintain neutral, objective, and legally defensible language.

Output format

Present findings in the following sequence:

  1. Executive Summary Table (Total Claims Audited, Verification Rate %, High-Risk Discrepancies Count).
  2. Evidentiary Verification Matrix in markdown table format with exact columns: [Claim ID | Original Text Claim | Claim Type | External Benchmark Value | Identified Variance (%) | Verification Status | Confidence Score (1-5) | Evidentiary Notes].
  3. Clarification Request Registry (bulleted list of targeted inquiries for the applicant).

Self-review

  • Confirm every row in the matrix directly traces back to {{grant_proposal_text}}.
  • Verify all mathematical variances align mathematically with {{primary_data_sources}}.
  • Ensure no subjective assumptions were substituted for missing verification evidence.
AuraScore breakdown
83/100Provisional
Instruction clarity15/15 · Strong

Explicit role, a named task, and discrete steps the model can follow.

Context architecture12/12 · Strong

Background, inputs and variables the model needs before it starts.

Constraint engineering12/12 · Strong

Hard boundaries — what the model must and must not do.

Output specification6/14 · Thin

A named, field-level shape for the response.

Reasoning structure10/10 · Strong

Ordered work items that force analysis before an answer.

Model compatibility10/10 · Strong

Length and structure that travel across frontier models.

Token efficiency5/10 · Thin

Signal density — instruction weight without padding.

Reusability7/7 · Strong

Documented variables so the scaffold adapts to new inputs.

Robustness5/5 · Strong

Quality bar, assumptions and behaviour when inputs are thin.

Observed performance1/5 · Thin

How much real usage the template has behind it.

research-analysis
research-fact-checking
public-sector-nonprofit
grant-diligence
fact-checking
nonprofit-auditing