A decision you can inspect.

The six required donor questions

Deep Research is not complete because a Form 990 was summarized. Every decision-grade investigation must address six questions:

1. What does the organization actually do with donor money? Operating model, programs, geography, beneficiaries, grants, cash versus in-kind support, and what donor money actually finances.
2. Is it financially healthy, and is capital being deployed rationally? Liquidity, reserves, restrictions, commitments, fundraising economics, compensation, capital use, and financial resilience.
3. Does the principal work actually work? Independent outcome evidence, author and funder affiliations, evaluation design, causal strength, outcome directness, and current applicability.
4. Do the major claims and numbers reconcile? Filings, audited statements, annual reports, grant schedules, websites, government data, and other comparable records are reconciled before concern language is used.
5. What material concerns should a donor know? Audit findings, related parties, unusual transactions, governance, fundraising practices, litigation, regulators, and credible independent reporting.
6. What would another unrestricted dollar likely accomplish now? Capacity, commitments, expansion plans, reserves, bottlenecks, need, room for more funding, and the likely productive use of additional capital.

These six questions are a minimum research contract, not a ceiling. If the investigation raises a new material question, No Halo's publication standard is to attempt to resolve it before handing it back to the donor. If it still cannot be answered, the report should state what was checked, why the answer remains unresolved, and what evidence or response would change the donor decision.

Donor action is separate from diagnostic scoring

No Halo may use numerical diagnostics because numbers can make patterns easier to see. They are never the final answer. A diagnostic score is shown only when enough relevant evidence exists to calculate it responsibly, and it must always be read with coverage and evidence confidence.

Five donor outcomes

A completed Deep Research file should end with a clear current action:

• Recommended. A high-confidence case for a meaningful unrestricted gift. This is intentionally a high bar.
• Proceed with reservations. Giving appears reasonable, but the donor should understand specific limitations or tradeoffs.
• Resolve these questions first. Discrete unanswered issues could change the decision and should be settled before giving.
• Hold pending enhanced diligence. The current evidence does not establish enough confidence to recommend a meaningful gift yet.
• Do not recommend. The current verified evidence affirmatively weighs against giving at this time.

A HOLD or DO NOT RECOMMEND conclusion does not require an accusation of misconduct. For example, a donor may reasonably pause because operating reality, controls, outcomes, or deployment capacity could not be established. The report must say that plainly.

Financial stewardship

The financial assessment is intended to describe financial resilience, stewardship signals, and capital deployment, not moral worth and not impact. No Halo separates reported accounting totals from the question a cash donor actually cares about.

• Reported program, management, and fundraising expense remain visible but are not treated as impact.
• Noncash contributions and gifts in kind are separated where the filing or audited statements permit it so donated goods do not mechanically drive a cash-donor interpretation.
• Fundraising economics can include the reported cost to raise $100 of cash-oriented contributions when the basis can be derived defensibly.
• Joint-cost allocations and solicitation-related program allocations are kept visible. No Halo does not reclassify them without evidence.
• Liquidity, donor restrictions, fixed assets, commitments, reserves, deficits, and capital campaigns are considered separately. A building is not treated as cash simply because the associated net assets are unrestricted.
• Compensation is presented with components and trends where available. No Halo does not call pay excessive or reasonable without a defensible benchmark.

Governance

Policy checkboxes are baseline disclosures, not proof that governance worked. No Halo considers the reported voting board and independence, control concentration, compensation-setting process, Schedule L interested-person transactions, Schedule R related organizations, audit related-party disclosures, significant-diversion disclosures, Federal Audit Clearinghouse findings, repeated findings, corrective-action history, and other material control evidence when available.

A governance concern must be tied to evidence. A missing policy or missing disclosure is not converted into misconduct merely because the organization is small.

Impact evidence

No Halo does not infer impact from spending ratios, beneficiary counts, mission statements, or fundraising claims. Impact evidence is evaluated on the quality of the evidence for the principal outcomes.

Transparency and small organizations

Transparency is evaluated proportionally. A small volunteer-run charity is not penalized merely because it does not purchase an audit or publish an institutional-style annual report. It is still expected to provide scale-appropriate evidence of current operations, governance, financial control, outcomes, legal standing, and ability to deploy additional funds responsibly.

Evidence confidence may remain low when those facts cannot be established. Low evidence confidence is not the same as evidence of wrongdoing.

Reconciliation before accusation

A difference between two numbers is not automatically a discrepancy, and a discrepancy is not automatically misconduct. No Halo first tests entity, period, definition, accounting basis, geography, population, restriction, and scope. When authoritative evidence explains the difference, the report says so. When a plausible explanation exists but is not established, the report says that too. When comparable primary sources genuinely remain in conflict, the disagreement stays visible.

Current evidence and the Material Event Sweep

Tax filings often lag current operations. No Halo separates the latest filed financial period from the date through which current evidence was researched. A Deep Research recommendation is not treated as current merely because a prior No Halo file exists.

Before a current donor conclusion is issued, the research system is expected to perform a fresh Material Event Sweep across the Event Ledger and current external evidence for potentially material legal, regulatory, audit, leadership, financial, impact, and operating developments.

AI research and independent verification

Deep Research uses an automated primary investigation followed by a separate automated verification pass designed to challenge the draft conclusion. Every accepted material claim still has to survive source validation.

The verifier is instructed to reject unsupported negative inference, invented reconciliations, allegations presented as findings, impact inferred from spending, compensation judgments without a benchmark, stale studies presented as current proof, material evidence that was ignored, and unresolved questions that the evidence already answers.

What happens when No Halo still cannot answer something?

An unresolved question should become a useful course of action, not a dead end. The report should identify the exact diligence question, why it matters, and what evidence would change the recommendation. Where verified contact information is available, No Halo can surface it. The consumer experience is also designed to provide copy-ready outreach so the donor can ask the charity directly.

A true research-process failure is different. If the investigation itself cannot complete because the model, source, identity, or verification pipeline failed, the result should not be presented as decision-grade.

Corrections, right of reply, and editorial review

Organizations can provide evidence, dispute facts, and request corrections. They cannot buy a better conclusion or removal of a supported finding. Extreme positive and adverse results may also enter No Halo's internal editorial-review queue before No Halo chooses to amplify them through the public Intelligence feed or a future Recommended list. That editorial process does not delay the paying donor's completed Deep Research answer.

Submit a correction · corrections@nohaloresearch.com · Open Deep Research