Free tool · private draft · no signup

Do not just finish the work. Publish the proof.

Turn one real test into a field report that says what happened, how you know, what the evidence cannot prove, and what you will do next.

The missing final mile

Evidence is only useful when another person can inspect it.

A dashboard can show a number without showing where it came from. A case study can tell a moving story without naming the denominator, the reporting window, or the people who disappeared. A proof pack keeps those missing pieces in the report.

It does not promise impact. It publishes the strongest level the record can support—intention, delivery, use, return, or evidenced change—and keeps the method and limitations attached.

Build the proof pack.

Private draft · saved on this device

Measure → report

Your scorecard. No retyping.

Bring a snapshot from the scorecard in this browser. Nothing is merged, synced, submitted, or published. Your scorecard stays unchanged.

Open the scorecard

01

Name the promise.

Give the report a boundary before adding numbers.

02

Keep the signals separate.

Leave unknown counts blank. Zero is a valid result.

03

Show your work.

The method and the limits belong beside the result.

04

Make the decision.

A report should change what happens next.

Decision earned by this evidence

The anatomy of a credible field report.

Small enough to read · complete enough to inspect
01

Name the promise

State who the work was for, what was delivered, and the change it intended to support.

02

Report each signal

Keep reach, delivery, use, return, and evidenced change as separate counts from one reporting window.

03

Show the method and limits

Explain how evidence was collected, what it can support, and what it cannot establish.

04

Make an evidence-shaped decision

Choose to continue, change, or stop, and name the next test the evidence earned.

Proof packs, plainly answered.

Why does my report say Needs review?

The reported evidence level changes to Needs review when entries contain invalid counts, conflicting counts, invalid reporting dates, or a change count without a meaningful method and limitations. For example, zero complete deliveries and three uses need explanation before claiming meaningful use. Correct the entries or explain the source mismatch; the tool does not independently verify evidence.

Which report format should I choose?

Use plain text for an email or a document editor, Markdown for a formatted working report, and JSON for a restorable backup. All three include the finding, method, limitations, checklist, and consistency warnings. Review the full on-screen report before sharing. If clipboard access is blocked, open Select report text and copy manually. Files are readable, not encrypted; remove private details first.

What is included when I print a proof pack?

The printed report includes the promise, people served, reporting dates, separate counts, finding, evidence method, limitations, decision, and before-sharing review. The review prints even when its on-screen panel is collapsed. Missing fields and consistency warnings remain visible; printing does not verify or publish the report.

What does Before you share check in my report?

It checks eight areas: project and promise, people served, reporting dates, signal formats and obvious consistency, finding, method, limitations, and decision with next test. Open any check to jump to its field. Blank counts may stay unmeasured, and zero is valid; fractions and invalid counts are flagged rather than rounded. The checklist and warnings are included in Markdown, plain text, copied reports, and JSON exports. You can still export an unfinished draft.

Does completing the review verify my impact or approve publication?

No. The checks inspect field completeness and obvious inconsistencies, not whether the statements are true, consent exists, outcomes were caused by the work, or an impact claim is independently verified. You still need to review accuracy, permissions, and private details before sharing. Nothing is automatically published.

Can I turn my impact scorecard into a report without retyping it?

Yes. Choose Review scorecard snapshot in the builder to inspect the scorecard saved in this browser. Back up the current report if needed, then explicitly choose Replace report with snapshot. Blank counts stay unmeasured and zero stays zero. The planned change is labelled as a plan, not a finding. Evidence notes move to the method field for review; add separate limitations, reporting dates, findings, and a decision yourself. The original scorecard stays unchanged and nothing is published.

How do I restore a report draft from a backup?

Choose Restore JSON backup in the proof pack builder and select a SIX33 proof pack JSON export. Review the project and promise, back up the current draft if needed, then explicitly replace it. Markdown and plain text are shareable reports, not a restorable backup. Older complete JSON exports are supported. Files stay in your browser and are readable, not encrypted; nothing is automatically published.

What is a proof pack?

A proof pack is a compact, dated record of a promise, the people it was for, what was delivered, what happened next, the evidence method, the limitations, and the decision that followed. It is designed to be inspected, not merely admired.

Is a proof pack the same as an impact report?

Not automatically. A proof pack can honestly end at delivery, use, or return. It should use impact language only when evidence supports an intended change and the method and limitations are published beside the count.

Can I publish a report when the result is zero?

Yes. Zero use, zero return, or zero evidenced change can be decision-grade information when the reporting window, method, and attempted delivery are clear. A useful record does not require a flattering result.

Does AGENT SIX33 receive the information I enter?

No. The builder saves the draft in local browser storage. Nothing is submitted to AGENT SIX33. Exported Markdown, plain-text, and JSON files are created on your device.

What should I do when my numbers do not share the same cohort?

Do not calculate a conversion rate across mismatched groups. Explain which source and reporting window belongs to each number, or leave the number out until the comparison is valid.