Trust-Proof Register

How do you keep a material claim verifiable from proposal through delivered outcome?

Problem: claims lose their method, limits, and owner as they move from proposal to delivery.

Question: which claims can a decision-maker verify now, and what evidence must be collected next?

Decision: publish, qualify, hold, or retire each claim.

This method produces one live register that connects a promise to its forecast, actual outcome, and evidence state.

Audience

The register serves operators, delivery leads, finance leads, and evidence owners. One named owner maintains each row. A commercial owner decides whether the claim may enter a proposal.

A delivery or finance owner verifies the outcome. No row becomes verified because its wording sounds plausible.

Inputs

• Each material public, proposal, or delivery claim.
• The method used to calculate or test it.
• The sample or population covered.
• The forecast made before delivery.
• The actual outcome after delivery.
• A durable source a reviewer can inspect.
• A named owner and review date.

If a reading does not exist, enter UNKNOWN. Name the owner and the collection step. Never replace a missing reading with an estimate unless the row is explicitly marked FORECAST.

Row schema

Field	Required reading
Claim	The exact statement a buyer or operator may rely on.
Method	How the number or conclusion is produced.
Sample	The projects, customers, period, or population covered.
Forecast	The outcome predicted before delivery, or N/A with a reason.
Actual outcome	The measured result, or UNKNOWN with owner and collection step.
Source	The inspectable record that supports the row.
Limit	What the evidence does not prove.
Owner	The person accountable for the row.
Review date	The next date the row must be checked.
Verification state	One state from the contract below.
Next evidence action	One observable step that can improve or retire the row.

Verification states

State	Meaning	Allowed use
UNKNOWN	A required reading or source is missing.	Discovery only. Do not publish as proof.
FORECAST	A dated prediction exists before the outcome.	Label as forecast and show assumptions.
OBSERVED	An actual outcome exists, but method or independent review is incomplete.	Qualified internal or proposal use.
VERIFIED	Method, sample, source, limits, and actual outcome have been reviewed by the named authority.	Buyer-facing use within the stated limits.
REJECTED	Evidence contradicts the claim or cannot support it.	Retire or rewrite the claim.
STALE	The review date passed or the operating context changed.	Hold until reviewed.

Build procedure

1. Inventory every material claim in scope. Output: one row per exact claim.
2. Attach method, sample, source, and limit. Output: an attributable evidence boundary.
3. Separate forecasts from actual outcomes. Output: no hindsight rewritten as prediction.
4. Assign an owner, review date, and verification state. Output: a controlled row.
5. Name one next evidence action. Output: an observable move that improves or retires the claim.
6. Review commercial use against the row state. Output: PUBLISH, QUALIFY, HOLD, or RETIRE.

Use procedure

Before a claim enters a proposal, filter the register to VERIFIED and current rows. Show the source and limit beside the claim. After delivery, compare forecast with actual outcome and update the row.

Record variance; do not overwrite the forecast. When a method, offer, or operating context changes, mark affected rows STALE until reviewed.

Quality criteria

• Every material claim has one row and one owner.
• Forecast and actual outcome occupy separate fields.
• VERIFIED rows contain method, sample, source, limit, review date, and reviewer authority.
• Every UNKNOWN names an owner and collection step.
• Sources are inspectable, not circular links to the claim itself.
• Limits are specific enough to prevent broader use than the evidence warrants.
• Every row ends with one next evidence action.

Review cadence

Review on every proposal-pattern change, after each material delivered outcome, and at least quarterly while a claim remains in active use. Review immediately after an adverse outcome, source failure, method change, or buyer challenge.

Checks

• Can a new reviewer reproduce the state from the listed source?
• Was the forecast recorded before the outcome?
• Does the sample support the scope of the claim?
• Is the review date current?
• Does the next action have an owner and observable completion signal?

Failure modes

• A homepage statement is treated as its own evidence.
• A forecast is overwritten after delivery.
• A strong average hides segment, period, or sample limits.
• VERIFIED means “approved by marketing” rather than source-checked.
• The register records claims but fires no review or evidence action.

Proof of done

The owner can select any active claim, reproduce its method and evidence boundary, state whether it may be used, and perform the named next action without asking the register author for context.

Kill signal: retire or hold any claim when its source is missing, its method cannot be reproduced, its actual outcome materially contradicts it, or its review date passes without a named recheck.

Changes my mind: repeated use shows that a lighter existing evidence instrument preserves forecast, actual outcome, limits, authority, and review state with less work.

Retrieval

Pull this method before publishing performance claims, preparing evidence-led proposals, reviewing delivered outcomes, or designing a machine-readable proof loop.

Version delta: adds the canonical operational guide for preserving claims from promise through verified outcome.

Context

• pairs-with Data Flow Map — move evidence through systems without losing ownership.
• pairs-with Unit Economics — verify the financial readings used in commercial decisions.
• contrasts-with Venture Proof Page — present a venture case; do not use it as the row-level verification contract.
• applies-to Performance — proof matters when it changes a real gauge and future decision.

Questions

Next question: which active claim has the greatest decision impact and the weakest verification state?

• Which claim is used most often?
• Which source can a new reviewer inspect now?
• Which UNKNOWN has the cheapest collection step?