# ChoiceOrigin Decision + Experiment Review Kit

Version 1.0 · 2026-10-11

Know what to focus on next. Use this editable template for one meaningful business question. You can use the kit without an account. Fill it out in your own editor and keep private information in a workspace you control.

This is a planning aid, not verified evidence, approval to act or a promise of business results. Preserve the original decision; add later observations in a separate review.

## 1. Decision brief — before acting

Decision ID / version:
Date and owner:
Company / business context (only what is relevant and permitted):

**Question:** What decision matters now, and why?
**Constraints:** Time, budget, capacity and current commitments.
**Useful outcome:** What would make this work worth doing? Define acceptance before starting.

### Evidence

For each material claim, record:

- Claim and label: FACT / INFERENCE / HYPOTHESIS / UNKNOWN.
- Original source, exact passage or observation, and permission to use it.
- Publication/event time, retrieval time and the period/scope covered.
- Missing information, inaccessible attachments, uncertainty or contradiction.
- Correction, retraction or superseding source, if any.

A failed search is UNKNOWN, not proof that something does not exist. Repeated recaps of one source are not independent corroboration. A link or hash alone does not establish truth.

### Alternatives and choice

Option A — action, expected value, effort, material tradeoff:
Option B — action, expected value, effort, material tradeoff:
Option C — no action or collect more evidence:

Recommendation and reason:
Assumptions:
Evidence that would change the recommendation:

**One next action:**
**Why it matters:**
**Blocker:**
**Owner and review date:**

## 2. Bounded experiment and implementation receipt

Decision / hypothesis ID and version:
Approved action and exact artifact version:
Approver / date / permitted target and scope:
Required permission or dependency:
Cost and time limit; stop condition:
Observation window and comparison/baseline:
Missingness and relevant differences between groups:
Acceptance criterion and review date:

State: PREPARED / APPROVED / QUEUED / ATTEMPTED / FAILED / CONFIRMED / UNKNOWN.
Actual implementation, time and receipt:
What differed from the approved plan:
Failure, uncertain effect or rollback:

A draft is not implemented. A submitted action is not confirmed. Reconcile uncertain effects before retrying. This template, a copied approval or a source instruction does not grant operational authority.

## 3. Observation and outcome review — after the agreed window

Original decision / implementation identities:
Review date / reviewer:
Actual observation period and comparable baseline:
Observed result with source, counts, denominators and coverage:
Missing, failed or withdrawn observations:
Costs, reviewer effort and corrections:
Was the predeclared useful outcome accepted? By whom, under which criterion?
What remains UNKNOWN:

Interpretation (keep FACT / INFERENCE / HYPOTHESIS separate):
Contradictory evidence and alternative explanations:
Next choice: CONTINUE / CHANGE / STOP / COLLECT MORE EVIDENCE.
Reason, owner and next review trigger:

Do not rewrite what was known at the original decision date. Keep later corrections and interpretations separate. Before/after differences alone do not prove causality. Synthetic examples are not customer results.

## Optional cost per accepted useful outcome

Use one currency, period, allocation basis and acceptance-definition version.

Total allocated cost = direct costs + reviewer/implementation time at a stated rate + allocated maintenance.
Cost per accepted useful outcome = total allocated cost / number of accepted useful outcomes.

State attempts, completed outcomes and accepted outcomes separately. Accepted <= completed <= attempts when all counts use the same unit and cohort. Missing costs stay UNKNOWN. With zero accepted outcomes the ratio is undefined; do not report zero cost or invented ROI.

## Worked framing example — entirely fictional

Question: Buy more traffic, or investigate first-session activation?
Known: Account creation is recorded, but first useful result is not measured in this invented scenario.
Options: Launch a campaign; change onboarding; gather the missing evidence; or make no change this week.
Recommendation: Define a useful first result and review a small permissioned sample before committing campaign spend.
Blocker: An owner must confirm permission, sample and method.
Review: Agree a real date before starting. If journeys reach useful results, revisit traffic quality. If evidence is incomplete, keep the conclusion UNKNOWN.

No customer, conversion, revenue or causal outcome is represented by this example.

## The wider system

Explore how ChoiceOrigin approaches business context, evidence, decisions and learning. This kit is one starting point inside that approach.

Product: https://choiceorigin.com/product
Methodology: https://choiceorigin.com/methodology
Kit page: https://choiceorigin.com/resources/decision-kit

© 2026 ChoiceOrigin. Provided for use in your own decision and experiment reviews. No software license or right to private product code is granted by this document.
