# Campaign Evidence and Human Decision Record

Editable Markdown working material.

Use this record to connect a campaign version, the evidence examined and the decision actually made. Reference detailed resource outputs instead of copying customer data into every section. Replace the italic instructions; duplicate rows as needed. Unknown, not supplied, not performed and not applicable are different states. An empty decision field means no recorded approval.

## 1. Decision and version

| Field | Working entry |
|---|---|
| Campaign / version / record date | *Complete* |
| Objective and decision requested | *State the business outcome and exact proposed action* |
| Audience, purpose, channel and market | *Separate each materially different use* |
| Accountable decision owner and authority reference | *Name only an actual owner; otherwise Unassigned* |
| Budget, capacity, timing and scope limits | *Reference approved limits; do not infer approval* |
| Creative, landing, offer and measurement versions | *Record the versions this decision applies to* |
| Permitted data/tool use | *Reference actual organisation rules, including AI-provider disclosure* |

## 2. Evidence index

| Evidence ID | Source / locator / version | Event date and extract date | Proposition supported | Classification and limitation |
|---|---|---|---|---|
| E1 | *Reference, not unnecessary personal data* | *Both where relevant* | *One material proposition* | *Fact / assertion / permission evidence / assumption / conflict / unknown* |

Record what was actually inspected. A link alone does not prove it was accessible; a screenshot may omit context. Preserve conflicting versions. A generated review is working material, not independent proof of the underlying fact.

## 3. Evidence-to-decision map

| Decision item | Proposed position / action | Evidence IDs and contrary evidence | Gap / consequence | Required human review |
|---|---|---|---|---|
| Customer need and audience criteria | *Rationale* | *Source references and sample limits* | *What cannot be concluded* | *Research / audience owner* |
| Claim and whole communication | *Exact wording and placement version* | *Proof, qualification and context references* | *Unsupported or omitted context* | *Brand/product and specialist where required* |
| Purpose, permission and recipient | *Specific use and downstream destination* | *Notice, authorised assessment/consent where applicable, restrictions* | *Do not infer from possession* | *Privacy/data owner* |
| Exclusion and suppression | *Proposed rule and propagation path* | *Source/destination observations and timestamps* | *Unobserved sync / conflicts* | *Channel/operations owner* |
| Channel and spend rationale | *Role and proposed allocation* | *Comparable outcomes, assumptions, capacity* | *No invented marginal return* | *Media/finance/budget owner* |
| Measurement | *Metric, denominator, horizon and source* | *Definition and QA references* | *Attribution/causal limits* | *Analyst* |

Roles above are review prompts, not assertions that the organisation has these titles or that each is legally required. Use its actual responsibility arrangement. Reference the separate measurement record for detailed definitions/tests.

## 4. Assumptions, unresolved questions and AI observations

| Item ID | Assumption / question / AI observation | Source or reason it is uncertain | Decision affected | Check / owner / due trigger |
|---|---|---|---|---|
| Q1 | *Keep AI interpretation distinct from source evidence* | *Evidence IDs or Missing* | *State consequence* | *No invented deadline or approval* |

Record AI resource, input/evidence version and output date/reference if used. Record the human correction or rejection of an AI observation; do not treat the existence of a Run Record as validation. Use an organisation-approved location and its actual access/retention rules.

## 5. Procedure and review evidence

| Check ID | Expected behaviour / proposed procedure | Actual supplied result / Not performed | Evidence and scope | Review result / unresolved action |
|---|---|---|---|---|
| QA1 | *Example: withdrawal excluded from destination audience* | *Record observation only after it exists* | *Date, version, synthetic/live scope* | *Supported / Conflicting / Incomplete / Not performed* |

An action marked completed still needs outcome verification. A test using synthetic data does not prove all production records are correct. Missing evidence cannot be marked passed.

## 6. Human decision — complete only from an actual decision

| Field | Working entry |
|---|---|
| Disposition | Pending / Hold / Rejected / Approved for precisely stated action |
| Human owner / decision date / authority | *Leave Pending until supplied* |
| Exact action and version covered | *Approval of a draft is not permission to launch* |
| Rationale and evidence considered | *Include unresolved risks and contrary evidence* |
| Conditions, exclusions and required further approvals | *State boundaries* |
| Expiry or review trigger | *Only if owner supplies/agrees it; no invented period* |
| Operational executor and execution evidence | *Separate later action; Not performed initially* |
| Verification of outcome | *Actual evidence or Pending* |

## 7. Follow-up and change history

| Date/version | Change or unresolved item | Effect on earlier decision | Owner / follow-up | Actual closure evidence or Open |
|---|---|---|---|---|
| *Complete* | *Include changed claims, audience purpose or metric definition* | *Re-review needed?* | *Agreed action* | *Completion alone is not verified closure* |

Before relying on this record, confirm that the decision references the same campaign/evidence versions, unsupported claims and permissions remain visible, and required humans actually reviewed their parts. This document does not authorise spending, publication, contact or legal compliance. AI output remains working material. Use is subject to the Amo.ng Terms; no new licence is asserted.