# Career Delivery Evidence

Copy this template into your portfolio project. Replace prompts with your work and evidence links. Label each source as observed, reported, inferred, or simulated. Leave missing observations and pending decisions explicit; never invent impact or approval.

## Workflow and Setback

- Career route, project, and your contribution:
- User, task, current workflow, and intended outcome:
- Evidence record, date, source type, and privacy treatment:
- Assumption or expectation that failed:
- What changed your understanding, and what remains unknown:

## Decision Brief

- Decision needed and why it matters now:
- Accountable decision owner and affected stakeholders:
- Work you may perform now, bounded choices, and actions requiring authorization:
- Recommendation and main uncertainty:
- Stakeholder objection, source, and response; label role-play:
- Decision deadline, actual response, or pending status:

| Option | User and interaction effects | Quality and risk | Effort and ongoing cost | Evidence still needed |
|---|---|---|---|---|
| Option A | | | | |
| Option B, including manual/no-build if credible | | | | |

## Experiment and User Walkthrough

- Learning question and chosen stage: prototype, pilot, or production:
- Audience, data, duration, authority, and rollback:
- Decision rules set before measurement: continue / change / stop / inconclusive:
- Task steps from starting point to completion:
- Incorrect-output or missing-data case and recovery:
- Where the user hesitated, needed help, reviewed, or corrected:
- Evidence links; state whether real user, peer simulation, or offline replay:

## Outcome Comparison

Use the same definition and comparable populations and windows before and after. State sample counts and exclusions. Preserve missed targets and missing values.

| Measure | Definition, target, and source | Baseline value/window/sample | Follow-up value/window/sample | Limitation |
|---|---|---|---|---|
| User outcome | | | | |
| Quality or safety guardrail | | | | |
| Human review and correction effort | | | | |
| Cost per successfully completed task | | | | |

- Cost calculation: model usage + retries + supporting services + review labor; divide by successfully completed tasks:
- Measured inputs versus estimates, labor-rate assumption, and unsuccessful task costs:
- Manual alternative or value assumption, and the product decision it supports:
- What the sample supports, what it cannot establish, and any changed comparison conditions:

## Owned Follow-Up

- Continue / change / stop / collect more evidence, with rationale:
- Decision owner's actual response or pending status:
- Next bounded action, owner, due date, and verification:
- One improvement to how you frame, build, review, or hand off the work:
- Improvement owner, review date, and keep/revise/retire criteria:
- Review result and evidence; leave pending until observed:

## Manual Acceptance

Have a peer apply the lesson's rubric. For each check, record **met**, **needs work**, or **not observed**, plus evidence or a missing observation: workflow grounded; authority respected; tradeoffs communicated; interaction tested; outcome compared; setback owned; process improved.

- Reviewer and date:
- Rubric findings and required revisions:
- Accurate portfolio claim, including simulation and real-user evidence limits: