# Final Assessment: Experimental Design

## Assignment

Write a review-ready experiment design memo for a product, policy, operations, or personal experiment.

Your memo must define the decision, causal question, estimand, randomization unit, metrics, power assumptions, validity checks, and decision rule.

## Required Artifact

Submit a completed experiment design memo.

Minimum sections:

- decision
- causal question
- estimand
- hypothesis
- primary metric
- guardrails
- design and assignment
- power or MDE rationale
- validity checks
- decision rule
- hostile review

## Rubric

| Criterion | Strong | Needs revision |
|---|---|---|
| Decision | Names the action this experiment will inform | Describes curiosity without a decision |
| Causal question | Names population, treatment, control, outcome, and horizon | Leaves one or more core elements vague |
| Estimand | Clearly distinguishes assignment, exposure, unit, and analysis population | Uses "effect" without operational definition |
| Metrics | Prioritizes one primary metric and meaningful guardrails | Lists many metrics without priority |
| Validity | Plans SRM, logging, interference, and missing-data checks | Treats analysis as valid by default |
| Decision rule | Specifies ship, do not ship, rerun, and iterate criteria | Leaves interpretation until after results |
| Critique | Identifies the most likely failure mode | Only repeats the design positively |

## Pass Criteria

Pass if the memo is specific enough that another analyst could implement the experiment and another reviewer could decide whether the result is valid.

## Submission Checklist

- [ ] The decision is explicit.
- [ ] The causal question is one sentence.
- [ ] The estimand is named.
- [ ] The randomization unit matches the interference structure.
- [ ] Primary and guardrail metrics are separated.
- [ ] Validity checks come before effect interpretation.
- [ ] Decision rules are written before results are known.

## Certificate Language

Completed Experimental Design by producing a review-ready experiment memo with estimand, metrics, guardrails, validity checks, and decision criteria.