1.6 KiB
1.6 KiB
name, description, usage
| name | description | usage |
|---|---|---|
| prioritize-hypotheses | Scores experiment backlog using impact, confidence, effort, and guardrail readiness. | /growth-experiments:prioritize-hypotheses --source backlog.csv --capacity 6 --framework rice |
Command: prioritize-hypotheses
Inputs
- source – backlog file, experiment tracker, or Notion database ID.
- capacity – number of experiments that can run in the next sprint/cycle.
- framework – ice | rice | custom; determines scoring weights.
- guardrails – optional JSON/CSV for mandatory guardrail requirements.
- filters – tags or OKRs to focus on (acquisition, activation, retention, monetization).
Workflow
- Data Ingestion – load backlog, normalize fields, and enrich with latest metrics.
- Scoring Engine – calculate ICE/RICE/custom scores, factoring guardrail readiness.
- Portfolio Mix – ensure balance across funnel stages and surfaces; flag conflicts.
- Capacity Planning – fit highest-value tests into available slots, accounting for owners + effort.
- Decision Pack – generate prioritized list, rationale, and trade-off notes for approval.
Outputs
- Ranked backlog with scores, dependencies, and guardrail status.
- Capacity plan showing selected tests plus waitlist.
- Decision memo summarizing trade-offs and next actions.
Agent/Skill Invocations
experimentation-strategist– orchestrates prioritization + governance alignment.insight-analyst– validates data quality and metric assumptions.hypothesis-libraryskill – links past learnings to current ideas.guardrail-scorecardskill – enforces readiness requirements.