analysis-qa-checklist

name	analysis-qa-checklist
description	Pre-delivery quality gate for analytical work. Activate before sharing any analysis, dashboard, or query result with stakeholders. Catches logic errors, presentation issues, and assumption gaps before they reach the audience.

When to use

• Before delivering any analysis, report, dashboard, or query result to stakeholders
• After completing a peer review to confirm all issues are resolved
• When returning to an analysis after a break to verify nothing was missed
• Any output that will inform a business decision

Process

1. Run automated checks — execute scripts/qa_runner.py on the output file; flags numeric anomalies (unexpected nulls, out-of-range values, division results), structural issues, and formatting problems
2. Complete the logic checklist — work through references/qa_checklist_master.md covering: question framing, data sourcing, transformations, statistical validity, finding statements, and presentation
3. Check for common errors — review references/common_analysis_errors.md for analysis-type-specific mistakes (e.g. survivorship bias in cohort analysis, multiple comparison issues in A/B tests)
4. Validate assumptions — confirm every assumption is documented, sourced, and has been sensitivity-tested where uncertain
5. Review the narrative — verify: the conclusion follows logically from the data, all caveats are stated, and recommendations are specific and actionable
6. Sign off — complete assets/qa_signoff_template.md documenting the reviewer, issues found, resolutions, and delivery decision

Inputs the skill needs

• Required: output file (CSV, notebook, SQL result, or document)
• Required: the original analysis question or brief
• Required: reviewer name and intended audience

Output

• QA runner report with automated flags
• Completed checklist with pass/fail for each section
• assets/qa_signoff_template.md (filled) — formal sign-off confirming the output is ready for delivery