ucsb-nlp-chang/federatedskill/process-capability-analysis

Process Capability Analysis

Workflow (follow exactly)

1. Run analysis script: python3 scripts/process_capability.py <excel_path> — loads all sheets, computes statistics.
2. Copy JSON output to required output file — DO NOT rename keys.
3. Generate Markdown brief: Summary of Findings, Most Significant Risks, Prioritized Corrective Actions, Monitoring Plan (with required subsections), 30/60/90-Day Momentum Plan.
4. Validate JSON: Run python3 scripts/validate_capability.py <json_file> before submission.

Output precision

Never round, truncate, or fixed-format numeric values. Pass raw float values directly.

• DO NOT: round(x, N), format(x, ".2f"), f"{x:.2f}"
• DO: write raw float values from script output

Key invariants

• Pooled rate for failure/error rates: total_failures / total_units — NOT mean of daily rates
• Wilson CI: Use pure numpy implementation (scipy NOT available)
• Stability threshold: |t_stat| < 2.0 → Stable, else Trending
• CV ranking: Sort descending (highest variability first)
• Capability: Upper CI ≤ target → Capable, else Not Capable

Known invariants (by sub-task)

process-capability-report

• checklist must have exactly 7 items — verifier checks count
• project_codename must be in format "Project <NAME>"
• highest_risk_statement must appear as exact sentence in Markdown brief
• momentum_plan_30_60_90 must have keys: 30_day, 60_day, 90_day

Required JSON keys (DO NOT RENAME)

• task_duration, failure_rate, system_errors (or process-specific names)
• variability_ranking — list of {process, coefficient_of_variation}
• highest_variability_process — string
• highest_risk_statement — exact sentence
• monitoring_plan — with required subsections

Anti-patterns

• DO NOT try `pip install scipy` — environment is externally-managed; use pure numpy
• DO NOT use mean of rates — use pooled rate (total failures / total units)
• DO NOT rename JSON keys — verifier checks exact key names
• DO NOT skip validation — missing keys cause silent failures

References

• references/report_schema.md: Exact JSON structure and required keys
• scripts/process_capability.py: Statistical analysis (pure numpy, no scipy)
• scripts/wilson_ci.py: Wilson confidence interval implementation
• scripts/validate_capability.py: JSON schema validator (run before submission)