SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillspdca-cycle

pdca-cycle

Testing & quality

View repository

Updated 10-15 23:47

About this skill

PDCA Cycle

Prerequisite Protocol

Environment Check

PROJECT_ROOT=$(git rev-parse --show-toplevel 2>/dev/null || echo "unknown")
BRANCH=$(git branch --show-current 2>/dev/null || echo "unknown")
COMMIT=$(git rev-parse --short HEAD 2>/dev/null || echo "unknown")

echo "PROJECT: $PROJECT_ROOT"
echo "BRANCH: $BRANCH"
echo "COMMIT: $COMMIT"

Prerequisite Skill Check

# 检查前置工件
GOAL_ARTIFACT="memory/artifacts/goal-oriented/latest.json"
MVP_ARTIFACT="memory/artifacts/mvp-first/latest.json"

if [ -f "$GOAL_ARTIFACT" ]; then
  echo "FOUND: goal-oriented artifact"
fi

if [ -f "$MVP_ARTIFACT" ]; then
  echo "FOUND: mvp-first artifact"
fi

mkdir -p memory/artifacts/pdca-cycle
PDCA Cycle

Overview

The PDCA cycle (Plan-Do-Check-Act) is a continuous improvement methodology that continuously optimizes processes and results through a cycle of four stages: Plan, Do, Check, and Act.

Core Value：

  • Data-driven decision-making
  • Small steps, fast iteration
  • Continuous improvement, spiraling upward
  • Reduce trial-and-error costs

Origin: Deming Cycle, popularized by W. Edwards Deming

When to Use

Applicable Scenarios：

  • Iterative tasks that require continuous improvement
  • Quality assurance and process optimization
  • Pilot validation of new solutions
  • Data-driven decision-making
  • Performance optimization and cost reduction
  • Improving user satisfaction

Inapplicable Scenarios：

  • One-time tasks that do not require improvement
  • Emergencies that require immediate action

The Process

digraph pdca {
    rankdir=TB;

    Plan [label="Plan\n计划", shape=box, style=filled, fillcolor="#c8e6c9"];
    Do [label="Do\n执行", shape=box, style=filled, fillcolor="#bbdefb"];
    Check [label="Check\n检查", shape=box, style=filled, fillcolor="#fff9c4"];
    Act [label="Act\n处理", shape=box, style=filled, fillcolor="#f8bbd0"];

    Plan -> Do [label="制定计划"];
    Do -> Check [label="收集数据"];
    Check -> Act [label="分析结果"];
    Act -> Plan [label="下一轮循环", style=dashed];
}

Detailed Explanation of the Four Stages

Plan

  • Clarify goals and metrics
  • Analyze the current situation and gaps
  • Develop a specific action plan
  • Estimate risks and response plans

Do

  • Pilot on a small scale
  • Execute strictly according to the plan
  • Collect data during execution
  • Record problems and observations

Check

  • Compare expected and actual results
  • Analyze the causes of deviations
  • Extract lessons learned
  • Present data visually

Act

  • Standardize successful practices
  • Improve shortcomings
  • Plan the next cycle
  • Expand the scope of application

Cycle Checklist

Use the following checklist to ensure each stage is fully executed:

Plan Stage

  • Clarify goals and measurement metrics
  • Collect current-state data
  • Analyze root causes (fishbone diagram, 5 Whys can be used)
  • Develop a specific action plan
  • Determine responsible persons and time nodes
  • Estimate risks and response measures

Do Stage

  • Select a small-scale pilot area
  • Prepare the required resources
  • Execute strictly according to the plan
  • Collect process data (time, cost, quality)
  • Record problems and anomalies encountered
  • Keep the pilot environment controllable

Check Stage

  • Compare actual results with expected goals
  • Analyze the causes of differences
  • Verify the effectiveness of the solution
  • Extract successful experiences and failure lessons
  • Data visualization (charts, reports)
  • Team discussion and consensus building

Act Stage

  • Standardize successful practices
  • Document best practices
  • Improve shortcomings
  • Plan the goals of the next cycle
  • Expand the scope of application (if applicable)
  • Share experience with relevant teams

Examples

Case 1: PDCA Application for Code Quality Improvement

Plan:

  • Goal: Reduce bug density by 50% (from 15 per thousand lines → 7.5 per thousand lines)
  • Current situation analysis: The main problems are insufficient boundary checking and missing exception handling
  • Action plan:
    1. Introduce static code analysis tools
    2. Enforce Code Review
    3. Write unit test templates

Do:

  • Pilot in Project A (1 month)
  • Introduce SonarQube
  • Weekly Code Review meetings
  • Collect data: number of bugs, discovery stage

Check:

  • Bug density reduced to 8 per thousand lines (close to target)
  • Static analysis found 120 issues, no new bugs after fixes
  • Code Review found 35 potential issues
  • But development time increased by 15%

Act:

  • Standardization: Integrate SonarQube into CI/CD
  • Improvement: Optimize the Code Review process to reduce time
  • Next round goal: Maintain bug density and shorten development time

Case 2: Iterative optimization to improve user satisfaction

Plan:

  • Goal: Increase user satisfaction from 3.2 to 4.0
  • Current situation: Slow response and incomplete features
  • Plan:
    1. Shorten response time (< 2 hours)
    2. Add high-frequency demand features

Do:

  • Pilot for 2 weeks
  • Add customer service staff
  • Develop the Top 3 demanded features
  • Collect feedback data

Check:

  • Satisfaction increased to 3.8 (did not reach 4.0)
  • Response time met the target
  • High usage rate of new features
  • But some features lack usability

Act:

  • Standardization: Write response time requirements into the SLA
  • Improvement: Optimize the UI/UX of new features
  • Next round goal: Satisfaction > 4.0

Common Pitfalls

Pitfall 1: Skipping the Check stage and directly entering the next round

  • Manifestation: Entering the next round immediately after execution, without checking the results
  • Correct approach: Effects must be verified with data, avoiding judgment based on feelings

Pitfall 2: Only Plan and Do, without continuous improvement

  • Manifestation: Ending after planning and execution, without performing Check and Act
  • Correct approach: PDCA is a cycle, not a linear process

Pitfall 3: Ignoring data collection and judging based on feelings

  • Manifestation: No quantitative data in the Check stage, subjectively judging "the effect is good"
  • Correct approach: Clarify the data collection plan in the Plan stage

Pitfall 4: Pilot scope too large, risks out of control

  • Manifestation: Full-scale rollout on the first attempt
  • Correct approach: Pilot on a small scale in the Do stage, then expand after successful verification

Pitfall 5: No standardization in the Act stage, loss of experience

  • Manifestation: Ending after improvement, without documentation
  • Correct approach: Write successful practices into standard processes

References

  • "The Essential Deming: Leadership Principles from the Father of Quality" - W. Edwards Deming
  • Lean Six Sigma - Continuous improvement methodology
  • Toyota Production System - Lean production
  • PDCA Cycle Practical Guide

Post-protocol

Artifact output

Save the PDCA cycle results to the artifact file:

TIMESTAMP=$(date +%Y%m%d-%H%M%S)
ARTIFACT_FILE="memory/artifacts/pdca-cycle/result-$TIMESTAMP.json"

cat > "$ARTIFACT_FILE" <<EOFJSON
{
  "skill": "pdca-cycle",
  "version": "2.0.0",
  "timestamp": "$(date -u +%Y-%m-%dT%H:%M:%SZ)",
  "project": "$PROJECT_ROOT",
  "branch": "$BRANCH",
  "commit": "$COMMIT",
  "input": {
    "user_request": "用户的原始请求"
  },
  "output": {
    "cycle_number": 1,
    "plan": {
      "objective": "",
      "actions": []
    },
    "do": {
      "completed_actions": [],
      "issues": []
    },
    "check": {
      "metrics": {},
      "findings": []
    },
    "act": {
      "decisions": [],
      "next_cycle": ""
    }
  },
  "next_skills": []
}
EOFJSON

echo "ARTIFACT SAVED: $ARTIFACT_FILE"
ln -sf "$ARTIFACT_FILE" memory/artifacts/pdca-cycle/latest.json

Target file update

If the target file exists, record the completion of the PDCA cycle.

Suggested follow-up skills

Based on the PDCA cycle results, it may be necessary to continue with the next cycle or complete the goal.