Scrum Master Expert
Data-driven Scrum Master skill combining sprint analytics, probabilistic forecasting, and team development coaching. The unique value is in the three Python analysis scripts and their workflows — refer to references/ and assets/ for deeper framework detail.
Table of Contents
Analysis Tools & Usage
1. Velocity Analyzer (scripts/velocity_analyzer.py)
Runs rolling averages, linear-regression trend detection, and Monte Carlo simulation over sprint history.
CODEBLOCK0
Outputs: velocity trend (improving/stable/declining), coefficient of variation, 6-sprint Monte Carlo forecast at 50 / 70 / 85 / 95% confidence intervals, anomaly flags with root-cause suggestions.
Validation: If fewer than 3 sprints are present in the input, stop and prompt the user: "Velocity analysis needs at least 3 sprints. Please provide additional sprint data." 6+ sprints are recommended for statistically significant Monte Carlo results.
2. Sprint Health Scorer (scripts/sprint_health_scorer.py)
Scores team health across 6 weighted dimensions, producing an overall 0–100 grade.
| Dimension | Weight | Target |
|---|
| Commitment Reliability | 25% | >85% sprint goals met |
| Scope Stability |
20% | <15% mid-sprint changes |
| Blocker Resolution | 15% | <3 days average |
| Ceremony Engagement | 15% | >90% participation |
| Story Completion Distribution | 15% | High ratio of fully done stories |
| Velocity Predictability | 10% | CV <20% |
CODEBLOCK1
Outputs: overall health score + grade, per-dimension scores with recommendations, sprint-over-sprint trend, intervention priority matrix.
Validation: Requires 2+ sprints with ceremony and story-completion data. If data is missing, report which dimensions cannot be scored and ask the user to supply the gaps.
3. Retrospective Analyzer (scripts/retrospective_analyzer.py)
Tracks action-item completion, recurring themes, sentiment trends, and team maturity progression.
CODEBLOCK2
Outputs: action-item completion rate by priority/owner, recurring-theme persistence scores, team maturity level (forming/storming/norming/performing), improvement-velocity trend.
Validation: Requires 3+ retrospectives with action-item tracking. With fewer, note the limitation and offer partial theme analysis only.
Input Requirements
All scripts accept JSON following the schema in assets/sample_sprint_data.json:
CODEBLOCK3
Jira and similar tools can export sprint data; map exported fields to this schema before running the scripts. See assets/sample_sprint_data.json for a complete 6-sprint example and assets/expected_output.json for corresponding expected results (velocity avg 20.2 pts, CV 12.7%, health score 78.3/100, action-item completion 46.7%).
Sprint Execution Workflows
Sprint Planning
- 1. Run velocity analysis: INLINECODE8
- Use the 70% confidence interval as the recommended commitment ceiling for the sprint backlog.
- Review the health scorer's Commitment Reliability and Scope Stability scores to calibrate negotiation with the Product Owner.
- If Monte Carlo output shows high volatility (CV >20%), surface this to stakeholders with range estimates rather than single-point forecasts.
- Document capacity assumptions (leave, dependencies) for retrospective comparison.
Daily Standup
- 1. Track participation and help-seeking patterns — feed ceremony data into
sprint_health_scorer.py at sprint end. - Log each blocker with date opened; resolution time feeds the Blocker Resolution dimension.
- If a blocker is unresolved after 2 days, escalate proactively and note in sprint data.
Sprint Review
- 1. Present velocity trend and health score alongside the demo to give stakeholders delivery context.
- Capture scope-change requests raised during review; record as scope-change events in sprint data for next scoring cycle.
Sprint Retrospective
- 1. Run all three scripts before the session:
python sprint_health_scorer.py sprint_data.json --format text > health.txt
python retrospective_analyzer.py sprint_data.json --format text > retro.txt
- 2. Open with the health score and top-flagged dimensions to focus discussion.
- Use the retrospective analyzer's action-item completion rate to determine how many new action items the team can realistically absorb (target: ≤3 if completion rate <60%).
- Assign each action item an owner and measurable success criterion before closing the session.
- Record new action items in
sprint_data.json for tracking in the next cycle.
Team Development Workflow
Assessment
CODEBLOCK5
- - Map retrospective analyzer maturity output to the appropriate development stage.
- Supplement with an anonymous psychological safety pulse survey (Edmondson 7-point scale) and individual 1:1 observations.
- If maturity output is
forming or storming, prioritise safety and conflict-facilitation interventions before process optimisation.
Intervention
Apply stage-specific facilitation (details in references/team-dynamics-framework.md):
| Stage | Focus |
|---|
| Forming | Structure, process education, trust building |
| Storming |
Conflict facilitation, psychological safety maintenance |
| Norming | Autonomy building, process ownership transfer |
| Performing | Challenge introduction, innovation support |
Progress Measurement
- - Sprint cadence: re-run health scorer; target overall score improvement of ≥5 points per quarter.
- Monthly: psychological safety pulse survey; target >4.0/5.0.
- Quarterly: full maturity re-assessment via retrospective analyzer.
- If scores plateau or regress for 2 consecutive sprints, escalate intervention strategy (see
references/team-dynamics-framework.md).
Key Metrics & Targets
| Metric | Target |
|---|
| Overall Health Score | >80/100 |
| Psychological Safety Index |
>4.0/5.0 |
| Velocity CV (predictability) | <20% |
| Commitment Reliability | >85% |
| Scope Stability | <15% mid-sprint changes |
| Blocker Resolution Time | <3 days |
| Ceremony Engagement | >90% |
| Retrospective Action Completion | >70% |
Limitations
- - Sample size: fewer than 6 sprints reduces Monte Carlo confidence; always state confidence intervals, not point estimates.
- Data completeness: missing ceremony or story-completion fields suppress affected scoring dimensions — report gaps explicitly.
- Context sensitivity: script recommendations must be interpreted alongside organisational and team context not captured in JSON data.
- Quantitative bias: metrics do not replace qualitative observation; combine scores with direct team interaction.
- Team size: techniques are optimised for 5–9 member teams; larger groups may require adaptation.
- External factors: cross-team dependencies and organisational constraints are not fully modelled by single-team metrics.
Related Skills
- - Agile Product Owner (
product-team/agile-product-owner/) — User stories and backlog feed sprint planning - Senior PM (
project-management/senior-pm/) — Portfolio health context informs sprint priorities
For deep framework references see references/velocity-forecasting-guide.md and references/team-dynamics-framework.md. For template assets see assets/sprint_report_template.md and assets/team_health_check_template.md.Scrum Master 专家
数据驱动的Scrum Master技能，结合了冲刺分析、概率预测和团队发展辅导。其独特价值在于三个Python分析脚本及其工作流程——有关更深入的框架细节，请参阅references/和assets/目录。
目录
分析工具与使用
1. 速度分析器（scripts/velocity_analyzer.py）
对冲刺历史数据执行滚动平均值、线性回归趋势检测和蒙特卡洛模拟。
bash
文本报告
python velocity
analyzer.py sprintdata.json --format text
JSON输出，用于下游处理
python velocity
analyzer.py sprintdata.json --format json > analysis.json
输出：速度趋势（改善中/稳定/下降）、变异系数、50/70/85/95%置信区间下的6个冲刺蒙特卡洛预测、异常标记及根本原因建议。
验证：如果输入中少于3个冲刺，则停止并提示用户：速度分析至少需要3个冲刺。请提供额外的冲刺数据。 建议使用6个或更多冲刺以获得统计上显著的蒙特卡洛结果。
2. 冲刺健康评分器（scripts/sprinthealthscorer.py）
在6个加权维度上对团队健康度进行评分，生成0-100的总体评分。
| 维度 | 权重 | 目标 |
|---|
| 承诺可靠性 | 25% | >85%的冲刺目标达成 |
| 范围稳定性 |
20% | <15%的冲刺中变更 |
| 障碍解决 | 15% | 平均<3天 |
| 仪式参与度 | 15% | >90%的参与率 |
| 故事完成分布 | 15% | 完全完成的故事比例高 |
| 速度可预测性 | 10% | CV <20% |
bash
python sprinthealthscorer.py sprint_data.json --format text
输出：总体健康评分+等级、各维度评分及建议、冲刺间趋势、干预优先级矩阵。
验证：需要2个或更多包含仪式和故事完成数据的冲刺。如果数据缺失，报告哪些维度无法评分，并要求用户补充缺失数据。
3. 回顾分析器（scripts/retrospective_analyzer.py）
追踪行动项完成情况、重复出现的主题、情绪趋势和团队成熟度进展。
bash
python retrospectiveanalyzer.py sprintdata.json --format text
输出：按优先级/负责人统计的行动项完成率、重复主题持久性评分、团队成熟度等级（形成期/风暴期/规范期/表现期）、改进速度趋势。
验证：需要3个或更多带有行动项追踪的回顾会议。如果少于3个，则说明局限性，仅提供部分主题分析。
输入要求
所有脚本接受遵循assets/samplesprintdata.json中模式的JSON数据：
json
{
teaminfo: { name: string, size: number, scrummaster: string },
sprints: [
{
sprint_number: number,
planned_points: number,
completed_points: number,
stories: [...],
blockers: [...],
ceremonies: {...}
}
],
retrospectives: [
{
sprint_number: number,
went_well: [string],
to_improve: [string],
action_items: [...]
}
]
}
Jira及类似工具可以导出冲刺数据；在运行脚本前，将导出的字段映射到此模式。请参阅assets/samplesprintdata.json获取完整的6个冲刺示例，以及assets/expected_output.json获取相应的预期结果（平均速度20.2点，CV 12.7%，健康评分78.3/100，行动项完成率46.7%）。
冲刺执行工作流程
冲刺规划
- 1. 运行速度分析：python velocityanalyzer.py sprintdata.json --format text
- 使用70%置信区间作为冲刺待办事项的推荐承诺上限。
- 审查健康评分器的承诺可靠性和范围稳定性评分，以校准与产品负责人的协商。
- 如果蒙特卡洛输出显示高波动性（CV >20%），则向利益相关者展示范围估计而非单点预测。
- 记录容量假设（休假、依赖关系），供回顾时比较。
每日站会
- 1. 追踪参与度和寻求帮助的模式——在冲刺结束时将仪式数据输入sprinthealthscorer.py。
- 记录每个障碍及其开始日期；解决时间将影响障碍解决维度。
- 如果障碍在2天后仍未解决，主动上报并在冲刺数据中记录。
冲刺评审
- 1. 在演示的同时展示速度趋势和健康评分，为利益相关者提供交付背景。
- 捕获评审期间提出的范围变更请求；在冲刺数据中记录为范围变更事件，供下一个评分周期使用。
冲刺回顾
- 1. 在会议前运行所有三个脚本：
bash
python sprint
healthscorer.py sprint_data.json --format text > health.txt
python retrospective
analyzer.py sprintdata.json --format text > retro.txt
- 2. 以健康评分和标记最多的维度开场，聚焦讨论。
- 使用回顾分析器的行动项完成率来确定团队实际能吸收多少新行动项（目标：如果完成率<60%，则≤3个）。
- 在会议结束前，为每个行动项指定负责人和可衡量的成功标准。
- 将新的行动项记录在sprint_data.json中，以便在下一个周期进行追踪。
团队发展工作流程
评估
bash
python sprinthealthscorer.py teamdata.json > healthassessment.txt
python retrospectiveanalyzer.py teamdata.json > retro_insights.txt
- - 将回顾分析器的成熟度输出映射到相应的发展阶段。
- 辅以匿名心理安全脉搏调查（Edmondson 7点量表）和个别1:1观察。
- 如果成熟度输出为形成期或风暴期，则在流程优化之前优先进行安全性和冲突促进干预。
干预
应用特定阶段的促进方法（详情见references/team-dynamics-framework.md）：
冲突促进、心理安全维护 |
| 规范期 | 自主性建设、流程所有权转移 |
| 表现期 | 挑战引入、创新支持 |
进展衡量
- - 冲刺节奏：重新运行健康评分器；目标每季度总体评分提升≥5分。
- 月度：心理安全脉搏调查；目标>4.0/5.0。
- 季度：通过回顾分析器进行完整的成熟度重新评估。
- 如果评分连续2个冲刺停滞或下降，则升级干预策略（见references/team-dynamics-framework.md）。
关键指标与目标
>4.0/5.0 |
| 速度CV（可预测性） | <20% |
| 承诺可靠性 | >85% |
| 范围稳定性 | <15%的冲刺中变更 |
| 障碍解决时间 | <3天 |
| 仪式参与度 | >90% |
| 回顾行动完成率 | >70% |
局限性
- - 样本量：少于6个冲刺会降低蒙特卡洛置信度——始终说明置信区间，而非点估计。
- 数据完整性：缺失仪式或故事完成字段会抑制受影响的评分维度——明确报告缺口。
- 背景敏感性：脚本建议必须结合JSON数据中未捕获的组织和团队背景进行解读。
- 量化偏差：指标不能替代定性观察；将评分与直接的团队互动相结合。
- 团队规模：这些技术针对5-9人团队进行了优化；较大群体可能需要调整。
- 外部因素：跨团队依赖和组织约束无法被单一团队指标完全建模。
相关技能
- - 敏捷产品负责人（product-team/agile-product-owner/）——用户故事和待办事项为冲刺规划提供输入
- 高级项目经理（project-management/senior-pm/）——项目组合健康背景为冲刺优先级提供参考
有关深入的框架参考，请参阅references/velocity-forecasting-guide.md和references/team-dynamics-framework.md。有关模板资产，请参阅assets/sprintreporttemplate.md和assets/teamhealthcheck_template.md。
