Series ID: EDKS-BASIC-ZH-0118 · Lesson No.118 · Expert Maintenance: Change Communication
Writing Change Plans, Rollouts and Rollbacks | 编写变更计划、发布与回滚
A change plan is not a list of tasks. It is a controlled path from the current state to a new state, with evidence, owners, monitoring and a safe way back if the change fails.
Change communication must connect intent with execution. It tells people why the change is happening, what will change, what will not change, who owns each stage, how success is measured, what signals require a pause, and how rollback works.
高质量变更计划不是“把任务排成时间表”。它必须同时说明目标、范围、依赖、发布步骤、监测指标、停止条件、回滚路径和沟通对象。
Chinese Edition Hub · 中文版入口 · ← Lesson No.117 · Communicating Risk, Controls and Assurance
The change-plan map | 变更计划地图
- objective
- scope
- current state
- target state
- dependencies
- rollout
- monitoring
- stop rule
- rollback
- communications
1. Change objective | 变更目标
State what problem the change solves.
2. Current state | 当前状态
Describe the baseline before change.
3. Target state | 目标状态
Describe what should be true after successful change.
4. Scope | 范围
State what changes and what does not.
5. Out of scope | 不在范围
Prevent uncontrolled expansion.
6. Assumptions | 假设
State assumptions that materially affect timing or risk.
7. Dependencies | 依赖项
- systems
- vendors
- people
- approvals
- data
8. Readiness criteria | 准备就绪标准
What must be true before rollout begins?
9. Go/no-go decision | Go/No-Go 决策
Define who decides and which evidence is required.
10. Rollout | 发布/上线
Move from current to target state in controlled stages.
11. Big-bang rollout | 一次性发布
All users/systems change at once. High coordination, potentially high impact.
12. Phased rollout | 分阶段发布
Change selected groups or regions first.
13. Pilot | 试点
A limited deployment designed to reduce uncertainty.
14. Canary-style rollout | 小范围先行
A small percentage receives the change while monitoring continues.
15. Choose rollout by risk | 按风险选择发布方式
Higher uncertainty often favours staged deployment when technically and operationally feasible.
16. Rollout stage | 发布阶段
Each stage should have:
- scope
- owner
- entry criteria
- success criteria
- exit criteria
17. Success criteria | 成功标准
Define measurable indicators before rollout.
18. Guardrail metric | 护栏指标
A metric that must remain within acceptable limits while another metric improves.
19. Example | 例子
Improve speed, but error rate must stay below 2%.
20. Monitoring window | 监测窗口
How long must performance remain stable before expanding?
21. Stop rule | 停止规则
Define when rollout pauses.
22. Rollback trigger | 回滚触发器
Define when the old state should be restored.
23. Rollback | 回滚
A planned reversal, not improvised panic.
24. Rollback prerequisites | 回滚前置条件
Can the previous configuration/data/process be restored?
25. Rollback owner | 回滚责任人
Someone must have authority to initiate it.
26. Rollback time | 回滚时间
Estimate how long recovery requires.
27. Irreversible change | 不可逆变更
If rollback is impossible, risk controls and approval requirements should reflect that.
28. Change window | 变更窗口
Choose timing based on user impact, staffing and monitoring capacity.
29. Freeze period | 冻结期
Some organisations restrict changes during critical periods.
30. Communication plan | 沟通计划
Identify audiences:
- users
- operators
- managers
- support
- partners
31. Audience message | 受众消息
Different audiences need different detail, but the core facts must remain stable.
32. Before-change message | 变更前通知
State:
- what changes
- when
- expected impact
- required action
- support route
33. During-change update | 变更中更新
State status, deviations, next checkpoint and risk.
34. Post-change update | 变更后更新
State outcome, monitoring status and unresolved issues.
35. Do not announce success too early | 不要过早宣布成功
Use monitoring while stability is still being verified.
36. Change owner | 变更责任人
Coordinates the overall change.
37. Stage owner | 阶段责任人
Owns a specific action or phase.
38. Decision owner | 决策责任人
Approves proceed/pause/rollback.
39. Escalation path | 升级路径
Define whom to contact when thresholds are crossed.
40. Dependency failure | 依赖失败
State contingency if an upstream dependency is unavailable.
41. Change record | 变更记录
Record:
- what changed
- who approved
- when
- evidence
- outcome
42. Change log vs communication log | 变更日志 vs 沟通日志
Technical change and stakeholder communication may require different records.
43. Risk register | 风险登记
Link rollout risks to controls and owners.
44. Pre-mortem | 事前复盘
Imagine the change failed. Ask what likely caused it.
45. Pre-mortem is not prediction | 事前复盘不是预测
It is a structured way to surface failure modes.
46. Rehearsal | 演练
Test rollback and communication before a high-stakes change where feasible.
47. Dry run | 彩排
Walk through steps without making the production change.
48. Decision checkpoint | 决策检查点
At each stage ask:
- continue?
- pause?
- rollback?
49. Evidence at checkpoint | 检查点证据
Use predefined metrics rather than mood.
50. Change fatigue | 变更疲劳
Too many simultaneous changes can reduce adoption and increase confusion.
51. Training | 培训
If user behaviour must change, training is part of the rollout.
52. Documentation update | 文档更新
Change is incomplete if SOPs, requirements and support documents remain old.
53. Knowledge handoff | 知识交接
Support and operations teams need new-state knowledge before they own incidents.
54. Adoption | 采用
Technical deployment and user adoption are different outcomes.
55. Adoption metric | 采用指标
Measure whether intended users actually use the new process or feature.
56. Outcome metric | 结果指标
Measure whether the change solved the original problem.
57. Deployment success ≠ business success | 上线成功不等于目标成功
58. Review trigger | 复审触发器
Set a date or condition for post-change review.
59. Post-implementation review | 实施后复盘
Compare expected vs actual outcome.
60. Mandarin transfer: “上线”可用 rollout / launch / deployment | 依场景选择
61. Mandarin transfer: “回滚”在技术语境常为 rollback | 非技术流程也可用 revert/return
62. Practice A | 练习 A
Write a three-stage rollout with entry/exit criteria.
63. Practice B | 练习 B
Write one stop rule and one rollback trigger.
64. Practice C | 练习 C
Write user, executive and operator versions of the same change notice.
65. Error clinic | 常见问题
| Problem | Repair |
|---|---|
| Tasks listed, no target state. | Define objective. |
| No readiness criteria. | Add go/no-go evidence. |
| No rollback. | Design recovery path. |
| Success declared immediately. | Add monitoring window. |
| Docs remain old. | Update knowledge layer. |
66. First weak link diagnosis | 第一个卡点诊断
- change starts too early → readiness.
- bad result expands → stage/guardrail problem.
- failure hard to reverse → rollback design.
- users confused → communication/training.
- old practice returns → adoption/maintenance.
67. Seven-day training cycle | 七天训练
| Day 1 | objective/scope | 目标范围 |
| Day 2 | readiness | 准备 |
| Day 3 | rollout stages | 分阶段 |
| Day 4 | monitoring | 监测 |
| Day 5 | rollback | 回滚 |
| Day 6 | communications | 沟通 |
| Day 7 | full change plan | 综合 |
68. Self-test | 自测
Write a change plan with objective, readiness criteria, stages, owners, monitoring, stop rules, rollback, communication and review.
69. For parents and teachers | 给家长和老师
Use school examples such as introducing a new study routine or project workflow. Teach students to define success and a way back if the change does not work.
70. Final real-world challenge | 最终真实任务
- Choose one change.
- Define current/target state.
- List dependencies.
- Set readiness criteria.
- Design three rollout stages.
- Choose guardrails.
- Set stop rule.
- Write rollback.
- Write communications.
- Run post-change review plan.
Next: Lesson No.119 | 下一课
The next lesson develops governance language: how to record decisions, authority, rationale, challenge, ownership and review without hiding behind process or turning accountability into blame.
Lesson No.119 · Decision Logs, Governance and Accountability · 决策记录、治理与问责
Reference floor: expert-maintenance change communication. Controlled change requires explicit go/no-go evidence, monitoring and recovery.
