Series ID: EDKS-BASIC-ZH-0110 · Lesson No.110 · Expert Maintenance: Data Communication
Communicating Metrics, Dashboards and Trends | 解释指标、仪表板与趋势
A metric is useful only when people understand what it measures, how it is calculated, what changed, and what decision follows.
Dashboards can create an illusion of certainty. Numbers look precise, colours look decisive, and trend lines look meaningful. Expert communication slows that down: define the metric, check the denominator, compare with baseline, distinguish signal from noise, and explain what the trend can and cannot support.
专家级数据沟通不是“把图表读一遍”。真正的工作是让读者知道：这个数字到底测什么？和什么比较？变化有多大？是趋势还是波动？对下一步决策意味着什么？
Chinese Edition Hub · 中文版入口 · ← Lesson No.109 · Writing Playbooks, SOPs and Decision Guides
The metric-communication map | 指标沟通地图
- metric
- KPI
- baseline
- target
- numerator/denominator
- trend
- variation
- distribution
- threshold
- decision
1. Metric | 指标
A metric is a defined measurement used to describe some aspect of performance or state.
2. KPI | 关键绩效指标
A KPI is a metric selected because it matters directly to an important objective.
3. Not every metric should be a KPI | 不是每个数字都应该是 KPI
Too many “key” metrics make priorities unclear.
4. Define the metric | 定义指标
State:
- name
- formula
- unit
- frequency
- source
5. Numerator | 分子
What is being counted?
6. Denominator | 分母
Out of how many opportunities, users, hours or cases?
7. Rate | 比率
30 errors means little without workload.
30 errors in 10,000 transactions gives a rate.
8. Percentage point vs percent change | 百分点 vs 百分比变化
From 2% to 3% = +1 percentage point = +50% relative increase.
9. Absolute and relative change | 绝对与相对变化
Use both when one alone could mislead.
10. Baseline | 基线
The starting reference point.
11. Benchmark | 基准
A comparison reference such as prior performance, peers or a standard.
12. Target | 目标
A desired future level.
13. Target ≠ forecast | 目标不等于预测
We want 95% is not the same as we expect 95%.
14. Target ≠ threshold | 目标不等于阈值
A threshold may trigger action; a target may be aspirational.
15. Leading indicator | 领先指标
May change before the outcome of interest.
16. Lagging indicator | 滞后指标
Reflects outcomes after the process has occurred.
17. Leading is not automatically better | 领先不一定更好
It may be faster but less directly tied to final outcome.
18. Dashboard | 仪表板
A dashboard should help a user notice state, change, risk or required action.
19. Dashboard is not a data warehouse | 仪表板不是所有数据的仓库
Include only what supports the user job.
20. Metric hierarchy | 指标层级
- headline KPI
- diagnostic metric
- operational detail
21. Headline metric | 核心指标
Answers: Are we broadly on track?
22. Diagnostic metric | 诊断指标
Answers: Why did the headline metric move?
23. Trend | 趋势
A trend is sustained directional movement, not one isolated point.
24. Variation | 波动
Some movement is normal noise.
25. Do not narrate every wiggle | 不要解释每一个小波动
Ask whether the change is meaningful relative to normal variation.
26. Time window | 时间窗口
One week, one month and one year can tell different stories.
27. Window selection can frame the result | 时间窗口会改变叙事
State why the chosen period is relevant.
28. Seasonality | 季节性
Some patterns repeat by day, week, month or season.
29. Compare like with like | 同类时间比较
Compare holiday periods with appropriate historical periods where relevant.
30. Rolling average | 滚动平均
Can smooth short-term noise and reveal broader movement.
31. Smoothing can hide sudden events | 平滑也会隐藏突变
Keep raw data accessible where spikes matter.
32. Average | 平均值
Useful, but can be distorted by extreme values.
33. Median | 中位数
Can better represent a skewed distribution.
34. Distribution | 分布
Averages can hide important subgroups.
35. Segment | 分群
Break down by:
- region
- product
- age group
- task type
- channel
36. Beware tiny segments | 小样本分群要谨慎
Small groups can produce unstable percentages.
37. Missing data | 缺失数据
A dashboard should distinguish zero from missing.
38. Data freshness | 数据时效
State when the data were last updated.
39. Stale data | 过期数据
A real-time-looking dashboard can contain yesterday’s data.
40. Data quality | 数据质量
Check:
- completeness
- consistency
- accuracy
- timeliness
41. Threshold colour | 阈值颜色
Red/amber/green should be tied to defined thresholds, not decoration.
42. Threshold rationale | 阈值理由
Why does 90% turn green? State the logic.
43. Alert fatigue | 告警疲劳
Too many alerts make important alerts easier to ignore.
44. Metric movement vs causal explanation | 指标变化不等于原因
Sales fell after the change does not automatically mean the change caused the fall.
45. Correlation language | 相关表达
- moved together
- was associated with
- coincided with
46. Causal language | 因果表达
Use only with appropriate evidence.
47. Annotation | 注释
Mark known events on a time series:
- campaign launch
- policy change
- system outage
48. Annotation helps; it does not prove cause | 注释帮助理解但不证明因果
49. Trend sentence frame | 趋势句型
Conversion rose from 4.2% to 4.8% over six weeks, with most of the increase occurring after week four.
50. Qualification frame | 限定句型
The increase is concentrated in returning users and is not visible in new-user traffic.
51. Dashboard briefing structure | 仪表板汇报结构
- what changed
- why it matters
- likely drivers
- uncertainty
- next action
52. Do not read the dashboard aloud | 不要逐个念数字
Interpret relationships and decision relevance.
53. Executive dashboard language | 管理层仪表板语言
Lead with deviation, risk and decision.
54. Operational dashboard language | 操作型仪表板语言
Lead with current state, threshold and action owner.
55. Public dashboard language | 公众数据语言
Explain denominator, time period and uncertainty plainly.
56. Metric gaming | 指标游戏化
When people are strongly rewarded for a metric, behaviour may optimise the number rather than the real objective.
57. Metric + countermetric | 指标 + 反向指标
Pair speed with quality, growth with retention, output with error rate.
58. Balanced measurement | 平衡测量
One metric rarely captures a complex goal fully.
59. Mandarin transfer: “增长 10%”要说清楚相对还是绝对 | clarify change type
60. Mandarin transfer: “同比/环比”需要明确 comparison period | comparison period matters
61. Practice A | 练习 A
Take a dashboard with five metrics and identify one KPI, two diagnostics and two low-priority metrics.
62. Practice B | 练习 B
Rewrite a misleading relative percentage with absolute context.
63. Practice C | 练习 C
Write a 90-second dashboard briefing with one uncertainty and one next action.
64. Error clinic | 常见问题
| Problem | Repair |
|---|---|
| Metric undefined. | State formula/unit. |
| Percentage has no denominator. | Add denominator. |
| One point called a trend. | Use time series. |
| Average hides subgroup. | Inspect distribution. |
| Dashboard has no action. | Link threshold to decision. |
65. First weak link diagnosis | 第一个卡点诊断
- number misunderstood → metric definition.
- change exaggerated → baseline/denominator.
- trend uncertain → variation/time-window.
- cause overclaimed → evidence.
- dashboard unused → decision relevance.
66. Seven-day training cycle | 七天训练
| Day 1 | metric definitions | 指标定义 |
| Day 2 | rates/denominators | 比率分母 |
| Day 3 | trend/variation | 趋势波动 |
| Day 4 | segments/distributions | 分群分布 |
| Day 5 | thresholds | 阈值 |
| Day 6 | dashboard briefing | 汇报 |
| Day 7 | decision memo from dashboard | 综合 |
67. Self-test | 自测
Explain a dashboard by defining metrics, checking denominators and time windows, separating trends from noise, identifying relevant segments, and connecting thresholds to decisions.
68. For parents and teachers | 给家长和老师
Use familiar data: test scores, reading time, exercise logs or class attendance.
Teach learners to ask what the number means before evaluating whether it is good or bad.
69. Final real-world challenge | 最终真实任务
- Choose a dashboard.
- Define five metrics.
- Identify baseline/target.
- Check denominators.
- Identify one trend.
- Identify one noisy metric.
- Inspect one subgroup.
- State one uncertainty.
- Recommend one action.
- Deliver a two-minute briefing.
Next: Lesson No.111 | 下一课
The next lesson turns expert language toward people: coaching, feedback and performance conversations that separate observable behaviour from interpretation and convert criticism into a clear path forward.
Lesson No.111 · Coaching, Feedback and Performance Conversations · 教练、反馈与表现对话
Reference floor: expert-maintenance data communication. Metrics are treated as defined evidence for decisions, not self-explanatory numbers.
