返回提示词目录第 1454 页

编写研发交付指标月报

汇总交付速度、稳定性与工时分配趋势,形成含流程瓶颈、事故分析和下月改进事项的月报。

完整提示词

<context>
An engineering metrics report should help leaders see whether delivery is getting faster and safer and where the system slows teams down; used badly, it makes teams game numbers. Engineering leaders who report well follow a few principles. The four DORA measures (deployment frequency, lead time for changes, change failure rate, time to restore) are read together, because speed bought with more failures is not progress. Each measure depends on its definition (what counts as a deployment or a failure), so definitions are stated and changes to them are flagged. Cycle time is more useful split into stages (coding, waiting for review, in review, waiting to deploy), because waiting time is usually larger than working time and points at process fixes. Medians and percentiles matter, since a few long-running items distort averages. Teams differ in what they own (a mobile app with store review, a regulated payments service, a legacy platform), so comparing teams or ranking them is misleading and invites gaming; the useful comparison is each team against its own trend. Individual-level metrics such as commits or lines of code per person do not measure contribution and should not appear. Work allocation shows how much capacity goes to unplanned work and maintenance, often the real reason delivery slows.

{{#context}}Context: {{context}}{{/context}}
</context>

<task>
<engineering_data>
{{engineering_data}}
</engineering_data>

1. State the definition and source of each metric and flag any definition or tooling change that breaks the trend.
2. Report the four DORA measures for the organisation and per team against each team's own trend, using medians where distributions are skewed.
3. Break cycle time into stages, find the stage with the most waiting, and suggest the likely process cause as a hypothesis.
4. Report incidents by severity, time to restore and recurring causes, and link change failures to their sources where the data shows it.
5. Report work allocation and how unplanned work moved.
6. List what teams or leadership are changing next month, each tied to a figure, and what will be watched to see if it worked.
</task>

<constraints>
- Do not rank teams against each other or present a league table; compare each team with its own history and note context differences.
- Do not report individual-level productivity metrics; refuse requests to do so and explain why.
- Read speed and stability together; never praise a speed gain without the failure rate beside it.
- Do not invent benchmarks or performance tiers unless supplied with their source.
- Mark causes as hypotheses unless the data shows them.
{{> output/uncertainty}}
</constraints>

<output_format>
## Summary
Three to five sentences: speed and stability trend, the biggest bottleneck, the main reliability issue, the main change planned.

## Delivery metrics
Table: Measure | This month | Last month | Three-month trend | Definition. Per-team trends as a separate table without ranking.

## Flow and cycle time
Table: Stage | Median | 85th percentile | Change. Bottleneck and hypothesis.

## Reliability and incidents
Incidents by severity, time to restore, recurring causes.

## Where time went
Table: Work type | Share this month | Last month.

## What we are changing
Numbered changes with the figure that motivates each and the signal to watch.

## Data notes
Bullets: definitions, tool changes, gaps.
</output_format>

可替换内容

  • {{#context}}
  • {{context}}
  • {{/context}}
  • {{engineering_data}}
  • {{> output/uncertainty}}

分类与标签

办公效率、研发月报、交付指标、流程瓶颈、事故复盘

前往 AI 图片创作