搜索

PythonReactSEOPDF
GitHub Skill

metrics-dashboard

KPI and metrics dashboard workflow covering metric definition, data sourcing, visualization design, and anomaly detection. Delivers actionable dashboards.

LerianStudio☆ 86 仓库 Stars10 仓库 Forks

安装与使用

请根据来源仓库和你的 AI 工具说明安装此技能。先阅读技能文件和依赖要求,再添加到工作流。

查看来源 ↗使用指南

技能说明

Metrics Dashboard Workflow

This skill provides a structured workflow for designing KPI dashboards using the metrics-analyst agent.

Workflow Overview

The metrics dashboard workflow follows 5 phases:

Phase Name Description
1 Requirements Define dashboard objectives and audience
2 KPI Design Define metrics and methodology
3 Data Architecture Map data sources and calculations
4 Visualization Design visual presentation
5 Implementation Build and validate

Phase 1: Requirements

MANDATORY: Define dashboard objectives before building

Questions to Answer
Question Purpose
What decisions will this support? Ensures relevance
Who is the primary audience? Tailors complexity
What frequency of update? Sets refresh requirements
What level of drill-down? Scopes depth
What benchmark comparisons? Defines targets
Dashboard Types
Type Audience Focus
Executive C-Suite High-level, strategic
Operational Managers Detailed, actionable
Departmental Department heads Function-specific
Board Directors Governance, strategic
Blocker Check

If ANY of these are unclear, STOP and ask:

  • Dashboard purpose
  • Primary audience
  • Key decisions supported
  • Update frequency required

Phase 2: KPI Design

MANDATORY: Define all metrics with methodology

KPI Definition Standard
Element Requirement
Name Clear, concise name
Definition Precise description
Formula Exact calculation
Unit Measurement unit
Target Performance target
Owner Accountable person
Frequency Update cadence
KPI Categories
Category Example KPIs
Financial Revenue, margin, EBITDA, cash flow
Operational Throughput, cycle time, utilization
Customer Retention, NPS, LTV, CAC
Growth ARR growth, customer growth, expansion
KPI Selection Principles
Principle Description
Relevance Supports specific decisions
Measurable Can be quantified objectively
Actionable Drives specific actions
Timely Available when needed
Owned Clear accountability

Phase 3: Data Architecture

MANDATORY: Document data lineage completely

Data Source Mapping
Element Documentation
Source system Where data originates
Extraction method How data is obtained
Transformation Any calculations or adjustments
Refresh frequency How often updated
Data quality Validation checks
Data Quality Requirements
Check Validation
Completeness All required data present
Accuracy Data matches source
Timeliness Data is current
Consistency Data consistent across sources

Phase 4: Agent Dispatch

Dispatch to specialist with full context

Agent Dispatch
Task tool:
  subagent_type: "ring:metrics-analyst"
  model: "opus"
  prompt: |
    Design metrics dashboard per these specifications:

    **Purpose**: [from Phase 1]
    **Audience**: [from Phase 1]
    **Update Frequency**: [from Phase 1]

    **KPIs Required**:
    [From Phase 2 - list with definitions]

    **Data Sources**:
    [From Phase 3 - source mapping]

    **Required Output**:
    - KPI definitions with formulas
    - Data source documentation
    - Calculation methodology
    - Visualization specifications
    - Anomaly thresholds
    - Implementation guide
Required Output Elements
Element Requirement
Metrics Summary Dashboard overview
KPI Definitions Complete definitions
Data Sources Source documentation
Calculation Methodology Formula details
Dashboard Design Visual specifications
Anomaly Analysis Threshold definitions
Recommendations Enhancement suggestions

Phase 5: Implementation

MANDATORY: Validate before deployment

Implementation Checklist
Check Validation
Data feeds working All sources connected
Calculations verified Outputs match expected
Visuals rendering Display correctly
Refresh working Updates as expected
Access controlled Right users have access
Validation Tests
Test Description
Data reconciliation Dashboard ties to source
Historical comparison Trends make sense
Edge cases Handles nulls, zeros
Performance Loads in acceptable time

Pressure Resistance

See shared-patterns/pressure-resistance.md for universal pressures.

Dashboard-Specific Pressures
Pressure Type Request Agent Response
"Just show the numbers" "Numbers without methodology cannot be trusted. I'll include documentation."
"Pick the most important KPIs" "KPI selection requires business input. Which decisions should these support?"
"Skip the data quality checks" "Unreliable data undermines dashboard value. I'll validate all sources."
"Copy the existing dashboard" "Each dashboard needs fresh design. I'll validate requirements."

Anti-Rationalization Table

See shared-patterns/anti-rationalization.md for universal anti-rationalizations.

Dashboard-Specific Anti-Rationalizations
Rationalization Why It's WRONG Required Action
"Everyone knows what revenue means" Definitions vary DEFINE specifically
"Data source is obvious" Lineage needs documentation DOCUMENT source
"Calculation is standard" Standard still needs documentation SHOW formula
"Refresh frequency doesn't matter" Stale data causes bad decisions SPECIFY frequency

Execution Report

Upon completion, report:

Metric Value
Duration Xm Ys
KPIs Defined N
Data Sources Mapped N
Visualizations Designed N
Anomaly Thresholds N
Result COMPLETE/PARTIAL
Quality Indicators
Indicator Status
All KPIs defined YES/NO
All sources documented YES/NO
All calculations shown YES/NO
Data validated YES/NO
Refresh tested YES/NO

使用示例

查看来源仓库中的使用示例与配置说明。

查看来源 ↗

来源历史

收录时间
2026-01-20
索引更新时间
2026-02-04
Branch
main

收录与同步日期属于本站记录。源码修改历史请查看来源仓库。

GitHub ↗

chat Comments (0)

chat_bubble_outline

No comments yet. Be the first to share your thoughts!