diagnosis-health

name	diagnosis-health
description	Use when you need to diagnose product health. Automated product health diagnosis that collects multi-dimensional data and performs comprehensive scoring, trend prediction, and bottleneck identification, outputting a health report. Keywords: health score, product diagnosis, multi-dimensional scoring, health check, product health, health rating, product checkup, status check.

Automated Product Health Diagnosis 🤖

Core Principles

1. Health is a leading indicator, not a lagging report: The purpose of health scoring is not to record the past, but to predict the future — discover hidden risks before users perceive problems
2. Layered scoring with independent tracking: Performance/availability/satisfaction/business are scored independently across four layers, ensuring each layer can be independently located and tracked
3. Bottlenecks determine resource allocation: The ultimate goal of health diagnosis is to identify bottlenecks — resources should be directed wherever bottlenecks are found

Interaction Mode

🤖→👤 AI suggests, human approves

Input

Input Item	Type	Required	Source	Description
Performance data	JSON	Yes	APM/monitoring system → Performance data	Response time, throughput, resource utilization
Availability data	JSON	Yes	Monitoring system → Availability data	SLA achievement rate, MTTR, MTBF
User satisfaction	JSON	Yes	Feedback system → Satisfaction data	NPS, CSAT, feedback, complaints
Business metrics	JSON	Yes	Data analytics platform → Business metrics	Conversion rate, GMV, DAU/MAU, retention
Competitor dynamics	JSON	○	output/pm-monitoring/diagnosis-competition/competitor report	Competitor health comparison data

Execution Steps

Step 1: Data Collection & Standardization [Core]

Goal: Collect and standardize data across all dimensions

Data Source Mapping:

Dimension	Data Source	Collection Method
Performance	APM/monitoring system	API/export
Availability	Monitoring system/incident management	API/export
Satisfaction	Feedback system/survey tools	API/export
Business	Data analytics platform	SQL/API

Standardization Processing:

• Time alignment (unify to the same time window)
• Metric normalization (0-100 or 0-1 range)
• Outlier handling (remove/smooth)
• Missing value imputation (interpolation/mean)

Output Format:

raw_data:
  collected_at: {ISO8601}
  time_window:
    start: {ISO8601}
    end: {ISO8601}
  dimensions:
    performance:
      metrics: {...}
    availability:
      metrics: {...}
    satisfaction:
      metrics: {...}
    business:
      metrics: {...}
  data_quality:
    completeness: {percentage}
    accuracy: {percentage}
    freshness: {minutes}

Step 2: Dimension Scoring [Core]

Goal: Calculate scores for each health dimension

Scoring Method:

Dimension	Metric Weights	Scoring Algorithm
Performance	Response time 40%, Throughput 30%, Resources 30%	Baseline deviation scoring
Availability	SLA 50%, MTTR 25%, Failure frequency 25%	Target achievement scoring
Satisfaction	NPS 40%, CSAT 30%, Complaint rate 30%	Industry benchmark scoring
Business	Conversion rate 30%, Retention 30%, GMV 40%	Trend period-over-period scoring

Scoring Formula:

score = Σ(metric_weight × metric_score)
metric_score = min(100, (actual / baseline) × 100)

Output Format:

scores_by_dimension:
  performance:
    score: 85
    trend: improving | stable | declining
    metrics:
      - name: response_time_p95
        value: 120ms
        baseline: 150ms
        score: 125
  availability:
    score: 92
    trend: stable
    metrics:
      - name: sla_achievement
        value: 99.9%
        target: 99.95%
        score: 99.5
  satisfaction:
    score: 78
    trend: declining
    metrics:
      - name: nps
        value: 35
        benchmark: 40
        score: 87.5
  business:
    score: 88
    trend: improving
    metrics:
      - name: conversion_rate
        value: 4.2%
        baseline: 4.0%
        score: 105

Step 3: Trend Prediction [Core]

Goal: Predict future health trends based on historical data

Analysis Methods:

• Moving Average (MA)
• Exponential Smoothing (Holt-Winters)
• Trend Extrapolation (linear/polynomial regression)

Prediction Dimensions:

• 7-day trend prediction
• 30-day trend prediction
• Seasonality analysis (if applicable)

Output Format:

trend_analysis:
  prediction_horizon: 7d | 30d
  overall_trend:
    direction: up | down | stable
    change_rate: {percentage}
    confidence: {percentage}
  dimension_trends:
    - dimension: performance
      current: 85
      predicted_7d: 83
      predicted_30d: 80
      risk_alert: true | false

Step 4: Bottleneck Identification [Core]

Goal: Identify key bottlenecks affecting overall health

Analysis Methods:

• Dimension-weighted contribution analysis
• Period-over-period/year-over-year anomaly detection
• Competitor benchmarking analysis

Bottleneck Classification:

Level	Definition	Response Requirement
P0	Severe threat to business metrics	Immediate action
P1	Significant impact on user experience	Resolve this week
P2	Potential risk	Planned resolution
P3	Optimization item	Ongoing follow-up

Output Format:

bottlenecks:
  - id: BOT-001
    dimension: satisfaction
    severity: P1
    description: "NPS declining for 3 consecutive weeks"
    root_cause: "Payment flow conversion rate decline"
    impact_score: 8
    affected_metrics: [nps, conversion_rate]
    recommendation: "Optimize payment flow"

Step 5: Comprehensive Report [Core]

Goal: Generate a comprehensive health diagnosis report

Report Structure:

1. Executive Summary (Overall Score + Key Findings)
2. Detailed Scoring by Dimension
3. Trend Analysis
4. Bottleneck List
5. Improvement Recommendations

Output Depth Classification

Depth Level	Output Scope	Description
quick	Health score and issue list	Core conclusions + minimum viable output
standard	Complete output (current default)	Full output including all Step outputs
deep	Full diagnosis + health trend analysis + root cause deep inference + improvement priority roadmap	Complete output + extended analysis + deep inference

Output

Output File Path: output/pm-monitoring/diagnosis-health/ Output Schema:

{
  "type": "object",
  "required": ["report_id", "overall_score", "scores_by_dimension"],
  "properties": {
    "report_id": {"type": "string", "description": "Unique report identifier"},
    "generated_at": {"type": "string", "description": "Generation timestamp"},
    "period": {"type": "object", "description": "Evaluation period, including start and end times"},
    "overall_score": {"type": "number", "description": "Overall health score"},
    "score_trend": {"type": "string", "description": "Score trend: improving/stable/declining"},
    "scores_by_dimension": {"type": "object", "description": "Scores by dimension, including performance/availability/satisfaction/business"},
    "trend_analysis": {"type": "object", "description": "Trend prediction, including 7-day and 30-day predicted changes"},
    "bottlenecks": {"type": "array", "description": "Bottleneck list, including severity and dimension"},
    "recommendations": {"type": "array", "description": "Improvement recommendation list, including priority and expected impact"}
  }
}

├── {date}/
│   ├── overall_score.md
│   ├── dimension_scores.yaml
│   ├── trend_analysis.yaml
│   ├── bottlenecks.yaml
│   └── recommendations.md
└── latest/
    └── health_report.md

Health Report Output Format

health_diagnosis:
  report_id: {uuid}
  generated_at: {ISO8601}
  period:
    start: {ISO8601}
    end: {ISO8601}
  overall_score: 86
  score_trend: improving
  scores_by_dimension:
    performance: 85
    availability: 92
    satisfaction: 78
    business: 88
  trend_analysis:
    predicted_change_7d: -1
    predicted_change_30d: -3
  bottlenecks:
    - id: BOT-001
      severity: P1
      dimension: satisfaction
  recommendations:
    - priority: high
      action: "Optimize payment flow"
      expected_impact: "+5 NPS"

Decision Rules

Scenario	Decision Rule
Dimension score < 60	Flag as red alert, generate specialized diagnosis
Dimension score 60-75	Flag as yellow alert, add to improvement plan
Dimension score > 85	Flag as green, maintain monitoring
Predicted downward trend	Generate preventive recommendations
Data missing > 30%	Flag data quality issue, skip scoring for that dimension

Quality Checks

P0 Checks (must pass for quick/standard/deep)

• Data collection completeness ≥ 90%
• Score calculation accuracy (sample verification)

P1 Checks (must pass for standard/deep)

• Trend prediction deviation ±10%
• Bottleneck identification coverage ≥ 90%
• Actionable recommendation rate ≥ 80%
• Report generation timeliness (< 5 minutes)

P2 Checks (must pass for deep only)

• Extended analysis complete (deep inference and roadmap generated)
• Decision records complete (key decisions have rationale and alternatives)

Degradation Strategy

Upstream File Missing Degradation Plan

Missing Upstream Input	Degradation Plan	Output Impact
Performance data	User provides key performance metric values such as response time and throughput; AI scores directly	Performance dimension direct scoring result, lacking automated collection data
Availability data	User provides SLA achievement rate, incident count, etc.; AI scores directly	Availability dimension direct scoring result, lacking automated collection data
User satisfaction	User provides NPS/CSAT scores and complaint rate data; AI scores directly	Satisfaction dimension direct scoring result, lacking automated collection data
Business metrics	User provides conversion rate, GMV, DAU/MAU, retention, etc.; AI scores directly	Business dimension direct scoring result, lacking automated collection data
Competitor dynamics	Skip competitor benchmarking; bottleneck identification based solely on internal data	Health report without competitor benchmarking

Data Acquisition Instructions

When upstream files are missing, obtain necessary data through the following methods:

1. Multiple dimension data missing: Ask the user to provide key metric values for each dimension; the AI will score dimensions directly based on user-provided data, skipping automated collection and standardization steps
2. Partial dimension data missing: Score only the dimensions with available data; mark missing dimensions as "insufficient data"; calculate overall score based on available dimensions only and note confidence level
3. Competitor dynamics missing: Skip competitor benchmarking analysis; bottleneck identification relies solely on internal data trends and threshold judgments; recommend supplementing competitor data later to improve analysis

Output Validation Rules

Field Path	Type	Required	Description
report_id	string	Yes	Unique report identifier
overall_score	number	Yes	Overall health score, range 0-100
score_trend	string	Yes	Score trend, only improving/stable/declining allowed
scores_by_dimension	object	Yes	Scores by dimension, must include performance/availability/satisfaction/business
trend_analysis	object	No	Trend prediction, must include predicted_change_7d/predicted_change_30d
bottlenecks	array	No	Bottleneck list, each item must include id/severity/dimension
bottlenecks[].severity	string	Yes	Severity, only P0/P1/P2/P3 allowed

Upstream Change Response

Upstream Change Impact Table

Upstream Source	Change Type	Impact Scope	Response Action
APM/monitoring system	Performance metric definition change	Performance dimension scoring and baseline comparison	Recalculate scores based on new definition
Monitoring system	Availability data format change	Availability dimension scoring	Adapt to new format, supplement missing fields
Feedback system	Satisfaction metric change	Satisfaction dimension scoring	Update scoring algorithm and weights
Data analytics platform	Business metric definition change	Business dimension scoring and trend prediction	Recalculate scores and predictions based on new definition

Downstream Notification Mechanism Table

Downstream Consumer	Notification Condition	Notification Method	Notification Content
diagnosis-orchestrator	Health diagnosis completed	Output file updated	Diagnosis completion status and key bottlenecks
monitoring-pipeline	Health score anomaly	Write to output file	Anomalous dimensions and bottleneck details
iteration-retrospective	P0/P1 bottleneck identified	Write to output file	Bottleneck description and improvement recommendations