Skip to content

METRICS_ARCHITECTURE.md

Latest commit

 

History

History
111 lines (77 loc) · 4.3 KB

METRICS_ARCHITECTURE.md

File metadata and controls

111 lines (77 loc) · 4.3 KB

Engineering Metrics Architecture (v3.0)

Version: 3.0
Last Updated: 2026-01-04
Scope: dbt Layered Analytics, Flow Framework, SPACE, DORA, Code Hotspots, Talent Radar, Capitalization Audit

1. Overview

This document defines the comprehensive engineering metrics system implemented in the DevOps platform. It features a robust dbt-based data warehouse architecture (Source -> Staging -> Intermediate -> Marts) to ensure data lineage, quality, and multi-dimensional analysis.

The system is built on four theoretical pillars:

  1. ELOC 2.0 (GitPrime Style): Quantifying engineer output with impact and rework contexts.
  2. DORA (Google): Measuring delivery speed (including wait/work breakdown) and stability.
  3. SPACE (Microsoft/GitHub): Balancing productivity across 5 dimensions.
  4. Value Stream Accounting: Automated R&D Capitalization (CapEx vs OpEx).

2. Core Metrics Definitions

2.1 ELOC 2.0 (Equivalent Lines of Code)

Standardizes "effort" by weighting code changes based on context, reducing the bias of raw LOC.

Metric Business Value Logic
ELOC Score Coding volume. (Additions + Deletions) * FileWeight * ContextWeight
Impact Score Value of work. ELOC Score * LegacyFactor
Churn Lines Rework/Waste. Lines modified within 21 days of previous commit.
Active Days Focus consistency. Count of distinct days with at least one commit.

2.2 DORA Metrics (Refined)

Moving beyond totals to lifecycle bottleneck identification.

  • Deployment Frequency: Success releases to production per month.
  • Lead Time for Changes (Refined):
    • Pickup Delay: Time from MR creation to first human review. (Measures Responsiveness).
    • Work Duration: Time from first review to merge. (Measures Collaboration Complexity).
  • Change Failure Rate: Success vs Failed deployments at production environment.
  • MTTR: Mean Time to Recovery (based on Incident-labeled issues).

2.3 Talent & Knowledge Radar

Quantifying technical leadership and organizational risk.

  • Ownership %: Calculated by int_file_knowledge_ownership based on historical code contributions per file/directory.
  • Bus Factor: Identifying sub-systems where a single contributor owns >80% of the knowledge.
  • Talent Archetypes:
    • Domain Specialist: High knowledge depth in specific repositories.
    • Collaborative Leader: High review count and cross-team impact.
    • Reliable Contributor: Consistent delivery output.

2.4 Code Hotspots (Michael Feathers F-C Analysis)

Identifying high-risk technical debt through fct_code_hotspots.

  • Risk Factor: Churn_90d * log(Estimated_LOC + 2).
  • Risk Zones:
    • RED_ZONE: High Churn + High Complexity. Critical Technical Debt.
    • AMBER_ZONE: Complex but stable (Core modules) or Simple but volatile.
    • CLEAR: Low risk maintenance files.

2.5 R&D Capitalization Audit

Automated financial classification of engineering effort.

  • CapEx (Capital Expenditure): Effort spent on Features, Requirements, and Epics.
  • OpEx (Operating Expenditure): Effort spent on Bugs, Refactoring (Technical Debt), and Support.
  • Audit Status: AUDIT_READY vs HIGH_CAPEX_INSPECTION_REQUIRED (for rate > 80%).

3. Implementation Architecture

3.1 Data Flow (dbt Layered)

graph TD
    Sources[(Raw DB)] --> Staging[Staging Layer: stg_]
    Staging --> Intermediate[Intermediate Layer: int_]
    Intermediate --> Marts[Marts Layer: dws_ / fct_]
    
    subgraph "Marts Layer"
        fct_dora[fct_dora_metrics]
        fct_hotspots[fct_code_hotspots]
        fct_talent[fct_talent_radar]
        fct_fin[fct_capitalization_audit]
    end
    
    Marts --> Dashboard[Streamlit Dashboard Pages]
Loading

3.2 Model Mapping Structure

Dashboard Page Primary dbt Model
1_DORA_Metrics.py fct_dora_metrics
6_Capitalization_Audit.py fct_capitalization_audit
8_Talent_Radar.py fct_talent_radar
15_Michael_Feathers_Code_Hotspots.py fct_code_hotspots

4. Key Configuration

  • DORA Benchmarks: Elite (< 1 day lead time), High (< 1 week), etc.
  • Bus Factor Threshold: Warning triggered when single-user ownership > 80%.
  • Risk Factor Ceiling: Files with Risk Factor > 40 are flagged as RED_ZONE.