Methodology Specification v1.0

Sustainability Scoring Methodology

The mathematical, dimensional, and operational foundations of the VERDANT software sustainability intelligence engine.

1. Methodological Position

VERDANT produces an estimated comparative sustainability index, not a direct certified lifecycle carbon audit unless empirical telemetry and regional emissions models are supplied.

The primary objective is consistent, deterministic decision support: identifying resource-efficiency weaknesses, preventing idle overprovisioning, and prioritizing engineering optimizations.

2. The Six Architectural Dimensions

Overall Score $S = \sum (D_i \times W_i)$ where weights sum strictly to $1.00$ ($100\%$).

Energy Efficiency

Weight: 25% (0.25)

Evaluates operational kilowatt-hour draw, workload duty cycle, and power proportionality.

Mathematical FormulaS_energy = 0.35 × S_runtime + 0.35 × S_proportionality + 0.30 × S_consumption

Evaluated Inputs: Runtime hours/day, autoscaling dormancy, CPU duty cycle, measured/estimated monthly kWh.

Compute Efficiency

Weight: 20% (0.20)

Assesses virtual CPU core allocation rightsizing and utilization sweet spots (45%–75%).

Mathematical FormulaS_compute = 0.40 × S_utilization_sweetspot + 0.35 × S_rightsizing + 0.25 × S_elasticity

Evaluated Inputs: Allocated vCPU cores, average CPU load %, horizontal autoscaling, async non-blocking concurrency.

Memory Efficiency

Weight: 15% (0.15)

Measures working set density, memory leak prevention safeguards, and runtime heap hygiene.

Mathematical FormulaS_memory = 0.40 × S_ram_utilization + 0.35 × S_density + 0.25 × S_heap_hygiene

Evaluated Inputs: Allocated RAM (MB), active working set %, garbage collection parameters, leak safeguards.

Network Efficiency

Weight: 15% (0.15)

Measures HTTP response compactness, Brotli/Gzip wire compression, and edge CDN cache leverage.

Mathematical FormulaS_network = 0.40 × S_payload_size + 0.35 × S_cdn_hit_ratio + 0.25 × S_wire_compression

Evaluated Inputs: Daily requests, average response payload (KB), Brotli compression, edge cache hit %.

Storage Efficiency

Weight: 10% (0.10)

Evaluates dataset growth velocity, composite index coverage, and data lifecycle TTL rules.

Mathematical FormulaS_storage = 0.35 × S_growth_rate + 0.40 × S_indexing + 0.25 × S_retention_ttl

Evaluated Inputs: Database volume (GB), asset storage, monthly growth %, query optimization, retention days.

Carbon Efficiency

Weight: 15% (0.15)

Models regional electricity grid emissions factor (gCO2e/kWh), data center PUE, and clean energy.

Mathematical FormulaS_carbon = 0.45 × S_grid_intensity + 0.25 × S_datacenter_pue + 0.30 × S_renewable_match

Evaluated Inputs: Regional grid factor (15–650 gCO2e/kWh), cloud data center PUE (1.1–1.8), renewable energy matching %.

3. Normalization & Clamping Logic

Every metric is converted into a normalized $0–100$ scale using explicit upper and lower domain thresholds:

Lower-is-Better (e.g. Payload Size, Growth Rate):

score = clamp(100 × (upper - val) / (upper - lower), 0, 100)

Higher-is-Better (e.g. Cache Hit %, Renewable %):

score = clamp(100 × (val - lower) / (upper - lower), 0, 100)

4. Score Rating Classification

90–100 (Exceptional)

Software demonstrates world-class resource stewardship with minimal carbon and compute waste.

75–89.99 (Excellent)

Strong sustainability characteristics with high efficiency across primary resource dimensions.

60–74.99 (Good)

Balanced baseline performance with notable opportunities for architectural and data optimization.

40–59.99 (Needs Improvement)

Significant resource overhead, overprovisioning, or uncompressed network payloads detected.

0–39.99 (Critical)

Severe operational inefficiencies, heavy idle draw, or highly carbon-intensive infrastructure.

5. Assumptions & Limitations

Self-Reported Telemetry: In self-assessment modes, accuracy depends on the fidelity of input operational figures.
Grid Mix Dynamics: Regional electricity grid emissions vary hour-by-hour; VERDANT utilizes standardized annual regional intensity averages unless real-time marginal emissions feeds are connected.
Hardware Embodied Emissions: Scope 3 supply-chain manufacturing emissions for server hardware are outside the MVP scope and modeled through PUE amortizations.
Methodology Versioning: All historical assessment records preserve their respective methodology version tag (e.g. v1.0) to ensure backwards reproducibility.