This article introduces sustainability proxy metrics as fine-grained alternatives to carbon emissions for tracking cloud workload efficiency and enabling continuous optimization cycles.

• Proxy metrics provide near-real-time, granular insights into compute, storage, and networking resource usage
• Normalization is essential: factor in instance types, vCPUs, and business metrics for meaningful comparisons
• Normalize resource usage by multiplying instance hours with vCPU counts or using EC2 Reserved Instance factors
• Calculate KPI ratios like Spot adoption hours divided by total vCPU hours
• Weight services differently using scoring systems to incentivize efficient resource choices
• Divide resource usage by business metrics (users, transactions, API calls) to track efficiency over time
• AWS Customer Carbon Footprint Tool provides high-level emissions; proxy metrics enable team-level optimization
• Real examples: Arm tracked vCPU hours per EDA job; Prime Video measured cost per concurrent streams

Sustainability proxy metrics enable organizations to establish accountability and drive continuous efficiency improvements across application teams through normalized, business-aligned KPIs.