Cloud Cost Optimization Template

For SaaS companies, infrastructure costs typically range from 15-30% of revenue at early stage, declining to 5-15% at scale. Best-in-class SaaS companies operate at 5-8% of revenue. If your infrastructure cost exceeds 25% of revenue and is growing faster than revenue, optimization should be a priority. Track cost per customer, cost per API call, or cost per transaction as [unit economics metrics](/glossary/unit-economics) alongside total spend.

Yes. Infrastructure cost directly affects gross margin, which affects pricing strategy, profitability, and investment capacity. PMs should understand the cost implications of product decisions (feature X adds $5K/month in compute, feature Y requires a $20K/month managed service) and factor them into prioritization. Use the [RICE calculator](/tools/rice-calculator) to weigh cost optimization projects against feature work.

Use a proportional allocation model. If 5 services share a Kubernetes cluster, allocate cluster costs based on each service's resource requests (CPU and memory) as a percentage of total cluster capacity. Use namespace-level resource quotas and Kubecost or CloudHealth for automated allocation. Tag shared resources with all consuming teams and split the cost in your FinOps dashboard.

Spot instances work well for workloads that are fault-tolerant, stateless, and can handle 2-minute termination notices. Good candidates: batch processing, CI/CD build agents, data pipelines, dev/test environments, and stateless web workers behind a load balancer. Bad candidates: databases, single-instance services, anything that cannot restart gracefully. ---