Why Unit Economics Change Everything
When engineering teams own a "cost per active user" KPI, optimization decisions become obvious. Caching a database query isn't just good engineering — it's measurably reducing cost per user by $0.003. That's motivating in a way that "reduce cloud spend" isn't.
Unit economics also enable better growth decisions. If acquiring a new customer costs $120 in CAC and your cloud cost per customer is $8/month, scaling 10x costs are predictable. Without unit economics, every growth decision is a financial unknown.
Key Cloud Unit Economic Metrics
| Cost Per Active User (CPAU) | Cloud spend / Monthly Active Users |
| Cost Per Transaction | Cloud spend / Transactions processed |
| Infrastructure Cost Ratio (ICR) | Cloud spend / Revenue (%) |
| Cost Per GB Stored | Storage cost / GB of customer data |
| Cost Per API Call | Cloud spend / API requests handled |
| Compute Cost Per Order | EC2/container cost / Orders fulfilled |
Calculating Cost Per Customer
Formula: Total cloud spend / Monthly Active Customers
The challenge: allocating shared costs. Start with direct costs (services that serve customers directly: web servers, databases, CDN). Allocate shared infrastructure proportionally. Exclude internal tooling and dev/test environments from customer-serving cost.
Infrastructure Cost Ratio (ICR)
ICR = Cloud Spend / Revenue. Define which cloud costs and revenue are included, and use a consistent period and denominator.
- Compare the trend for the same product or business model over time.
- Explain changes in workload mix, growth, architecture and capacity commitments.
- Record whether an AI-heavy workload is training, inference or another operation before comparing it with a different product.
- Use your own measured baseline rather than an unsourced industry percentage target.
A rising ICR as you scale is a warning sign — it means your cloud costs are growing faster than revenue. A falling ICR is the goal: each additional dollar of revenue requires less marginal cloud spend.