From cloud bill to unit economics: cost per customer
Total cloud spend is a number finance cannot act on. Cost per customer, per environment and per feature turns infrastructure into a business conversation.
"Our cloud bill went up 18 percent" is not a useful sentence. "Our infrastructure cost per active customer fell 6 percent while volume grew 24 percent" is.
Step one: allocation
You cannot compute a unit cost without allocation. That means a tagging standard applied consistently — owner, environment, product, cost centre — and enforcement in infrastructure-as-code rather than in a policy document. Shared costs (networking, observability, control planes) get split by a documented rule. An imperfect rule applied consistently beats a perfect one applied never.
Step two: choose the denominator
Pick a metric the business already tracks: active customers, transactions processed, documents parsed, seats. The right denominator is the one that scales with the workload, not the one that flatters the chart.
Step three: publish it monthly
Unit economics only change behaviour when they are routine. One slide, same definition every month, alongside revenue metrics. Trends matter more than absolute precision.
What it unlocks
- Pricing decisions grounded in real marginal cost
- Gross margin forecasts that survive due diligence
- Architectural trade-offs argued with numbers instead of instinct
- Early warning when a feature quietly becomes unprofitable
A caution
Do not let the model become a project of its own. A rough unit cost, available every month, delivered to the people who make decisions, beats a precise model that takes two quarters to build.