Unit Economics of a System
The cost to serve one unit of business value — a customer, an order, a request — which determines whether a system's economics improve or worsen with scale.
Total cloud spend is a number that provokes anxiety and supports no decision. Cost per unit is the number that tells you whether the business works.
The unit must be one the business already uses: cost per active customer per month, cost per order, cost per thousand API calls, cost per gigabyte ingested. Then the question becomes answerable: is this above or below what we charge, and is the trend improving?
What makes it revealing is the trend against scale. A healthy system's unit cost falls as volume grows, because fixed costs amortise. A unit cost that is flat means everything scales linearly and there is no operating leverage. A unit cost that rises with scale is the alarming case and it happens more often than expected — coordination overhead, cross-zone traffic, an n-squared retrieval pattern, or a per-tenant resource that was cheap at fifty tenants and is not at five thousand.
The architectural leverage this exposes is specific rather than general. Knowing that 60% of cost per order sits in one enrichment call directs effort precisely, whereas a mandate to reduce cloud spend by 20% produces indiscriminate cuts.
The prerequisite is cost allocation — tagging, per-tenant attribution, and the ability to divide spend by a business denominator. Without it, unit economics is estimated in a spreadsheet once a year and believed by nobody.