DevOps

Your AWS bill went down. Did your cloud cost per customer?

Measure cloud cost per customer alongside your AWS bill. Allocate shared costs, account for commitments and check whether savings improve service economics.

Your AWS bill went down. Did your cloud cost per customer?

You cut the AWS bill, but fewer customers used the platform that month. To judge the saving, compare cloud cost per customer as well as total spending. You need the same comparison when a growing customer base pushes the bill up.

Choose a unit that reflects the service

For a subscription platform, an active customer account may be useful. For a billing system, completed billing runs may be better. For a wholesale ordering service, count orders the system processed without errors.

Define the denominator before you report it. Does an active customer mean a paying account, an account with a login, or one that completed a core workflow? Changing that definition changes the result even when infrastructure stays identical.

Track workload as well as account count if customers use the service at different volumes. Budget for the difference between serving a customer who processes ten orders and one who processes ten thousand.

Define the costs in your calculation

A starting calculation is:

Allocated cloud cost for the period ÷ completed business units for the same period.

For an illustrative platform spending AUD $6,000 in a month with 300 active customer accounts, the average is $20 per active account. If next month costs $5,400 but active accounts fall to 200, the average rises to $27. The bill fell while average unit cost increased.

Check fixed capacity and customer mix before attributing that increase to waste.

Cost groupA workable starting allocation
Dedicated customer infrastructureAssign directly to that customer
Shared application and databaseAllocate with a documented usage proxy
Development and stagingReport separately, or show an explicit allocation
Platform support infrastructureAllocate consistently and disclose the method
Credits and one-off adjustmentsShow separately so they do not disguise the operating trend

Make shared cost understandable

Start with an allocation method your team can explain. Invest in more detail for the expensive or disputed areas, where it could change a decision.

If you allocate shared database cost by successful transactions, label the result as a management estimate. Record the limits of that estimate: customers running expensive reports or storing large amounts of data may consume more than their transaction count suggests.

Separate total platform unit cost from the incremental cost of another customer. A shared database can make average cost fall as usage grows, until a capacity increase changes the picture.

Note

Use a consistent currency, period and treatment of commitments. Compare cash invoices separately from any amortised cost view, and explain the difference to finance.

Use AWS recommendations as inputs

AWS Cost Optimization Hub brings together recommendations such as rightsizing, idle-resource removal and commitment discounts, accounting for relevant commercial terms. Test the workload and check your billing arrangement before accepting an estimated saving.

For each accepted recommendation, identify the cost group you expect to reduce, the date and the effect on unit cost. Check for overlap before adding the estimates together.

Check the service after the saving

  • Compare similar reporting periods and explain customer or workload changes.
  • Confirm the cost change on the bill using the same accounting basis as the baseline.
  • Check request latency, error rates, queue delays and recovery capability.
  • Investigate high-cost customer segments before changing platform-wide capacity.

After an AWS cost optimisation review, track unit cost to check whether you retain the savings as the service grows. DevOps consulting can put the measurement into the operating routine.

Frequently asked questions

No. Cloud cost is one input. Gross margin depends on your accounting treatment and other costs of delivering the service, so reconcile the model with finance.

Start with reliable totals and a documented allocation. Add detailed metering where the likely improvement in decisions justifies its cost.

One that reflects the work the service does, such as paying accounts, completed billing runs or processed orders. Define it before you report it and keep the definition stable between periods.

If the number of units falls faster than spending, fixed capacity is spread across fewer customers. Check fixed capacity and customer mix before treating the increase as waste.

Find out what growth really costs your platform

MyTechTeam can connect cloud spending to workload, capacity and reliability decisions.