More paying customers can bring more revenue. How much of it you keep depends partly on what it costs to serve those customers. If cloud spending rises more slowly than revenue, it can help your margins. If a new customer needs far more capacity than the last one, your growth plan may need a different budget.
Before committing to a growth plan, you need an estimate of what those customers will cost to serve. That estimate can inform pricing and hiring decisions, but it is only useful if you understand the assumptions behind it.
Start by connecting your current cloud bill to how customers use the product.
Choose a measure that fits the business
Cloud cost per customer is straightforward: take the cloud costs you want to track and divide them by the customers served over the same period. Be clear about whether you mean paying accounts, active accounts or individual users. Changing that definition can make the number look better without changing the business.
For some products, cost per transaction is more useful. A customer who processes a handful of transactions puts a different load on the system from one who processes thousands. Pick a measure that helps explain both the work your product performs and how you charge for it. AWS recommends connecting workload costs to business outputs, such as reports produced or pages served.[1]
Keep the scope consistent too. A measure that includes production, development and testing answers a different question from one that includes production alone. Either can be useful if everyone knows what it covers. Show temporary credits separately so you can see the underlying cost when those credits run out.
Track the measure alongside revenue. A falling cloud cost per customer does not, by itself, prove that your margins are improving: discounts, support costs and other expenses may be moving in the opposite direction.
Today's average is only the starting point
Consider a simplified example with invented figures. A startup spends $4,000 a month serving 1,000 paying customer accounts. Its average cloud cost is $4 per account. Multiplying that by a target of 2,000 accounts gives a $8,000 estimate, but that estimate assumes costs grow in direct proportion to customers.
Suppose $2,500 of the bill pays for capacity that can support the target customer base. The remaining $1,500 varies with usage. If usage per customer and unit prices stay the same, doubling customers would double that variable portion to $3,000. The resulting monthly bill would be $5,500, or $2.75 per account.
In an alternative scenario, suppose the team finds that reaching the target also requires another $2,000 a month of capacity. The estimate becomes $7,500, or $3.75 per account. Both scenarios serve 2,000 accounts, but they need different capacity budgets.
These invented US dollar figures illustrate the calculation; they are not AWS prices or a savings forecast. To choose a capacity assumption for your own budget, you need evidence of how much work the current setup can handle.
Find the next increase in capacity
Different parts of the bill respond to growth in different ways. Some charges follow the amount of work performed or data retained. Other costs stay relatively steady within a range of usage, then increase when you add capacity. A larger database or another server can create that step.
Ask the team where it expects the next constraint to appear and what evidence supports that view. An estimate based on a recent test is different from an assumption based on a quiet month's average. AWS's cost modelling guidance recommends comparing costs under different predicted loads and accounting for workload cycles.[2]
You also need to know when spending must happen. Capacity required before a large customer launches affects cash planning differently from a cost that grows gradually with their use of the product. A change may involve engineering work as well as a larger monthly bill. Include that work in the decision, even if it sits outside the cloud invoice.
If the team does not know that threshold yet, record it as an open assumption and agree how to check it.
Double customers and usage separately
Two thousand customers will not necessarily use twice as much capacity as one thousand. New accounts may start slowly. A larger customer may bring more users, retain more data or run demanding features more often. Existing customers may also increase their usage while the number of accounts stays flat.
That makes customer mix part of the forecast. If you are moving from small teams towards larger organisations, check whether the average customer in your growth plan resembles the average customer behind today's bill. The same question applies when you introduce a feature that changes how much work the system performs.
Start with a scenario that keeps usage per customer steady. Then consider what changes if usage is heavier or customers arrive more quickly than expected. Keep the assumptions visible, including the revenue attached to each scenario. This helps you see whether pricing still supports the cost of serving the customers you want to win.
Take three questions to the team
1. What is driving our cloud bill?
Ask the team to explain the bill through the activity in your product. Knowing which service costs the most is useful, but you also need to understand which customer activity creates that cost. That connection helps you judge whether a change in spending follows expected growth, a product decision or something that needs investigation.
Focus first on the costs large enough to affect the plan. A short explanation of the main drivers is a better basis for a decision than a detailed inventory with no connection to customer behaviour.
2. What are we paying for whether customers use it or not?
This question identifies the spending that continues during a quiet period. It can include spare capacity, development environments and commitments you have already made. Ask what each substantial cost supports, who owns it and when it should be reviewed.
Low utilisation alone does not make a resource wasteful. Some capacity supports resilience or expected peaks. The useful distinction is between capacity with a clear purpose and spending whose purpose nobody can explain. Check that purpose before treating a reduction as a saving you can safely put in the budget.
3. If customer numbers double, how would usage and our bill change?
Ask the team to separate costs that grow with activity from capacity you already pay for. Then identify the first change the extra usage would require, including what it would cost to make and operate.
The answer can be a range. What matters is knowing what would move you towards either end, which assumptions have been checked and what still needs evidence. Give any outstanding check an owner so the forecast can improve before you depend on it.
Use the answer in your next growth decision
Bring the expected customer numbers, usage assumptions, monthly cloud cost and next capacity change into the same growth plan. Review them when customer behaviour or the product changes enough to affect the estimate. That gives the technical and commercial teams a shared set of assumptions to discuss.
Whether the current setup has room to grow or needs more capacity soon, you can use the estimate to decide when to invest and whether your pricing supports that spending.
If you want help working through this for your AWS setup, talk to us at StationOps about your cost drivers and what your next stage of growth could require.
