Your infrastructure bill rose 40% this quarter. What do you need to know before deciding whether that is a problem?

What is being tested

Whether you normalise cost against value before reacting. The absolute bill is nearly always the wrong first question.

The reasoning

A bill growing 40% while orders grow 60% means unit cost fell — the architecture became more efficient and the business grew. Reporting that as a cost problem and demanding reductions is exactly backwards, and it teaches teams that growth is punished.

The inverse is equally important: a flat bill while volume halves is a serious efficiency problem that a total figure hides completely.

Unit cost — cost per order, per active user, per transaction — is the metric that distinguishes these, and it is the only form in which cost is comparable to revenue. "£0.004 per order against a £28 average order value" is a sentence a business can act on; "£140,000 a month" is not.

The other questions, which come second

The other three options are all useful follow-ups once the first is answered:

• Which service? Necessary for action, and it requires cost allocation to be possible at all.
• Gradual or sudden? A step change points at a specific event — a deployment, a configuration change, a runaway job. Gradual points at growth or drift.
• Against budget? Relevant to finance and irrelevant to whether the architecture is efficient.

But without the volume comparison, all three lead to the wrong conclusion about whether there is a problem at all.

What to build so this is answerable

• A unit chosen for business value, not convenience. Requests are convenient and often wrong — a health check is a request.
• All costs allocated, including shared ones: observability, networking, platform teams, licences. Unallocated shared cost is where inefficiency hides.
• Segmentation, because cost per request for search and for a static asset differ by orders of magnitude, and an aggregate hides which endpoint is expensive.
• Tracked over time, with the expectation that unit cost falls as volume grows. If it is flat, nothing is amortising and something is linear where it should not be.

What a strong answer adds

That unit cost sets a floor on viability: if it approaches unit revenue, the product is structurally unprofitable and no operational tuning will fix it. That is a finding worth surfacing early rather than discovering at scale — and it is invisible in a total.