A reliability and cost audit answers two questions that usually get answered separately, by different people, with numbers that never meet: what is this platform costing us, and what is it risking? It is a point-in-time review, done by a person, that ends with a written report and a plan.
What gets looked at
The audit follows four surfaces, and every finding belongs to one of them.
- Cloud spend. Rightsizing, commitment coverage, storage tiers, egress, idle non-production environments, and the resources nobody remembers creating. Waste tends to concentrate in a handful of items, so the fix is usually smaller than the bill suggests.
- Reliability. Whether the SLOs are written down or assumed. Alert volume measured against alert value. The on-call rotation, its escalation path, and what actually pages at 3 a.m. Single points of failure, and whether the backups have ever been restored.
- Delivery. How much of the platform is genuinely in Terraform, and how far the running system has drifted from it. How long a deploy takes, what a rollback looks like, and whether a new engineer can ship on their first day.
- Audit trail. The controls a SOC 2, PCI DSS, or HIPAA auditor actually tests — access, change management, logging, encryption — and whether you could produce the evidence next month.
How the week runs
A scoping call sets the boundary and signs the NDA. Then read-only access: a cloud role that can list and read but not change, Kubernetes read, repository read, and a billing export. Four working days of analysis follow, and the fifth produces the report and a walkthrough.
Read-only is a design choice rather than a limitation. An audit that can change the environment is an audit whose findings are hard to separate from its own side effects, and nobody wants a consultant holding a credential that can delete a production cluster on a Tuesday.
What you get at the end
- A ranked report. Every finding says what it is, what it costs in dollars, what it risks, and roughly how much effort it takes to fix.
- A total. The report adds up the recoverable spend, so the case for acting is a number rather than an argument.
- A 90-day plan. What to do in the first two weeks, what to do next, and what to deliberately leave alone.
- The evidence. Queries, configuration, and screenshots, in a form that survives being handed to an auditor or a board.
What an audit is not
- It is not a penetration test. Nothing is exploited. If you need an attacker's view, that is a different engagement.
- It is not an implementation. The audit finds and prices the work; it does not do it. That separation is deliberate, because it keeps the audit honest about scope.
- It is not a compliance certificate. It tells you where you stand against the controls an auditor will test. Your auditor certifies; we get you ready.
The honest version of the sales pitch is this: the audit is worth buying if the report finds more than it costs. On most platforms it does, and when it does not, the report says so and you have a document that proves the platform is in better shape than you thought.