I had a conversation last month with a CIO at a 600-person Melbourne business. They’d hired a FinOps consultant who’d recommended setting up a Cloud Center of Excellence, hiring two dedicated FinOps practitioners, implementing a chargeback model across twelve cost centres, and building custom dashboards in their data warehouse. The total program cost: about $450,000 in year one. Their annual cloud spend: $3.8 million.

I told them to fire the consultant. Not because the recommendations were wrong in principle — they’re straight out of the FinOps Foundation playbook — but because they’re built for companies an order of magnitude bigger.

Mid-market cloud cost management is its own thing, and the standard frameworks don’t fit. Here’s what I actually do with clients in this segment.

The honest size threshold

Most published FinOps guidance assumes you’ve got enough cloud spend to justify a dedicated team. That math works at roughly $20 million annual spend and up. Below that, the overhead of running a “real” FinOps program eats most of the savings.

For mid-market — call it $1M to $10M annual cloud spend — you need a leaner approach. The goal isn’t to build a function. The goal is to build the smallest set of habits that keeps your spend under control without creating a bureaucracy that costs more than it saves.

The four habits that matter

I’ve narrowed this down through repetition. There are four things mid-market companies need to do consistently, and almost everything else is overhead.

One: tag everything, ruthlessly. Every resource needs an owner, an environment, and a cost centre tag. No exceptions. Untagged resources should fail to deploy. This is a one-time engineering investment that pays off forever. Without it, every other piece of cost management is theatre.

Two: weekly cost review with the people who can act. Thirty minutes, every Friday, with whoever owns the largest cost centres. Not a steering committee. Not a CCoE. Just engineers and a finance partner looking at what changed week-on-week and whether anything looks weird. Most cost surprises are caught here within seven days instead of at the end of the month.

Three: kill non-prod aggressively. Mid-market companies typically have non-production environments running 24/7 because nobody wrote the automation to shut them down. This is usually 25-40% of the bill. A weekend shutdown schedule saves more than any reserved instance optimization. The engineering effort is two days.

Four: commit modestly. Reserved instances and savings plans are great, but mid-market companies routinely over-commit. The right starting point is committing to roughly 60-70% of your stable baseline workload, not 100%. The flexibility premium is worth it because your demand patterns aren’t predictable enough to justify deeper commits.

Where I see mid-market companies waste money

The patterns repeat. Almost every audit I run finds the same five things.

Idle databases that someone spun up for a project that ended six months ago. Old snapshots accumulating at $200/month each, multiplied by hundreds of them. Oversized instances that were sized for peak load that never materialized. Data egress charges from poorly architected backup processes pulling production data through the internet instead of through private endpoints. And, increasingly in 2026, AI service costs from teams experimenting with Azure OpenAI or Bedrock without any guardrails on token usage.

That last one is the new frontier. I had one client in March where their monthly Azure OpenAI bill went from $4,000 to $38,000 in six weeks because a developer had built an agent loop without circuit breakers. Nobody noticed until the invoice arrived. This is now a category I check first on every audit.

The chargeback question

CFOs love chargeback. They imagine a world where engineering teams feel the pain of their cloud usage and self-regulate.

In practice, chargeback at mid-market scale creates more problems than it solves. The accounting overhead is significant. The arguments about allocation methodology consume actual hours. And the behaviour change is modest, because engineering teams in mid-market companies usually don’t have the autonomy to make architectural changes that would meaningfully shift their costs.

What works better is showback: every team sees what they’re spending, but it doesn’t actually move money between budgets. You get 80% of the behavioural benefit at 10% of the implementation cost.

What to actually do this quarter

If you’re a mid-market CIO reading this and your cloud bill makes you nervous, here’s the order I’d run things in.

Tag your resources properly. Spend two weeks getting this right. Then run a weekend shutdown schedule for non-prod and watch your bill drop 20-30%. Then implement weekly cost reviews with engineers, not executives. Then look at your commitment posture and trim it back to 60-70% of stable baseline. Then, only if you still have a problem, talk about whether you need any kind of formal program.

If you’ve worked through that list and you still need help, that’s the point at which bringing in outside expertise — whether that’s the kind of strategic advisory work I do, a specialist FinOps practice, or a hybrid — actually starts to pay off. Not before.

The mistake most mid-market companies make is buying the enterprise-grade solution before they’ve done the basics. The basics get you most of the way there. The fancy stuff is for when the basics aren’t enough.