How big the waste actually is
Industry survey data is consistent on the shape of the problem. The Flexera State of the Cloud Report puts self-reported waste in the high-20s to low-30s percent range year after year[1]. The FinOps Foundation’s annual practitioner survey has ranked reducing waste and managing commitments as the top priority for FinOps teams since 2024[2]. Flexera’s 2025 survey finds 84% of organisations struggle to manage cloud spend[3].
For a mid-market Australian organisation spending $1M to $5M annually on cloud, that puts $200K to $1.5M of waste on the table. The savings are not strategic, they are operational: things that should not have been running, sized, or kept the way they are.
Where the waste reliably sits
- Over-provisioned compute: instances sized for peak loads that never materialised, or that materialised once and were never re-evaluated
- Idle and orphaned storage: snapshots and unattached volumes from terminated workloads, plus storage tiers that should have moved to infrequent-access months ago
- Unused reserved instances or savings plans: commitments made for projected demand that did not arrive
- Forgotten non-production environments: dev, staging, and test resources running 24/7 instead of being shut down outside business hours
- Data transfer charges: cross-region or cross-AZ traffic patterns that could be co-located, with the saving never quantified
- Licence-included instances paying for software the organisation already owns under enterprise agreement
Why the incumbent MSP is rarely the right reviewer
The MSP that provisioned the cloud is being asked to assess whether the provisioning was right. This is the same conflict that exists in MSP oversight generally, but in cloud cost it shows up in specific ways: instances were sized for SLA safety margins that suited the MSP’s operational risk appetite, not the client’s commercial appetite; reserved instance purchases were timed to the MSP’s billing cycle rather than the client’s demand curve; cost-optimisation effort competes with revenue-generating project work for the same engineers.
None of this requires the MSP to be acting in bad faith. Structurally, the incentive to surface waste is on the client side, not the supplier side.
A practical first review
Week 1: extract
- Twelve months of cost and usage data exported from the cloud provider’s billing console (Cost Explorer, Azure Cost Management, GCP Billing)
- Resource inventory across compute, storage, network, and managed services
- Reserved instance / savings plan portfolio with utilisation rates
- Tagging coverage by cost centre or product line
Week 2: analyse
- Right-sizing recommendations from the provider’s own recommendations engine, validated against actual utilisation
- Idle resource sweep (compute, storage, IPs, gateways)
- Commitment portfolio review: coverage, utilisation, expiry alignment
- Non-production environment scheduling analysis
Week 3: prioritise
- Stack-rank findings by annual saving versus implementation effort
- Identify the changes that need no operational risk acceptance (delete the unattached volume) versus those that do (downsize the production cluster)
- Build a 90-day execution plan that captures 70% of the saving with the lowest 30% of effort
Beyond the first review
A first review surfaces waste; an ongoing discipline prevents it accumulating again. The FinOps Foundation framework distinguishes Inform (visibility), Optimise (action), and Operate (continuous improvement) phases[2]. Most mid-market organisations live permanently in Inform, occasionally enter Optimise after an alarming bill, and rarely reach Operate. The board’s job is to keep the discipline at Operate, regardless of who is doing the implementation.