Case study · FinOps
Six-figure savings from a 27-subscription Azure estate
An enterprise Azure estate of 27 subscriptions had grown the way most do: quickly, and without anyone able to say what each workload cost or why the bill kept rising. I established a monthly cost-governance discipline across the estate — subscription-by-subscription baselines, owner accountability, orphaned-resource cleanup and per-namespace Kubernetes cost apportionment — and identified six-figure annual savings, with anomalies now caught in days rather than at invoice time.
- —A 27-subscription estate was baselined subscription by subscription, each with a named owner — cost became accountable, not ambient.
- —A monthly cost-review cadence catches anomalies and drift in days instead of at invoice time.
- —Shared Kubernetes platform costs were apportioned per namespace using Prometheus utilisation data, ending the single-line-item AKS bill.
- —Six-figure annual savings were identified across orphaned resources, unused subscriptions and right-sizing.
The challenge
The client's Azure estate — 27 subscriptions supporting industrial digital products, name withheld — had grown fast enough that nobody could confidently answer the basic questions: what does each subscription cost, who owns it, and why is this month higher than last? Kubernetes made it worse: the shared platform arrived as a single opaque line item no team felt responsible for.
The approach
- Baseline and ownership. Every subscription was baselined and given a named owner — the precondition for any cost conversation that changes behaviour.
- A monthly governance cadence. A structured monthly review works through the estate: variances explained, anomalies chased, actions owned and carried forward.
- The reliable seams worked hard. Orphaned resources, unused and unclaimed subscriptions, over-provisioned compute and storage tiering — the unglamorous list that always pays.
- Per-namespace Kubernetes costing. Shared AKS platform cost was apportioned to teams per namespace using utilisation shares from managed Prometheus, turning one opaque bill into accountable team-level numbers.
The results
- Six-figure annual savings identified across cleanup, cancellations and right-sizing — found by discipline, not by a tool licence.
- Anomalies caught in days. A cost spike now surfaces at the next review or sooner, not on a quarterly invoice shock.
- Kubernetes cost made accountable. Teams see their actual platform share, and platform right-sizing decisions finally have owners.
- A rhythm, not a rescue. The governance cadence continues without heroics — which is what keeps the savings from quietly returning.
Why it matters
Azure waste is not a technology problem; it is an ownership problem with a technology surface. The estates that stay cheap are the ones where every pound has a name attached and someone looks at it monthly. That discipline is buildable in weeks — and it pays for itself many times over, every year, for as long as it runs.
Client identity withheld under confidentiality. Related services: Azure & AI cost management, Azure cost optimisation and Azure managed services.
Asked and answered.
How do you find savings in a large Azure estate?+
Baseline every subscription with a named owner, then work the reliable seams: orphaned and idle resources, subscriptions nobody claims, over-provisioned compute, missing reservations and untagged spend. On estates that grew without governance, that discipline typically surfaces six-figure annual savings.
How do you allocate shared AKS costs to teams?+
By apportioning cluster cost per namespace using utilisation data from managed Prometheus — each team sees its actual share of the platform rather than one opaque line item. Visible cost changes behaviour; invisible cost never does.
What stops the waste coming back?+
Cadence and ownership: a monthly review per subscription with its owner, anomaly detection between reviews, and tagging discipline so new spend lands accountable. Cost optimisation as a one-off project decays; as a governance rhythm it compounds.