SpecialisedUser storyConsultative playbook

We're paying retail prices for workloads that have run for two years

The platform team's stable workloads have been running for years on pay-as-you-go pricing. Reservations and savings plans have come up in conversation but never happened because no one owns the decision. A FinOps practitioner has just been hired; the first deliverable is a defensible reservation strategy.

Trigger: Renewal opportunity; new FinOps practitioner in post.
Good outcome: The customer is paying 15–30 % less in the first year on the same workloads — no architecture change, no risk. The unused resources are gone, and there's a quarterly check to keep it that way.

Discovery — signals and questions

Signals validating this story

·Stable workloads on pay-as-you-go for 12+ months
·Renewal moment or budget cycle approaching
·FinOps practitioner hired or assigned
·No documented reservation strategy
·Right-sizing and idle cleanup pending

Discovery questions

1.Which workloads have been running stably for 12+ months?
WhyIdentifies reservation candidates.
2.What's your current commitment level — none, 1-year, 3-year?
WhyEstablishes the gap between pay-as-you-go and optimised state.
3.Who has authority to commit to reservations?
WhyProcurement / finance ownership question — often the bottleneck.
4.What's the over-provisioning baseline — VMs running at <30% CPU?
WhyRight-sizing opportunity. Often the biggest single lever.
5.How do you track idle and orphaned resources today?
WhyCleanup is the cheapest savings. Surfaces whether it is operationalised.

Baseline architecture → Target architecture

Baseline architecture

Pay-as-you-go dominant. No reservation discipline. Right-sizing ad-hoc. Idle resources accumulating without a cleanup cadence. Defender for Cloud may surface them but no remediation backlog.

Typical concerns

·Paying retail for stable workloads
·Over-provisioning on VMs and managed services
·Idle resources surviving for months
·No documented reservation decision authority
·Renewal opportunities missed

Capability gaps

·Reservation strategy with named owner
·Right-sizing as platform discipline
·Idle resource cleanup cadence
·Decision authority for commitments
·Quarterly optimisation review

Target architecture

Reservation strategy aligned to forecast: 3-year reserved instances or savings plans for the most-stable tier; 1-year for moderately predictable; pay-as-you-go for variable. Right-sizing automated where signal is clear; reviewed where ambiguous. Idle and orphaned resources flagged weekly with a cleanup backlog. FinOps cadence runs the quarterly optimisation review.

Key capabilities

Reservation portfolio aligned to forecast
Right-sizing as platform discipline
Weekly idle-resource flagging
Quarterly optimisation review cadence
Decision authority documented

Architecture decisions

1.Reservation term — 1-year vs 3-year vs mix
3-year
Fits whenWorkload demonstrably stable over multi-year horizon.
Trade-offsLarger commitment with less flexibility.
1-year
Fits whenModerate predictability; growth uncertainty.
Trade-offsLower discount tier.
Mixed by tier
Fits whenMature FinOps; differentiated workload patterns.
Trade-offsMore portfolio management overhead.
Default recommendationMixed by tier. 3-year for production database VMs; 1-year for steady-state compute; pay-as-you-go for variable.
2.Right-sizing — automated vs platform-team reviewed
Automated
Fits whenWorkloads with clear utilisation signal; tolerant of size changes.
Trade-offsRisk of unintended impact on latency-sensitive workloads.
Platform-team reviewed
Fits whenWorkloads with ambiguous signal; high blast-radius.
Trade-offsSlower cadence; relies on platform-team bandwidth.
Default recommendationAutomated for low-blast-radius; reviewed for high.
3.Reservations vs Savings Plan vs Spot
Reservations (per-VM-family)
Fits whenStable workload pinned to specific VM family.
Trade-offsFlexibility limited if family changes.
Savings Plan (compute hours)
Fits whenCompute usage stable but mixed family.
Trade-offsSlightly lower discount than RIs.
Spot
Fits whenInterruptible workloads (batch, big data, dev/test).
Trade-offsEviction risk; not for production-critical.
Default recommendationSavings Plan as the base; Reservations for very stable workloads; Spot for interruptible.

Low-risk trial — proof of value

45-day FinOps optimisation — reservations + right-sizing + idle cleanup

~6 weeks

Reservation analysis on stable workloads with named owners. Right-sizing review on top ten spend lines. Idle and orphaned resource cleanup pass. First quarterly optimisation cadence run with platform and finance.

Success criteria

Realised first-month savings of 10%+ on the trial scope
Reservation portfolio sized to forecast and committed
Idle and orphaned resources flagged and 50%+ actioned
Quarterly cadence runbook authored and trialled

InvestmentCost Management free for Azure. Reservation commitments are cost-positive from day one. Advisory engagement only.

Proof metrics

·15–30% first-year savings on the trial scope
·Reservation coverage above 60% on stable workloads
·Idle-resource count trending down quarter-over-quarter
·Right-sizing actions delivered without production impact

Recommended cards

The SKUs and capabilities most likely to be part of the solution, with the editorial rationale for each in the context of this story. Add the ones that fit your situation.

SKUMicrosoftSaaS

Microsoft Cost Management

Azure's first-party FinOps console — budgets, anomaly alerts, rightsizing recommendations, showback / chargeback, and exports to data warehouse. Free for Azure-only estates; paid connectors extend it to AWS and GCP for multi-cloud FinOps.

In: Azure consumption

Cloud Foundation

Why for this story

Reservation recommendations + utilization reporting. The source of truth for which workloads are reservation candidates.

SKUMicrosoftAzure

Azure Monitor

Azure's first-party observability platform — metrics, logs, traces, and alerts across Azure resources, hybrid workloads (via Azure Arc), and other clouds. The substrate platform teams use to know what's actually happening across the estate.