skie.io
Sage for Azure

Right-sizing and reservations, decided together

Sage is skie's cost optimization engine for Azure. Every run returns two things at once: how your resources should be sized, and which reservations to hold. They are produced together because deciding either one on its own gets the other wrong.

Sage customers typically cut cloud spend by 25-50%. Across the whole customer base running Sage through 2025, realised spend fell 41% on average against the pre-optimization run rate.

Get a free assessment
  • Trusted & qualified
  • AWS Advanced Tier Partner
  • AWS Well-Architected
  • SAP Competency
  • SOC 2 Type II

One run, two kinds of decision

Optimization tools usually separate right-sizing from commitment buying, which is how you end up being told to resize a virtual machine that a reservation is already paying for. A Sage run on Azure returns both sets of recommendations side by side, and each proposed resize carries the reservation coverage of the machine it applies to.

  1. Output 1

    Right-sizing

    Per machine: current size and proposed size, subscription, location, and an efficiency score before and after.

  2. Output 2

    Reservation decisions

    Which commitments to add, shown as new purchases by region and series, priced against the post-resize footprint.

  3. Shared

    Coverage on every row

    Each resize proposal shows what the machine is currently covered by, so a change is never recommended blind to the commitment paying for it.

  4. Shared

    What to leave alone

    Resources the optimizer decided not to touch are labelled explicitly rather than dropped from the list, so you can see the whole estate was considered.

Virtual machine right-sizing recommendations listing each machine with its location, the size before and after the proposed change, an efficiency score before and after, and the reservation coverage on that machine. Rows are labelled either actionable or no change.
Coverage sits next to every proposal. The right-hand column shows what each machine is already covered by, so a resize is never suggested blind to the reservation paying for it. Machines the optimizer decided to leave alone stay in the list, labelled as no change. The subscription column has been removed from this image because those names identify the customer; nothing else has been altered.

What a run gives you

Every optimization reports the cost position before and after, split by how you are paying rather than only by how much.

ReportedDetail
Projected savingsMonthly and annualised, with the percentage change against current spend
Cost split by commitment typeCurrent and optimized monthly cost broken into reserved and pay-as-you-go, so you can see spend moving from one to the other rather than only the net figure
Actionable countHow many of the analyzed resources have a recommendation worth acting on, against the total examined
Reservation flowWhere new commitment would go, by region and machine series
Run historyEvery run stored and dated, with analyses older than thirty days flagged as stale, plus CSV export of any recommendation set
Cost analysis comparing current and optimized monthly cost, with a table splitting the total into reserved instances and pay-as-you-go, showing pay-as-you-go falling to zero as reserved spend rises.
The split matters as much as the total. Pay-as-you-go spend moves onto reservations rather than simply disappearing, so you can see where the saving comes from instead of taking the net figure on trust.
Reservation management flow diagram showing a single new commitment splitting into seven separate reservation purchases, each labelled by region and machine series, above a summary of the decision, the number of reservations and the total commitment affected.
Where the new commitment actually goes. One decision resolves into seven specific purchases, each tied to a region and a machine series, with the total commitment it puts on the books stated plainly rather than buried.
Two panels: cost change split into right-sizing savings and a smaller increase in commitment spend, and the net change in virtual machine counts by series, with two series growing and two shrinking.
Right-sizing falls, commitment rises, and some series grow. The plan spends more on reservations because right-sizing changed what needs covering, and it grows two machine series while shrinking two others. The total drops even though several lines move up.

Which Azure services does Sage cover?

Optimization runs against virtual machines, SQL databases and managed disks. Visibility and spend attribution extend across the wider estate.

ServiceDepthWhat Sage does
Virtual machinesOptimizedRight-sizing with efficiency scoring, plus reservation decisions in the same run
SQL databasesOptimizedRight-sizing and reservation planning, including managed instances
Managed disksOptimizedDisk sizing and tier optimization, with snapshot inventory
VM scale setsVisibilityInventory of scale sets and their instances, with spend attribution
Reservations and savings plansVisibilityFull inventory of what you hold, with reservation spend reporting
Storage accounts, load balancersVisibilityInventory and spend attribution

Understanding where Azure spend goes

Azure billing is reported by meter rather than by service, which is a large part of why Azure costs are harder to attribute than AWS costs. Sage reports spend the way Azure actually bills it, then adds the views you need on top.

By meterSpend by meter category and sub category, over eighteen months
By locationRegional spend distribution across subscriptions
By groupResource group attribution for chargeback and showback
By commitmentA separate reservation spend view, split from on-demand
Azure spend dashboard showing total, on-demand and reserved spend for the month with trend indicators, the previous month broken down by meter category as a ring chart, and an eighteen month history of spend by meter category.
Reported the way Azure bills. Total spend split between on-demand and commitment, then the month broken down by meter category rather than forced into a service taxonomy that Azure does not actually use.

Common questions

Will Sage tell me to resize a machine that a reservation already covers?

No. Every right-sizing row carries the reservation coverage of the machine it applies to, and reservation decisions are produced in the same run as the sizing recommendations rather than separately.

How does Sage decide what to leave alone?

Resources the optimizer examined but chose not to change are labelled as no change, and stay in the list. You can see the whole estate was considered rather than only the parts with a recommendation attached.

What is an efficiency score?

A measure of how well a machine's size matches its actual workload, reported before and after the proposed change so the improvement is visible per resource.

Can I export the recommendations?

Yes. Any recommendation set can be exported to CSV, and every run is stored with its date so plans can be compared over time.

How does this differ from Sage for AWS?

The AWS product solves commitments across EC2, ECS and EKS in one portfolio, because Savings Plans are shared between them, and it can apply approved changes for you during a patch window. On Azure, right-sizing and reservation decisions are produced together per service. See Sage for AWS for that side of the platform.

Cloud cost optimization skie has delivered with Sage

See what Sage finds in your subscriptions

A free assessment runs against your real usage and returns the same right-sizing recommendations and reservation decisions described on this page.

Get a free assessment