Effective savings rate: the commitment number finance should ask for
In brief
Coverage and utilization can both look healthy while commitments save little. Why finance should ask for effective savings rate, and how to check the math.
Steve Elliott · · 7 min read
On this page
- Why don't coverage and utilization tell finance what was saved?
- How does the FinOps Foundation define effective savings rate?
- A worked example: high coverage, low savings
- How do FOCUS columns feed ESR, and where can the sum go wrong?
- What should finance ask before accepting an ESR figure?
- How might the Azure exchange change affect ESR?
- What is a good effective savings rate?
- Sources
Finance should ask for effective savings rate (ESR): the share of on-demand-equivalent spend that cloud commitments actually saved, after unused commitment is counted as a cost. Coverage and utilization each describe one side of a commitment portfolio, and both can look healthy while the money saved is thin.
ESR is easy to define and easy to calculate in a way that flatters the result. This post shows why the familiar metrics mislead and what to check in a FOCUS-based ESR before it reaches a board pack.
Why don't coverage and utilization tell finance what was saved?
Commitment coverage measures how much eligible usage ran under a commitment. AWS calculates it in on-demand-equivalent terms: covered usage divided by covered usage plus eligible usage billed at on-demand rates. Commitment utilization measures the other side: how much of what you bought you used. The FinOps Foundation's Commitment Utilization Score is used commitment divided by total commitment.
Neither answers the CFO's question. Coverage ignores what you paid for commitments you didn't use. Utilization ignores discount depth and uncovered spend. Shallow one-year savings plans can run at 100% utilization and save little.
The FinOps Foundation's Rate Optimization capability makes the same point: "under utilization of a commitment discount would also negatively impact ESR," as would significant uncovered usage. ESR nets both effects into one rate.
How does the FinOps Foundation define effective savings rate?
The Rate Optimization capability lists "Effective Savings Rate Percentage" as a return-on-investment measure for commitment discounts and gives two formulas:
- Savings form. Commitment discount savings, minus the cost to achieve them, divided by on-demand-equivalent spend.
- Spend form. 1 minus actual spend with discounts, divided by the same usage priced at on-demand rates.
The Foundation's ESR playbook, last updated March 17, 2026, lists the cost to achieve as software, services, time and tooling. Its steps cover AWS compute over a full month. It also offers a shorthand: ESR equals utilization times coverage times discount, with the example 90% × 100% × 50% = 45%.
That shorthand matches the exact figure at full utilization. Below that, for discounts up to about 50%, it overstates ESR, and the gap widens as utilization falls, because unused commitment is a cost, not just a missing saving. With deeper discounts it can understate. By our arithmetic, the exact relationship under the spend form is coverage × (1 − (1 − discount) ÷ utilization). It also gives a break-even rule worth putting in front of finance: a commitment loses money when utilization drops below 1 minus its discount. At a 30% discount, that line is 70%.
A worked example: high coverage, low savings
These numbers are illustrative, not from any company. Each portfolio has $1 million a month of commitment-eligible usage at on-demand rates.
| Portfolio A | Portfolio B | Portfolio C | |
|---|---|---|---|
| Coverage | 95% | 60% | 85% |
| Utilization | 80% | 99% | 100% |
| Discount vs on-demand | 30% | 40% | 15% |
| Monthly commitment cost | $831,250 | $363,636 | $722,500 |
| Of which unused | $166,250 | $3,636 | $0 |
| Uncovered on-demand spend | $50,000 | $400,000 | $150,000 |
| Saved vs on-demand | $118,750 | $236,364 | $127,500 |
| ESR | 11.9% | 23.6% | 12.8% |
| Shorthand (U × C × D) | 22.8% | 23.8% | 12.8% |
Portfolio A looks best on coverage. It over-bought: a fifth of its commitment went unused, eating most of the discount. The shorthand reports nearly double the real rate.
Portfolio B looks worst on coverage. It committed only to its steady base at deep discounts and saves twice as much as A.
Portfolio C has perfect utilization but shallow discounts, so it saves about the same as A. Coverage ranks these portfolios backwards, utilization rates C best, and only ESR ranks them by money saved.
How do FOCUS columns feed ESR, and where can the sum go wrong?
The FOCUS specification, whose version 1.4 was ratified June 4, 2026, gives finance four cost columns to rebuild ESR across providers:
- ListCost. The pre-discount baseline.
- ContractedCost. Cost after negotiated discounts. It defaults to ListCost when there are none.
- EffectiveCost. The accrual view, with commitment purchases amortized onto the usage they cover.
- BilledCost. The cash view: what was invoiced in the billing period.
The spec's cost comparison feature splits savings cleanly. ListCost minus ContractedCost is negotiated discount savings. ContractedCost minus EffectiveCost is commitment discount savings. ListCost minus EffectiveCost is the total. FOCUS's published ESR use case takes the middle comparison: (ContractedCost − EffectiveCost) ÷ ContractedCost, summed over the period.
Use EffectiveCost, not BilledCost. BilledCost puts an upfront purchase in the month it was invoiced, so an ESR built on it swings between a terrible purchase month and months that look free; EffectiveCost spreads that cost across the usage it covers, as our post on mapping FOCUS data into the TBM taxonomy explains. How your organization accounts for prepaid commitments is a separate question; confirm that with your advisers.
Watch the unused rows. FOCUS records unused commitment as rows with a CommitmentDiscountStatus of "Unused." In the spec's words, "these unused rows still have EffectiveCost to reflect the wasted commitment value." In its examples those rows also carry a ListCost equal to their EffectiveCost. So they add to cost and baseline in equal measure: they show zero savings rather than a loss.
Take the usage rows in the spec's own 50% utilization example. By our arithmetic, its 12 used hours carry ListCost of $1,424.04, its 12 unused hours add $949.32 of ListCost, and total EffectiveCost is $1,898.64. Summed, that is a 20% savings rate. But usage that would have cost $1,424.04 at list cost $1,898.64 under the commitment. Measured against what was actually consumed, the rate is about −33%.
The FOCUS ESR use case query sums every row in the period, with no filter on charge category or commitment status. That causes two problems. First, the ContractedCost and ListCost column pages say to leave purchase charges out of accrual-basis savings totals, because the usage rows already carry that cost. The query keeps them, so any month with a commitment purchase or recurring fee inflates ESR. In an illustrative month with full coverage and utilization, a 30% discount and a monthly fee row carrying the fee as ContractedCost, the true ESR is 30% and the query returns about 59%. Second, in the spec's examples, where there are no negotiated discounts, unused rows' ContractedCost would, by our reading, equal their ListCost, so the query overstates ESR as the Foundation defines it, in the same way as the sum above.
The columns are sound; the spec's own cost comparison query handles the double count, and FOCUS's coverage rate use case notes that unused rows inflate both sides of its ratio. We think the safe ESR build filters to usage charges, keeps unused rows' EffectiveCost in the cost line and drops them from the baseline. Test that on your own data before you trust a trend.
For coverage, FOCUS 1.4 adds a Commitment Program Eligibility Details column listing the programs a charge qualifies for, applied or not. FOCUS's coverage rate use case uses it to limit the denominator to eligible charges.
What should finance ask before accepting an ESR figure?
We suggest six questions, in this order:
- Baseline: list or contracted? A ListCost baseline mixes negotiated discounts into the result. A ContractedCost baseline isolates what commitments added. In our view, both are defensible. The report should say which.
- Is unused commitment in the cost, and out of the baseline? If it sits in both, or in neither, the figure is closer to a gross discount rate than to ESR.
- Amortized or billed? We think ESR should run on amortized, accrual-based cost.
- Are purchase rows excluded? On an accrual basis, their cost is already in the usage rows.
- Net of cost to achieve? The Foundation's savings form subtracts it: software, services, time and tooling.
- What scope? Which providers, services and accounts are in the denominator, and is the window a full month?
Ask for the trend, too: the playbook notes that ESR can move daily.
How might the Azure exchange change affect ESR?
From February 1, 2027, Azure reservations bought for services that savings plans also cover can't be exchanged. Microsoft lists Virtual Machines, App Service and SQL Database among them. Earlier reservations keep one final exchange. Cancellation, with its $50,000 rolling 12-month limit, and trade-ins to savings plans are unchanged. Microsoft says the change gives customers a clear, predictable framework for choosing a commitment type for each workload.
Microsoft's own guidance says "when fully utilized, reservations provide the greatest savings." A reservation bought after February 1, 2027 for one of these services can't be exchanged if the workload later moves. The exits left are cancellation within the limit or a trade-in to a savings plan. Directions on Microsoft, an independent analyst firm, wrote that the change "closes a door that offered flexibility and potentially significant savings." That is its view, not a measured outcome.
In our view, the effect depends on behavior. Reservations bought for workloads that drift will add unused commitment and lower ESR. A shift to savings plans trades some discount for steadier utilization. Reservations kept to steady workloads may change little. Coverage won't show which is happening, so ask for ESR by commitment type before and after February 2027.
What is a good effective savings rate?
There is no standard target. The playbook cites data from ProsperOps, a commitment-discount software vendor whose staff contributed to the playbook. It covers thousands of anonymized AWS organizations: median ESR of 0%, a 75th percentile of 23% and a 98th percentile of 46%. Treat those as one vendor's AWS-only sample. Your own trend, split by provider and commitment type, is the better yardstick.
We think ESR belongs next to forecasts and funding decisions, because every commitment is a multi-year bet on demand that finance underwrites. For how to put cloud cost in front of a CFO, see how to present IT cost to the CFO.
Sources
- FinOps Foundation, FinOps Framework capability: Rate Optimization, accessed November 2026
- FinOps Foundation, How to Calculate Effective Savings Rate (ESR), last updated March 17, 2026
- FinOps Foundation, Introducing FOCUS 1.4, June 10, 2026
- FOCUS Specification 1.4, Cost Comparison
- FOCUS Specification 1.4, Effective Cost, Contracted Cost and List Cost columns
- FOCUS Specification 1.4, Appendix: Commitment Discounts examples
- FOCUS Use Cases, Determine Effective Savings Rate by Service (v1.3–1.4)
- FOCUS Use Cases, Calculate Commitment Discount Coverage Rate with Eligibility-Adjusted Denominator (v1.4)
- AWS, Understanding coverage metrics and calculations, Savings Plans User Guide
- Microsoft Learn, Self-service exchanges and refunds for Azure Reservations, updated July 29, 2026
- Microsoft Tech Community, Reservation exchanges for Azure services covered by savings plans end starting Feb. 1, 2027, July 30, 2026
- Microsoft Learn, Decide between a savings plan and a reservation, updated March 14, 2026
- Directions on Microsoft, Most Azure Reservation Exchanges End in 2027, Wes Miller, August 7, 2026
Frequently asked questions
Can effective savings rate be negative?
Yes. ESR turns negative when the cost of unused commitment exceeds the discount earned on what was used. By our arithmetic, a commitment breaks even when utilization equals 1 minus its discount, so a 30% discount needs at least 70% utilization. The FOCUS spec's 50% utilization example, measured against usage actually consumed, works out to about −33%.
How is effective savings rate different from commitment coverage?
Coverage measures how much eligible usage ran under a commitment. ESR measures how much money commitments saved against on-demand rates, net of unused commitment. A portfolio can show 95% coverage and a low ESR if it over-bought. Another can show 60% coverage and a higher ESR because it committed only to steady usage at deep discounts.
Should ESR include negotiated discounts such as an enterprise agreement?
It depends on the question. FOCUS separates the two: ListCost minus ContractedCost is negotiated discount savings, and ContractedCost minus EffectiveCost is commitment discount savings. A ContractedCost baseline shows what commitments added on top of negotiated rates. A ListCost baseline shows total rate savings. Either works if the report states which one it uses.
Which cost column should ESR use, billed or effective?
EffectiveCost. In FOCUS, BilledCost is the cash view, so an upfront purchase lands entirely in the month it was invoiced. EffectiveCost amortizes that purchase onto the usage it covers and records unused commitment as cost. ESR built on BilledCost swings between a very poor purchase month and months that look almost free.
Steve Elliott
Published November 22, 2026
More on FinOps & Cloud
FOCUS, commitments, shared cost.
Keep reading
Should product teams own their cloud budgets?
Product teams should hold cloud budgets once allocation puts spend in their name and shared costs have an owner. Prerequisites, guardrails and failure modes.
AI capacity reservations aren't savings plans: how to compare PTUs, model units and GSUs
Provisioned AI capacity is billed every hour, used or not, and often breaks even only at high use. How to compare PTUs, model units and GSUs, with the math.
Should FinOps report to the CTO, CIO or CFO?
Most FinOps teams report to the CTO or CIO; 8% report to the CFO. When each line fits, what it costs, and how to make a dotted line to finance work.
What AI agents should and shouldn't do in FinOps
Agents suit FinOps work that is read-only or easy to undo. Commitments, policy and chargeback disputes need a person. Where the line sits, with guardrails.