.png)
AWS
Microsoft Azure
Google Cloud
Key takeaways
1
High utilization means your purchased reservation benefits are being used; it does not mean most eligible usage is covered.
2
Separate genuine PAYG usage from Savings Plan-covered usage and charges outside reservation coverage before buying more.
3
Optimize retained savings against future demand by using coverage and utilization together, not by chasing a universal percentage target.
Your Azure reservations show 98% utilization, but Finance is still asking why the PAYG bill is so high. With almost all purchased capacity in use, the commitment portfolio can appear optimized.
Before approving another purchase, you need to understand what is driving that remaining spend. The answer may point to more coverage, better use of existing benefits or demand that should stay on PAYG. That choice depends on the full portfolio cost and the workloads Engineering expects to run not the utilization percentage alone.
When utilization is high and coverage is low, investigate the remaining workload before purchasing more. Confirm that it is genuinely uncovered, likely to persist after planned engineering changes, and cheaper to cover after accounting for full commitment costs and interactions with existing Savings Plans.
Before approving another purchase, you need to understand what is driving that remaining spend. The answer may point to more coverage, better use of existing benefits or demand that should stay on PAYG. That choice depends on the full portfolio cost and the workloads Engineering expects to run not the utilization percentage alone.
Short Answer
Optimize savings while tracking both coverage and utilization. Utilization shows how much purchased reservation capacity found matching usage; coverage shows how much eligible consumption received the benefit. Even 100% utilization can leave substantial PAYG usage.When utilization is high and coverage is low, investigate the remaining workload before purchasing more. Confirm that it is genuinely uncovered, likely to persist after planned engineering changes, and cheaper to cover after accounting for full commitment costs and interactions with existing Savings Plans.
Coverage vs Utilization: What Each Metric Measures
The key difference between Azure reservation coverage and utilization is the denominator. For a compatible VM segment measured over the same period:| Metric | Formula | What it tells you |
|---|---|---|
| Reservation utilization | Applied reservation unit-hours ÷ purchased reservation unit-hours × 100 | How much of the capacity you paid for was consumed |
| Reservation coverage | Reservation-covered eligible unit-hours ÷ total eligible consumed unit-hours × 100 | How much eligible consumption received a reservation benefit |
For this VM calculation, eligible consumption includes covered and uncovered infrastructure usage that matches the defined SKU or applicable instance-size flexibility group, region, and benefit scope. Exclude software, storage and networking charges from this denominator.
Keep the period and workload boundary visible beside both percentages. Reports with different scopes do not assess the same portfolio.
When comparing different VM sizes, use Microsoft’s instance size flexibility ratios within the relevant flexibility group. Raw VM counts do not establish equivalent capacity.
For a spend-based portfolio view, value covered and uncovered eligible consumption at the same no-commitment rates. Dividing discounted covered costs by a PAYG-valued denominator mixes pricing bases and distorts the result.
Keep unlike service units separate: VM-hours and database vCore-hours cannot simply be added. Aggregate compatible numerators and denominators instead of taking an unweighted average of reservation percentages.
Label reservation-only coverage separately from combined reservation and Savings Plan coverage. Usage without a reservation benefit is not necessarily PAYG.
For the matching and scope mechanics behind these calculations, our Azure Reservations guide provides a deeper explanation.
Assume illustrative rates of $1 per PAYG unit-hour and $0.60 per purchased reservation unit-hour, amortized over the period. These are hypothetical prices, not Azure quotes. The example excludes software, storage, networking, taxes, fees, cashback, and other discounts.
Keep the period and workload boundary visible beside both percentages. Reports with different scopes do not assess the same portfolio.
When comparing different VM sizes, use Microsoft’s instance size flexibility ratios within the relevant flexibility group. Raw VM counts do not establish equivalent capacity.
For a spend-based portfolio view, value covered and uncovered eligible consumption at the same no-commitment rates. Dividing discounted covered costs by a PAYG-valued denominator mixes pricing bases and distorts the result.
Keep unlike service units separate: VM-hours and database vCore-hours cannot simply be added. Aggregate compatible numerators and denominators instead of taking an unweighted average of reservation percentages.
Label reservation-only coverage separately from combined reservation and Savings Plan coverage. Usage without a reservation benefit is not necessarily PAYG.
For the matching and scope mechanics behind these calculations, our Azure Reservations guide provides a deeper explanation.
How 98% Utilization Can Still Leave High PAYG
Consider one compatible VM segment over one period, with no Savings Plans. It consumes 10,000 normalized unit-hours. You purchase 4,000 reservation unit-hours, and matching workloads consume 3,920.Assume illustrative rates of $1 per PAYG unit-hour and $0.60 per purchased reservation unit-hour, amortized over the period. These are hypothetical prices, not Azure quotes. The example excludes software, storage, networking, taxes, fees, cashback, and other discounts.
| Calculation | Result |
|---|---|
| Utilization: 3,920 ÷ 4,000 | 98% |
| Coverage: 3,920 ÷ 10,000 | 39.2% |
| Uncovered usage: 10,000 − 3,920 | 6,080 unit-hours |
| PAYG cost: 6,080 × $1 | $6,080 |
| Full reservation cost: 4,000 × $0.60 | $2,400 |
| Total cost: $6,080 + $2,400 | $8,480 |
| Same-usage baseline without reservations: 10,000 × $1 | $10,000 |
| Savings: $10,000 − $8,480 | $1,520, or 15.2% |
The reservation rate is 40% below PAYG, but the discount applies to less than half the consumption. It cannot produce a 40% saving across this segment.
The unused 80 reservation unit-hours cost $48. That amount is already included in the $2,400 reservation cost, so subtracting it again would double-count waste. Equivalently, $1,568 of discount on covered usage minus $48 of unused cost produces $1,520 in savings.
Even if all 4,000 purchased unit-hours were consumed, coverage would reach only 40%. Eliminating the final 2% of unused capacity would not close the larger gap.
Microsoft’s reservation recommendation examples demonstrate the same trade-off: reducing purchase quantity can improve utilization while reducing overall savings because more usage remains PAYG. A savings-maximizing recommendation does not need to deliver 100% utilization.
For your own portfolio, replace both illustrative rates with your enterprise prices. The useful result is the difference between the same-usage baseline and total portfolio cost, not the advertised discount.
The unused 80 reservation unit-hours cost $48. That amount is already included in the $2,400 reservation cost, so subtracting it again would double-count waste. Equivalently, $1,568 of discount on covered usage minus $48 of unused cost produces $1,520 in savings.
Even if all 4,000 purchased unit-hours were consumed, coverage would reach only 40%. Eliminating the final 2% of unused capacity would not close the larger gap.
Microsoft’s reservation recommendation examples demonstrate the same trade-off: reducing purchase quantity can improve utilization while reducing overall savings because more usage remains PAYG. A savings-maximizing recommendation does not need to deliver 100% utilization.
For your own portfolio, replace both illustrative rates with your enterprise prices. The useful result is the difference between the same-usage baseline and total portfolio cost, not the advertised discount.
What Is Driving Your Remaining PAYG?
Break down the bill before treating uncovered spend as a purchase opportunity. Use evidence from the affected workload rather than applying the portfolio headline to every region or service.| Finding | Evidence to inspect | Decision implication |
|---|---|---|
| Durable demand exceeds existing coverage | Hourly consumption, growth, new regions and expirations | Evaluate additional coverage for the persistent portion |
| Demand is intermittent or temporary | Operating schedules, peaks and project end dates | PAYG may remain cheaper than a long-term commitment |
| Existing benefits are stranded elsewhere | Reservation scope, matching configuration and unused cost | Investigate existing benefit use before adding capacity |
| Charges fall outside the VM reservation benefit | Software, storage, networking and service meters | Review these charges separately from VM commitments |
| The apparent gap is a reporting or pricing issue | Savings Plan allocation, dates, cost views and enterprise rates | Correct the comparison before changing purchases |
Scope and configuration failures matter, but a comparable 98% utilization rate leaves little unused capacity in the measured purchase. It does not show that matching failures explain the entire PAYG bill.
Similarly, Azure VM reservations cover infrastructure charges, while storage, networking, and software can remain separately charged. Another VM reservation will not discount those charges.
Reservation utilization also differs from CPU utilization. A stopped but allocated VM can remain billable and consume reservation benefits. Financial coverage can look healthy while unnecessary resources continue running or remain allocated.
An aggregate percentage can hide an expensive underused reservation among many smaller, well-used purchases. Check unused dollars alongside percentages: the financial exposure may be concentrated in a few commitments.
Similarly, Azure VM reservations cover infrastructure charges, while storage, networking, and software can remain separately charged. Another VM reservation will not discount those charges.
Reservation utilization also differs from CPU utilization. A stopped but allocated VM can remain billable and consume reservation benefits. Financial coverage can look healthy while unnecessary resources continue running or remain allocated.
An aggregate percentage can hide an expensive underused reservation among many smaller, well-used purchases. Check unused dollars alongside percentages: the financial exposure may be concentrated in a few commitments.
Audit the Portfolio Before Changing Commitments
Use a repeatable review that Finance and Engineering can both validate:
1
Align the measurement. Use the same billing scope, dates, currency, and eligibility rules. Establish the rates your organization would actually pay without commitments, including negotiated discounts.
2
Reconcile the full portfolio. Inventory reservations and Savings Plans, including quantity, configuration, scope, term, expiry, and renewal settings. Match consumed usage to its benefit and include paid unused capacity.
3
Inspect hourly demand. Compare applied benefits with uncovered consumption across representative hours. Unused reservation hours cannot roll forward, so monthly totals can conceal simultaneous waste and PAYG.
4
Forecast the optimized workload. Ask Engineering what will be resized, retired, migrated, or rescheduled. Size the decision around future required usage instead of preserving unnecessary resources to keep utilization high.
Use Actual Cost for invoice reconciliation and, where supported, Amortized Cost for costs spread across the benefit period. Cost Analysis does not support viewing amortized reservation costs for pay-as-you-go MS-AZR-0003P subscriptions. Do not add purchase charges to the same amortized costs. Unused costs are not attributed to a specific resource or subscription, so a subscription-only view can miss them.
Azure’s reservation recommendations already account for existing commitments and private on-demand discounts. Start with native reporting and recommendations, then validate them against planned changes and confirm that recent purchases are reflected.
For a consistent financial baseline across reviews, use our commitment benchmarking guide.
Record an owner, proposed action, and expected cost difference for each finding. Have Engineering validate the demand forecast and Finance agree on the baseline before approving purchases.
Azure’s reservation recommendations already account for existing commitments and private on-demand discounts. Start with native reporting and recommendations, then validate them against planned changes and confirm that recent purchases are reflected.
For a consistent financial baseline across reviews, use our commitment benchmarking guide.
Record an owner, proposed action, and expected cost difference for each finding. Have Engineering validate the demand forecast and Finance agree on the baseline before approving purchases.
Which Metric Should You Improve First?
Use these conditions as investigation signals, not universal “healthy” thresholds. Here, utilization and coverage refer to reservations for the same defined workload. Check Savings Plan-covered usage before treating remaining consumption as PAYG.| Portfolio condition | First priority |
|---|---|
| High utilization, low coverage | Test durable uncovered demand for additional commitments |
| Low utilization, high coverage | Investigate stranded benefits, timing and matching |
| Both low | Address unused commitments and uncovered demand by segment |
| Both high, weak savings | Reconcile actual rates, full costs and the savings baseline |
For durable demand, consider VM reservations when the instance family and region are stable, or consider a compute savings plan when eligible usage shifts across families, services, or regions.
The financial test is projected total portfolio cost after the change versus the same forecast workload on existing commitments plus PAYG. Include unused costs and the effect on other commitments.
Azure applies reservation benefits before Savings Plans. A new reservation can displace usage that was consuming a Savings Plan; confirm that the plan can still find sufficient other matching usage.
Where existing benefits can match more demand, investigate permitted scope changes within the same billing context first. Where demand is uncertain, keeping it on PAYG may be the better decision. Our Reservations vs Savings Plans comparison helps evaluate the right instrument.
Exchange policy: Under Microsoft’s exchange-policy change, reservations purchased after February 1, 2027 are not eligible for exchange when the corresponding service is supported by Savings Plans. Affected reservations purchased before that date retain one final exchange after the policy change. Separately, reservation refunds have product exclusions and a $50,000 canceled-commitment limit per rolling 12 months in the relevant billing scope. Verify those terms before treating recovery as an exit.
This approach follows the FinOps rate-optimization framework: align commitments with workload variability, engineering requirements, and the organization’s financial posture.
We analyze usage at the billing layer, including existing-commitment metadata. Our read-only Savings Test lets you evaluate potential savings without granting purchase authority.
We track existing Azure Reservations and Savings Plans, recommend supported commitment purchases, and provide reporting on savings and accrued cashback. For supported Flex Insured Commitments, we call provider APIs to execute approved purchases.
With our Flex Insured Commitments, teams can get up to 65% savings with an eligible three-year Azure compute Savings Plan and none of the commitment risk.
If an eligible Flex Commitment costs more than equivalent pay-as-you-go usage, we provide cashback protection for the qualifying difference under the program terms. Our protection does not cancel the underlying Azure commitment, and we do not automatically protect existing customer-owned commitments as Flex Commitments.
Customers pay us an agreed percentage of realized savings, billed monthly in arrears after Azure billing data is finalized.
The financial test is projected total portfolio cost after the change versus the same forecast workload on existing commitments plus PAYG. Include unused costs and the effect on other commitments.
Azure applies reservation benefits before Savings Plans. A new reservation can displace usage that was consuming a Savings Plan; confirm that the plan can still find sufficient other matching usage.
Where existing benefits can match more demand, investigate permitted scope changes within the same billing context first. Where demand is uncertain, keeping it on PAYG may be the better decision. Our Reservations vs Savings Plans comparison helps evaluate the right instrument.
Exchange policy: Under Microsoft’s exchange-policy change, reservations purchased after February 1, 2027 are not eligible for exchange when the corresponding service is supported by Savings Plans. Affected reservations purchased before that date retain one final exchange after the policy change. Separately, reservation refunds have product exclusions and a $50,000 canceled-commitment limit per rolling 12 months in the relevant billing scope. Verify those terms before treating recovery as an exit.
This approach follows the FinOps rate-optimization framework: align commitments with workload variability, engineering requirements, and the organization’s financial posture.
How Usage.ai Fits an Azure Commitment Portfolio Review
When recommendations, existing purchases, and engineering changes require continual coordination, we help assess the savings opportunity through our Azure commitment optimization approach.We analyze usage at the billing layer, including existing-commitment metadata. Our read-only Savings Test lets you evaluate potential savings without granting purchase authority.
We track existing Azure Reservations and Savings Plans, recommend supported commitment purchases, and provide reporting on savings and accrued cashback. For supported Flex Insured Commitments, we call provider APIs to execute approved purchases.
With our Flex Insured Commitments, teams can get up to 65% savings with an eligible three-year Azure compute Savings Plan and none of the commitment risk.
If an eligible Flex Commitment costs more than equivalent pay-as-you-go usage, we provide cashback protection for the qualifying difference under the program terms. Our protection does not cancel the underlying Azure commitment, and we do not automatically protect existing customer-owned commitments as Flex Commitments.
Customers pay us an agreed percentage of realized savings, billed monthly in arrears after Azure billing data is finalized.
Final Verdict: Optimize Savings and Track Both Metrics
A healthy utilization report is a starting point. Identify genuine uncovered demand, reconcile the full cost, and test the next action against future matching usage. Improve the constraint limiting savings instead of purchasing to reach a percentage target.Frequently asked questions
Where can we check Azure reservation coverage and utilization?
Use Azure’s Reservations view for utilization history, with the required access. Enterprise Agreement customers can use the Cost Management Power BI app’s VM RI Coverage reports. For other reporting arrangements, use available usage and reservation-benefit data to calculate coverage for the same eligible workload and period; do not assume a prebuilt coverage report is available for every billing arrangement.
How often should we review the commitment portfolio?
Use a monthly financial reconciliation, with more frequent monitoring where demand changes quickly. Reassess before renewals and after material growth, migrations, rightsizing, or purchase activity. Assign an owner to act on findings so the review does not end with a dashboard update. This is an operating recommendation, not an Azure-required cadence.
How should we use Azure recommendations after a recent commitment purchase?
Confirm that the recommendation reflects the purchase before adding another commitment. Microsoft’s Savings Plan recommendation guidance warns against buying both instruments simultaneously; update timing varies by instrument and scope. Reconcile the new allocation, then reassess the remaining opportunity.
Share
Facebook
X
LinkedIn
Reddit
