AWS Cost Explorer provides both a utilization report and a coverage report for Savings Plans and Reserved Instances. What is the difference between the two, and what does a low value on each one tell you?
answer
- one starts from the commitment, one from the usage
- unused commitment is money already gone
- low coverage is savings not taken
- read the pair, not either alone
- utilization decays as the fleet changes
basics
~20 sUtilization is the share of a purchased commitment that was actually used; coverage is the share of eligible on-demand usage that a commitment discounted. Low utilization means you are paying for unused commitment; low coverage means discountable usage is still billed at On-Demand rates.
solid answer
~50 sThey look at the same discount from opposite ends. **Utilization** starts from the commitment: of the hourly spend or reserved hours you promised to buy, how much did matching usage actually consume? Utilization below 100% means you are paying for a commitment nothing used — money already gone. **Coverage** starts from the usage: of the eligible On-Demand usage you ran, what share was discounted by a commitment? Low coverage means usage that could have been cheaper was billed at full rate — savings you did not take. The diagnostic value is in reading them together. High utilization with low coverage says the commitment is working but is too small, or scoped too narrowly, for the fleet. Low utilization with reasonable coverage says the commitment outlived or outgrew the workload it was bought for — an instance family changed, a region moved, a service was decommissioned. Both near 100% is unusual and often means the commitment was sized against a very stable fleet.
go deeper
Know that AWS Cost Explorer reports on commitments you already hold, and be able to state the one-line difference: utilization is about the commitment being used, coverage is about usage being discounted.
Explain what each report is denominated in and why a shortfall means something different in each — unused commitment is spend already lost, uncovered usage is a discount you did not take.
Read the two numbers as a pair and diagnose from the combination: undersized versus scope-mismatched versus outlived commitments, and why utilization decays quietly as the estate changes around it.
Own the review cadence and the alerting: who watches these numbers, what floor triggers action, and the honest caveat that a perfectly fitted commitment says nothing about whether the underlying workload should be that size.
## Two views of one discount A Savings Plan or a Reserved Instance is a promise: you commit to a level of spend or a quantity of capacity for a term, and AWS discounts matching usage. Two things can go wrong with that promise, and they are opposites. You can commit to more than you use, or you can use more than you committed to. Cost Explorer gives each failure mode its own report. ## Utilization — did the commitment get used? The **utilization** report is denominated in the commitment. For a Savings Plan it asks: of the committed dollars-per-hour, how many were actually consumed by matching usage each hour? For reservations it asks: of the reserved instance hours, how many were matched by running instances? Unused commitment is **pure waste**. You paid for the hour whether or not anything ran in it, and unlike On-Demand there is no way to get it back. A utilization report showing 70% means roughly 30% of what you committed was spent on nothing. Utilization drops for recognizable reasons: the workload the commitment was bought for was decommissioned or migrated; the fleet moved to a different instance family or region that the reservation's scope does not match; capacity was consolidated onto fewer, larger instances; or the purchase was simply sized against a peak that no longer occurs. Utilization is also the metric that quietly rots — it is fine on the day of purchase and degrades as the estate changes underneath it, which is why it deserves a standing review rather than a one-time check. ## Coverage — did the usage get the discount? The **coverage** report is denominated in usage. Of the eligible On-Demand usage you ran during the period, what percentage was covered by a commitment? The rest ran at undiscounted rates. Low coverage is not money lost in the same sense — you got the compute you paid for — it is **savings foregone**. Nothing is wasted; you simply paid list price for something that had a cheaper path available. Coverage drops when the fleet grows past the commitment level, when new workloads land outside the scope of existing commitments, or when a commitment reaches the end of its term and nothing replaces it. Coverage will also never reach 100% for a fleet with genuinely variable demand, and it should not: the portion of usage that only exists at peak is exactly the portion you do not want to commit to. ## Reading them together Each number alone is ambiguous; the pair is diagnostic. - **High utilization, low coverage.** The commitment is fully consumed but only covers a slice of the fleet. The commitment is too small for the steady-state baseline, or scoped too narrowly to reach the rest of it. - **Low utilization, decent coverage.** Something the commitment was bought for went away or moved. Look for a decommissioned workload, a family or region change, or a scope mismatch. - **Low on both.** Usually a scope problem rather than a sizing one — the commitment cannot match the usage that exists, so it is simultaneously unused and leaving the fleet undiscounted. - **High on both.** A stable fleet with a well-fitted commitment. Legitimate, but check that the commitment term is not about to expire with nothing queued behind it. ## Operational use These reports are for **monitoring a commitment you already hold**, not for deciding what to buy — the purchasing decision is a different conversation involving term, payment option and flexibility. The practical habits are: review utilization on a fixed cadence rather than at renewal, because it decays silently; watch coverage as a leading indicator of fleet growth, since a steadily falling coverage percentage means the baseline has outgrown the commitment; and set an automated alert on utilization in particular, since a commitment quietly going unused is the most expensive of the two failures and the least visible in a total cost chart. AWS Budgets supports commitment utilization and coverage budget types precisely so this can be alerted rather than remembered. One further nuance worth stating: a report showing a healthy percentage does not prove the commitment is the right one. A commitment that covers the fleet perfectly, and is fully used, may still be attached to workloads that should not exist at that size at all. Utilization and coverage measure how well the discount fits the usage — never whether the usage itself is justified.
- Should the goal be 100% coverage?No. Coverage should track the steady-state baseline of the fleet, not its peak. The portion of usage that only appears at peak or in short bursts is exactly the portion you do not want to commit to, because committing to it converts a variable cost into a fixed one that goes unused most of the time — which then shows up as poor utilization.
- Utilization on a reservation drops sharply one month with no change in total spend. What do you look for?A scope mismatch rather than a spend change. The usual causes are the fleet moving to a different instance family, size or region that the reservation cannot match, workloads consolidating onto different instance shapes, or an availability-zone-scoped reservation whose capacity no longer lines up. Total cost stays flat because the same usage is simply being billed On-Demand instead of discounted.
- How would you get alerted on this rather than remembering to look?AWS Budgets supports budget types that track Savings Plans and reservation utilization and coverage directly, with thresholds and notifications like any other budget. Setting a utilization budget with an alert below your acceptable floor turns a silent, ongoing loss into a notification, which matters because unused commitment never appears as a spike in a cost chart.
Utilization is how much of a season ticket you actually travelled on; coverage is how many of your journeys the season ticket paid for. Both can be low at once if you bought the ticket for the wrong line.
saying these in an interview costs you the question
- Using utilization and coverage as interchangeable terms
- Treating 100% coverage as the target for every fleet
- Assuming high utilization proves the commitment is right-sized
- Only checking these reports at renewal time
- Reading low coverage as wasted money rather than foregone savings