Effective Savings Rate
Coverage tells you what's reserved. ESR tells you what it's worth.
Effective Savings Rate is the FinOps industry's standard output metric for rate optimization: the single number that folds coverage, discount depth, and utilization into one percentage, comparable at any fleet size. Here's the formula, why coverage alone can't tell this story, and what a renewal lapse actually costs on this scale.
First, the formula
ESR = coverage × [1 − (1 − discount) / utilization].
Take your on-demand-equivalent (ODE) cost (what the fleet would cost with nothing reserved) and your amortized cost (what you actually pay, reservations included). ESR is the % you saved: 1 − amortized / ODE. AWS Cost Explorer and the Cost and Usage Dashboard both surface amortized and ODE cost natively; the FinOps Foundation's Rate Optimization capability uses this exact ratio as its output metric.
Expand amortized cost into its three moving parts: coverage (C), discount depth (D), and utilization (U). The identity reduces to one line: ESR = C · [1 − (1−D)/U]. Every number on this page comes out of that one formula, plugged with disclosed inputs, never a hardcoded percentage.
Why coverage alone lies
Two fleets, the same 27.0% ESR, and wildly different coverage.
A coverage report alone can't distinguish these two fleets. Only once discount depth and utilization are folded in does it become clear they're earning the identical rate, for very different reasons.
- Coverage
- 60%
- Discount depth
- 45%
- Utilization
- 100%
- Coverage
- 90%
- Discount depth
- 30%
- Utilization
- 100%
Fleet A reserves less than Fleet B (60% vs 90%) but at a deeper discount (45% vs 30%). The two trade-offs cancel out exactly. A coverage-only dashboard would call Fleet B the clear winner. ESR calls it a tie, correctly.
The three levers
Each lever has its own way of quietly costing you ESR.
Coverage
The share of reservable running usage matched to a reservation you own.
Fails by: Fails by under-reserving: steady usage sits on-demand for no reason, and every point of coverage left on the table is a point of ESR left unclaimed.
Discount depth
How deep the reservation's rate cuts below on-demand, set by term, payment option, and instance family.
Fails by: Fails quietly: a shallow 1-year, no-upfront reservation still counts as "covered" while leaving most of the available discount unclaimed.
Utilization
The share of what was bought that's actually being claimed by matching running usage.
Fails by: Fails hardest: a family that shrinks, resizes, or moves region keeps billing the reservation's full commitment while claiming less and less of it.
The utilization cliff
Below 55% utilization, this reservation costs MORE than on-demand.
Holding coverage and discount depth at the Growth-preset baseline (60% coverage, 45% discount) and varying only utilization: ESR crosses zero exactly at 1 − discount, 55% here. Nobody publishes this plainly, because it means a reservation can be a net loss even while every other number about it looks fine.
The renewal lapse, priced in ESR
A missed renewal isn't a footnote. It's arithmetic.
A 1-year book renews once a year, so a median expiry-to-repurchase gap of N days leaves that share of the book uncovered N days out of the year: C_eff = C · (1 − (N/365)/termYears). Plug straight back into the same ESR formula.
The dollar column below is priced against the Growth preset's $30,000/mo total AWS compute spend, of which 65% ($19,500/mo) is reservable. That reservable slice, priced at on-demand rates, is ESR's denominator: the on-demand-equivalent spend. Savings percentages never apply to the parts of a bill that were never eligible to be reserved.
| Gap | ESR | Points lost | Cost/yr |
|---|---|---|---|
| 8 days | 26.4% | −0.59 pts | $1,385/yr |
| 14 days | 26.0% | −1.04 pts | $2,423/yr |
| 30 days | 24.8% | −2.22 pts | $5,193/yr |
An 8-day median expiry-to-repurchase gap is not hypothetical. It's the median gap from the site's own multi-region SaaS case study. On the Growth preset that costs $1,385/yr, against Autopilot's $1,788/yr, most of a year of the product, recovered from one slipped renewal. The crossover, on these inputs, is at 11 days: a gap that long pays for Reserver outright, and the 14-day row already clears it at $2,423/yr. At Enterprise scale the same 8-day gap is $13,848/yr, over 7× the price, and not a close call.
Calculate your own
Effective Savings Rate. One number, three levers.
Coverage, discount depth, and utilization all move ESR — drag any of them, or the renewal gap below. The size presets only change the dollar stakes: ESR itself reads on the same axis whether you're carrying ten reservations or ten thousand.
A renewal missed by 8 days costs $1,385/yr on $19,500/mo of reservable on-demand-equivalent spend — pays for about 9 months of Reserver.
See my real numbers →Estimates from disclosed inputs, computed the same way for every preset. The three levers are whatever the sliders read. The dollar figure converts ESR points into money against the preset's on-demand-equivalent (ODE) spend — the reservable slice of its monthly AWS compute bill, shown on each preset above and taken as 65% of total spend, the same share the savings calculator uses. A savings rate never applies to the parts of a bill that were never eligible to be reserved. Reserver reads your real coverage, discount, and utilization straight from AWS Cost Explorer's own amortized and on-demand-equivalent cost lines — this calculator lets you check the shape of the math first.
Get the two real numbers
Pull amortized and on-demand-equivalent cost straight from Cost Explorer.
In AWS Cost Explorer or the Cost and Usage Dashboard, group cost by "Amortized cost" and compare it to the same usage priced "On-Demand" (or "Unblended" with RI/Savings Plans excluded) over the same period. ESR = 1 − amortized / ODE from those two totals directly. No coverage math is required to get the headline number, though coverage, discount, and utilization are what move it.
The honest limits
What ESR does not tell you.
ESR is the best single number this industry has for rate optimization. It is still one number, and it has blind spots worth stating plainly rather than papering over.
Commitment risk isn't in the number
A higher ESR usually means a bigger, longer, more exposed commitment. A 3-year All Upfront term scores better than a 1-year No Upfront one, even when the shorter term is the safer bet for a family that might change.
Realized vs. potential savings
ESR reports what you actually captured, not what a perfectly-timed, perfectly-sized fleet could have captured. A high ESR on a badly-oversized reservation still looks good on this one number.
One blended number hides per-service variance
A fleet-wide ESR can sit comfortably positive while one service or region is well below its own break-even. The aggregate doesn't tell you where to look.
It says nothing about what you paid to get here
ESR measures the discount captured, not the price of capturing it. Exit obligations, resale losses, or a vendor fee taken as a cut of the very savings this number reports are all invisible to it.
At every size
The percentage doesn't change with fleet size. The dollars do.
ESR is scale-free. A percentage reads the same whether it's measured against one account or a thousand. What changes is the dollar stakes riding on the same percentage point.
Same 27.0% ESR baseline. An 8-day renewal gap here costs $138/yr.
Same 27.0% ESR baseline. An 8-day renewal gap here costs $1,385/yr.
Same 27.0% ESR baseline. An 8-day renewal gap here costs $13,848/yr.
One flat price at every size, never a cut of what ESR reports you saved. See pricing. Or start with the coverage playbook: coverage is one of the three levers behind this number, not the whole story.
Stop overpaying for a steady fleet.
Reserver is in early access. Join the list and we'll onboard you as capacity opens up. Your first recommendation lands within minutes of connecting.