skip to content

Your commitment report shows high coverage and low utilisation — what does each measure, and which one is costing money?

level: seniorimportance: must knowfreq 55%

answer

  1. two ratios, two different questions
  2. usage side against purchase side
  3. buying more moves them oppositely
  4. size to the trough, not the mean
  5. unconsumed promise is money for nothing

basics

~20 s

Coverage is the share of eligible usage billed at a committed rate; utilisation is the share of the commitment actually consumed. High coverage with low utilisation means the commitment was sized above the workload's floor, and the unconsumed part is cash spent on nothing.

solid answer

~50 s

The two ratios look at opposite ends of the same purchase. **Coverage** divides discounted usage by all eligible usage and answers whether you left discountable work at the undiscounted rate. **Utilisation** divides the commitment actually consumed by the commitment paid for and answers whether you bought more promise than the workload turns up to use. Buying more commitment pushes them apart: above the workload's floor each added unit is consumed only in the busier hours, so coverage keeps climbing while utilisation falls. That is the pattern you are describing, and it means the commitment was sized above the floor — probably against a monthly average rather than the trough. Low utilisation is the one costing real money, because unconsumed commitment is billed for nothing at all; low coverage only means a missed discount on usage you were going to pay for anyway.

code

pseudocode · 17 lines
pseudocode
committedUnitsPerHour = 45
eligibleUsagePerHour  = [30, 30, 35, 55, 60, 30]

coveredUsage   = 0
eligibleUsage  = 0
commitmentUsed = 0
commitmentPaid = 0

for each hourUsage in eligibleUsagePerHour:
    eligibleUsage  = eligibleUsage + hourUsage
    usedThisHour   = min(hourUsage, committedUnitsPerHour)
    coveredUsage   = coveredUsage + usedThisHour
    commitmentUsed = commitmentUsed + usedThisHour
    commitmentPaid = commitmentPaid + committedUnitsPerHour

coverage    = coveredUsage / eligibleUsage      // 215 / 240 = 0.90
utilisation = commitmentUsed / commitmentPaid   // 215 / 270 = 0.80

go deeper

for a junior

Learn the two words apart: coverage is about how much of your usage got the discount, utilisation is about how much of what you paid for was actually used.

for a middle

Explain the formulas and why they diverge once the commitment is sized above the level the workload never drops below.

for a senior

Diagnose from the pair: name which sizing error each combination implies, compute both at the granularity the instrument charges at, and correct an over-sized commitment by letting it expire.

for a principal

Set what the organisation optimises for. A standing target of very high coverage quietly instructs teams to over-commit, so the target has to name a utilisation floor alongside it.

## Two ratios that answer different questions A commitment is judged after it is bought, not by the headline rate it advertised. Two measures do that judging, and they are routinely swapped: - **Coverage** = discounted usage / all eligible usage. It looks at the **workload side**: of everything that could have been billed at a committed rate, how much was? Low coverage means usage ran at the undiscounted rate that need not have. - **Utilisation** = commitment consumed / commitment paid for. It looks at the **purchase side**: of everything you promised to pay for, how much did the workload actually turn up to use? Low utilisation means paying for nothing. Both are ratios over a period, and the period matters: a month is long enough to hide the hours that decide both numbers. ## Why they pull in opposite directions Start from nothing committed and add commitment one unit at a time. 1. **Below the floor** — the level the workload never drops below — every added unit is consumed in every hour. Coverage rises, utilisation stays high. This is free ground. 2. **Above the floor**, an added unit is consumed only in the busier hours and idles in the quiet ones. It still covers usage when it is consumed, so coverage keeps rising, but it is paid for in every hour, so utilisation falls. That asymmetry is the whole reason the pair is reported together. Coverage on its own can always be driven to a high number by buying more, which is precisely the purchase that destroys utilisation. ## Reading the two together | Coverage | Utilisation | What it means | What to do | |---|---|---|---| | High | High | Sized close to the floor | Leave it; renew at the same size | | High | Low | Sized above the floor, part idles every hour | Do not renew at that size; let it expire down | | Low | High | Sized below the floor, discount left on the table | Room to commit more against the durable part | | Low | Low | The commitment does not match the usage that exists | Check the shape and scope, not the quantity | The fourth row is the one people miss: if a narrow commitment is pointed at a resource shape the workload no longer runs, it is simultaneously unconsumed and covering nothing, and buying more of it makes both numbers worse. ## Size against the floor, not the average A variable daily curve has three candidate sizing points, and only one is right: - **The peak** covers every hour and leaves most of the commitment idle for most of the day. - **The average** feels balanced and is not: hours above it are uncovered and hours below it are wasted, and the waste does not cancel the missed discount, because they are different money. One is a full charge for nothing; the other is a discount you did not take on a charge you were paying anyway. - **The floor**, the level the curve never drops below over a representative period, keeps utilisation high by construction and leaves the variable top of the curve on the undiscounted meter, where it belongs. ## Measurement traps - **Reporting lags.** Spend is reported in arrears, so a decision made on this week's figures is made on last week's shape. - **Monthly averages hide the trough.** Utilisation computed over a month can look acceptable while a handful of quiet hours every night are pure waste. Compute it at the granularity the instrument charges at. - **The coverage denominator is not the whole bill.** Only usage the instrument is eligible to discount belongs in it. Including everything makes coverage look terrible and invites buying more commitment than the eligible usage can absorb. - **Shared consumption.** Where a commitment applies across a grouping of accounts, whoever runs first consumes it, so a healthy group utilisation figure can sit on top of one team whose own usage is never covered. ## Judging the purchase after the fact 1. Compute both ratios at hourly granularity over a full representative period, including weekends and any quiet season. 2. Fix the low-utilisation problem by **not renewing**, not by buying more. An over-sized commitment corrects itself at expiry; buying another one on top does not. 3. Fix the low-coverage problem by committing more **only against the floor** you measured, and leave the variable top uncommitted. 4. Re-measure after each change, because the two ratios move in opposite directions and a fix aimed at one will move the other.

  • Why does buying more commitment raise coverage and lower utilisation at once?
    Above the workload's floor, an added unit is consumed only in the busier hours. In those hours it covers usage that was previously undiscounted, so coverage climbs; in the quiet hours it is paid for and idle, so utilisation falls. Below the floor both rise together.
  • What does high utilisation with low coverage tell you?
    That the commitment is fully consumed but small relative to eligible usage, so a large part of the workload is still billed at the undiscounted rate. It is the safe direction of error: nothing is being wasted, and there is room to commit more against the durable part of the baseline.

saying these in an interview costs you the question

  • Uses coverage alone to judge whether a commitment was a good buy
  • Says low utilisation is harmless because the rate is still lower
  • Sizes the commitment at the average of a variable curve
  • Reads a monthly average and misses the quiet hours below the floor
  • Believes buying more commitment improves both numbers together
  • Fixes low utilisation by buying an additional commitment