What a service credit is, and what it is not
A service credit is a percentage of what you pay the vendor, refunded against a future invoice when availability falls below a committed figure. It is not damages, it is not insurance, and in most contracts it is the exclusive remedy — meaning you agree in advance not to pursue anything else for the outage. That last clause is the reason this calculation matters: you are signing away a claim in exchange for a discount whose size you can compute before you sign.
The mismatch is structural, not a case of a stingy vendor. The credit is bounded by your spend with them. The loss is bounded by your customers' spend with you. For most businesses those two numbers differ by an order of magnitude or more, so even a 100% credit — a full month of fees returned — typically covers a small fraction of what a long outage costs. The calculator makes that ratio explicit instead of leaving it as a vague suspicion.
The second thing the calculation reveals is subtler. Even when the vendor performs exactly to its commitment, the allowed downtime still costs you something. A 99.9% monthly SLA permits 43.8 minutes of outage every month, and you absorb all of it with no credit at all. That figure — what full compliance costs you — is often larger than the credit you would earn for a modest breach, which is a useful thing to know before negotiating a target upward.
How the two sides are computed
Downtime. A billing month is treated as 730 hours, which is 8,760 ÷ 12 and matches the monthly-uptime-percentage definitions used by the major cloud providers. Downtime is 730 × (1 − availability), so 99.5% availability is 3.65 hours and 99.9% is 43.8 minutes. Some SLAs use actual calendar-month minutes instead, which moves the figure by up to about 1.6% between February and a 31-day month — immaterial to the comparison, but check which convention your contract uses before disputing a number.
The credit. Credit schedules are step functions, not proportional. The typical shape has three tiers: a small percentage once you drop below the target, a larger one below 99.0%, and the largest below 95.0%. The consequence is that the credit does not increase with the length of the outage within a tier — an outage of two hours and one of six hours can earn exactly the same credit. This calculator applies your own tier percentages, and it returns zero credit whenever the achieved availability meets the target, regardless of how the tiers are set.
Your loss. Downtime hours multiplied by the revenue riding on the service, multiplied by the share you never recover, plus your own response cost. The recovery share is what separates delayed revenue from destroyed revenue and is worth measuring rather than assuming; the incident response cost calculator breaks the same loss down in more detail across responder time, displaced work and credits owed to your own customers.
Coverage. Credit divided by loss. Because the credit steps and the loss rises continuously, the coverage share does not move in one direction as availability falls — it jumps upward each time a tier boundary is crossed and declines between boundaries. The table below shows that shape across the whole range at your figures.
Worked example: 99.5% delivered against a 99.9% commitment
Take the defaults: you pay $20,000 a month, the SLA commits to 99.9%, the vendor delivered 99.5%, the credit schedule is 10% / 25% / 100%, your service carries $15,000 an hour of revenue, you lose 70% of it permanently, and the outage cost you $8,000 of your own response effort.
- Downtime incurred. 730 × (1 − 0.995) = 3.65 hours.
- Downtime allowed. 730 × (1 − 0.999) = 0.73 hours, or 43.8 minutes.
- Excess. 3.65 − 0.73 = 2.92 hours beyond the commitment.
- Credit tier. 99.5% is below the 99.9% target but at or above 99.0%, so the first tier applies: 10%.
- Credit earned. $20,000 × 10% = $2,000.
- Effective loss rate. $15,000 × 70% = $10,500 an hour.
- Revenue lost. 3.65 × $10,500 = $38,325.
- Total loss. $38,325 + $8,000 = $46,325.
- Coverage. $2,000 ÷ $46,325 = 4.32%, leaving $44,325 uncovered.
- Cost of full compliance. 0.73 × $10,500 = $7,665 — what the SLA's own allowance costs you in a month where the vendor performs perfectly.
The last two lines together are the argument. The credit for missing the target is $2,000; the downtime the target permits costs you $7,665. You are being compensated with less than a third of what compliance itself costs you, and with 4.32% of what this particular breach cost. Neither figure is unusual.
How to use the coverage figure
Read it as a negotiating input, not as a grievance. Vendors set credits to be a service-quality signal, not an indemnity, and no amount of arguing changes the structural cap. What the number does support is a specific ask: a higher committed target, a shorter measurement window, or an outage-based termination right that lets you leave after repeated breaches. Those are worth far more than a larger credit percentage.
Check the exclusive-remedy clause. If credits are the sole remedy, the coverage percentage is the entire recovery you will ever see for an outage, and the uncovered figure is the risk you are carrying yourself. That is the risk to price into your own architecture — the backup and DR cost calculator and the cost of additional nines calculator both price ways of buying it down, and either is more likely to pay off than a credit negotiation.
Watch the claim procedure. Most SLAs require you to request the credit within a defined window — often thirty days — with your own evidence of impact, and unclaimed credits are simply not paid. A credit worth 4.32% of your loss is still worth having, and the administrative cost of claiming it is small. Put the deadline in the incident process rather than relying on someone remembering.
Compare targets on allowed downtime, not on nines. The gap between 99.9% and 99.95% sounds trivial and halves your monthly allowance from 43.8 to 21.9 minutes. The uptime and downtime calculator converts any target into minutes for any window, which is the form in which SLA targets can actually be compared.
Downtime allowed by common SLA targets
| Target | Per month | Per quarter | Per year |
|---|---|---|---|
| 99.0% | 7 h 18 m | 21 h 54 m | 87 h 36 m |
| 99.5% | 3 h 39 m | 10 h 57 m | 43 h 48 m |
| 99.9% | 43.8 m | 2 h 11.4 m | 8 h 45.6 m |
| 99.95% | 21.9 m | 65.7 m | 4 h 22.8 m |
| 99.99% | 4.38 m | 13.14 m | 52.56 m |
| 99.999% | 26.3 s | 78.8 s | 5.26 m |
A monthly window is more forgiving to the vendor than an annual one at the same percentage, because a bad month cannot be offset by good ones and, more importantly, twelve separate allowances add up to the annual figure with no memory between them.
Clauses that change the arithmetic
- Exclusive remedy. Credits are your only recourse. This is the default in most standard terms and it converts the coverage percentage into your total recovery.
- Credit cap. Many schedules cap total monthly credits at a percentage of the fee even when several services breach at once. Check whether the top tier here is really reachable.
- Claim window and evidence. A deadline of thirty days is common, along with a requirement to supply your own logs. Miss it and the credit is forfeited regardless of the outage.
- Excluded downtime. Scheduled maintenance, force majeure, problems attributed to your own configuration, and anything on a beta or preview service are typically outside the measured availability. Your monitoring will show more downtime than the SLA counts.
- Measurement definition. Availability may be measured per region, per service, per API call or as an aggregate, and the choice moves the number materially. An outage that takes down one region may not breach a global aggregate at all.
- Credit form. Credits are usually applied against future invoices rather than refunded in cash, which means they are worth nothing if you leave the vendor — the exact circumstance a serious outage makes likely.
A credit is not a substitute for architecture
If the uncovered figure is large enough to matter, the answer is not a better contract but a design that survives the vendor's failure: a second provider, a degraded read-only mode, queueing that absorbs a short outage, or cached data that keeps the critical path alive. Those cost money and this calculation tells you how much they are worth. A vendor who offers a 99.99% target and a 25% credit is telling you plainly that they expect to be down sometimes and that they will not be paying for it.
Where this sits in vendor management
SLA credits are one of three levers in a reliability negotiation and the weakest of them. The strongest is the committed target itself, because it changes the vendor's own engineering priorities rather than compensating you afterwards. The second is a termination or price-review right triggered by repeated breach, which gives you an exit rather than a discount. Credits are the third, and they are best treated as a signal about how seriously a vendor takes availability rather than as a financial instrument.
When renewal comes around, the accumulated credit history is evidence rather than a bargaining chip — a year of breaches is a much better argument for a price concession than for a bigger credit percentage. The renewal uplift calculator covers the price side of that conversation.
Internally, the mirror image applies. If you sell an SLA to your own customers, the same arithmetic runs in reverse and your credits are a liability. Model your own exposure with your customers' fees in the fee field and your own reliability in the achieved field, and set your committed target against what the error budget calculator says you can actually deliver. Committing to a target your error budget does not support is how a reliability problem becomes a revenue problem.
Key terms
- Service credit
- A percentage of the fee for the affected service, applied against a future invoice when availability falls below the committed level.
- Exclusive remedy
- A contract term making the credit the only compensation available for a breach of the availability commitment.
- Monthly uptime percentage
- The vendor's measure of availability over a billing month, usually excluding scheduled maintenance and customer-caused problems.
- Coverage share
- Credit earned divided by actual loss. The single figure that shows how much of an outage a contract really pays for.