SLA Uptime Calculator Guide: Turn a Percentage Into Allowed Downtime

Service level agreements, or SLAs, are written in percentages, and percentages are good at sounding reassuring while hiding the numbers that matter. A promise of 99.9 percent uptime sounds almost perfect, and a promise of 99.99 percent sounds better still, yet the difference between them is not a small refinement. It is the difference between about forty-three minutes of allowed downtime in a month and about four minutes. When a service you depend on goes down, you do not experience a percentage. You experience minutes of waiting, customers unable to check out, or staff unable to work. An uptime calculator translates the percentage into those minutes, and that translation is the single most useful thing you can do before signing or relying on an SLA.

This guide explains the arithmetic, shows how to compare SLAs fairly, and explains the exclusions and measurement details that often matter more than the headline percentage. It is written for buyers, managers, and small business owners rather than lawyers, and it does not replace legal advice on a specific contract. But the arithmetic is universal, and once you can do it, no SLA headline will mislead you again.

The core formula

Allowed downtime is the failure percentage multiplied by the length of the period. If uptime is U percent, the failure percentage is 100 minus U. For 99.9 percent uptime, the failure percentage is 0.1 percent, which as a decimal is 0.001. Multiply that by the period: a 30-day month contains 30 times 24 times 60 minutes, which equals 43,200 minutes, and 0.001 times 43,200 equals 43.2 minutes. That is the whole formula, and every uptime figure in this guide comes from it. For a year, use 365 days, which is 525,600 minutes, and for a 31-day month, use 44,640 minutes. The choice of period matters, because an SLA quoted monthly and an SLA quoted annually allow different patterns of failure: a monthly SLA limits each bad month, while an annual SLA can hide a terrible month inside a good year.

The famous "nines" shorthand follows directly from the formula. Each additional nine divides the failure percentage by ten, and therefore divides allowed downtime by ten. At 99 percent, allowed downtime in a 30-day month is 1 percent of 43,200 minutes, which is 432 minutes, or 7.2 hours. At 99.9 percent, it is 43.2 minutes. At 99.99 percent, it is 4.32 minutes. At 99.999 percent, it is 0.432 minutes, about 26 seconds. The jump from three nines to four nines is therefore not a 0.09 percent improvement in any meaningful sense; it is a tenfold reduction in allowed downtime, and it usually costs far more than ten times as much to deliver, because the engineering required to remove the last minutes of downtime is much harder than the engineering that removed the first hours.

Worked example with sample values

Suppose a sample service promises 99.95 percent uptime in a 30-day month. First find the failure percentage: 100 minus 99.95 equals 0.05 percent, which as a decimal is 0.0005. Next find the minutes in the period: 30 times 24 equals 720 hours, and 720 times 60 equals 43,200 minutes. Multiply: 0.0005 times 43,200 equals 21.6 minutes, because 43,200 divided by 2,000 is 21.6. So the SLA allows 21.6 minutes of downtime in that month. Check the result against its neighbours: 99.9 percent allows 43.2 minutes and 99.99 percent allows 4.32 minutes, so 21.6 minutes sits sensibly between them, closer to the 99.9 percent figure, which matches 99.95 percent sitting halfway between 99.9 and 99.99 on a logarithmic scale rather than a linear one.

Now convert the same SLA to a year. A year has 525,600 minutes, and 0.0005 times 525,600 equals 262.8 minutes, which is 4 hours and 22.8 minutes, since 262.8 divided by 60 is 4.38 hours. Notice that twelve months at 21.6 minutes each would be 259.2 minutes, slightly less than the annual figure, because twelve 30-day months total 360 days rather than 365. Small differences like this are why you should always state the period alongside any downtime figure. A downtime budget quoted without a period is incomplete, and a vendor who quotes an annual figure for a monthly SLA, or the reverse, may be obscuring the real commitment, intentionally or not.

What an SLA usually does not cover

The headline percentage is only the start of the contract. Most SLAs exclude scheduled maintenance, which the vendor announces in advance, and that maintenance can be substantial if it is scheduled weekly. Most exclude downtime caused by factors outside the vendor's control, such as internet outages between you and the vendor, problems with your own equipment, or force majeure events. Many measure uptime per service or per region rather than for your specific account, so your experience can be worse than the reported figure if you happen to be in the affected region. Some measure only whether the service responds at all, not whether it responds quickly or correctly, so a service that is technically up but unusably slow may still count as available.

Remedies matter as much as the percentage. A typical SLA remedy is a service credit, a percentage of the monthly fee refunded or credited, rather than compensation for your actual losses, which are usually excluded. Credits are often tiered: a small credit for missing the target slightly, a larger credit for a major breach, and almost never a full refund. Credits usually must be claimed within a set time, with evidence, and some contracts cap total credits at a fraction of the fee. Before relying on an SLA, read the remedy section and ask what happens in the worst realistic case: if the service is down for a day, do you get a small credit, or does the contract give you a right to leave? The answer shapes how much redundancy you need on your side, regardless of the percentage.

Comparing SLAs fairly

Compare SLAs on the same period, with the same exclusions in mind, and convert every percentage to minutes before judging it. A 99.9 percent monthly SLA and a 99.95 percent annual SLA are not directly comparable, because the annual figure can absorb a bad month. A vendor offering 99.99 percent but excluding four hours of weekly maintenance is offering less availability than a vendor offering 99.9 percent with no exclusions, if you need the service during those maintenance windows. Ask vendors how uptime is measured, who measures it, and whether you can see the measurement data or only a summary. Independent monitoring from your side, even a simple check every few minutes, gives you evidence if you ever need to claim a credit, and it often reveals patterns, such as regular short outages, that monthly averages hide.

Think about your own tolerance, not the vendor's marketing. A brochure website can tolerate an hour of downtime at night with little harm. An online store cannot tolerate an hour during a sale. A system that supports emergency work cannot tolerate minutes at any time. Match the SLA to the tolerance, and where the tolerance is tight, do not rely on a single vendor's SLA at all: build redundancy, such as a second provider, cached content, or a manual fallback process, because even a 99.999 percent SLA allows about five minutes of downtime a year, and those five minutes may arrive at the worst possible moment.

Common mistakes to avoid

The first mistake is reading 99.9 percent as "almost never down" without converting it to 43 minutes a month. The second is comparing percentages without converting them to minutes on the same period. The third is ignoring exclusions, especially scheduled maintenance and third-party causes. The fourth is assuming a service credit compensates for lost business; it almost never does, and the contract usually says so. The fifth is forgetting that uptime is an average: a service can meet its monthly SLA while still suffering several short outages that each disrupt your work, because the SLA counts total minutes, not the number or timing of incidents. If short, frequent outages harm you more than one long outage, ask about incident frequency as well as total uptime, and design your own monitoring and fallback around what you actually experience.

Try it next

Try the SLA Uptime Calculator Guide tool with your own numbers, then explore the Percentage Calculator and the Average Calculator for related calculations. For a companion read, see /blog/average-calculator-guide/, which applies the same measure-carefully habit to a different problem.