99.9% uptime, spoken as "three nines," means a service is allowed to be down for at most about 8 hours 46 minutes over a full year. That works out to roughly 43 minutes 50 seconds per month and about 10 minutes 5 seconds per week. It is the availability level most software companies promise, because it is credible to hit with good hosting and a competent on-call process, yet it still leaves real room for the occasional bad deploy.
The number looks tiny next to a plain 99%, but the gap between them is a factor of ten in allowed downtime. 99 percent uptime lets you be dark for more than three and a half days a year. Three nines cuts that to under nine hours. Knowing what does 99.9 uptime mean in actual minutes is the only way to tell whether the promise on your marketing page is one you can keep.
What "three nines" actually means
Each "nine" is a digit in the availability percentage, and every nine you add divides the downtime budget by ten. Two nines is 99%, three nines is 99.9%, four nines is 99.99%. So 99.9% availability means the service is reachable 99.9% of the time and unreachable at most 0.1% of the time. Multiply 0.1% by the number of minutes in a year and you get the yearly budget: about 526 minutes, or 8 hours 45 minutes and change.
99.9% became the common baseline SLA for a simple reason. It is honest. A well-run team on solid infrastructure can hold it without heroics, and 8 hours 46 minutes a year absorbs a couple of unplanned incidents plus routine maintenance if the contract excludes scheduled windows. Promise more and you are on the hook for automatic failover engineering. Promise less and customers notice. Three nines sits at the point where the promise is both marketable and achievable.
The 99.9% downtime table
The only honest way to read an availability target is to convert it into time you are allowed to be down. Here is the full ladder around 99.9%, so you can see what each step buys and costs.
| Availability | Down / year | Down / month | Down / week |
|---|---|---|---|
| 99% (two nines) | 3d 15h 36m | 7h 18m | 1h 40m 48s |
| 99.5% | 1d 19h 48m | 3h 39m | 50m 24s |
| 99.9% (three nines) | 8h 45m 36s | 43m 50s | 10m 5s |
| 99.95% | 4h 22m 48s | 21m 55s | 5m 2s |
| 99.99% (four nines) | 52m 34s | 4m 23s | 1m 0s |
| 99.999% (five nines) | 5m 15s | 26s | 6s |
For any target between the rows, enter the exact percentage to get its allowed downtime in hours and minutes:
alertping ▸ downtime calculator
- allowed downtime per year
- 52m 36s
- per quarter
- 13m 9s
- per month
- 4m 23s
- per week
- 1m 1s
- per day
- 8.6s
A 5-minute check interval can miss a whole month of budget at four nines. AlertPing checks every 30 seconds, so the number above is one you can actually measure.
Read across the 99.9% row and the promise stops being abstract. Forty-four minutes a month is one long incident, or three or four short ones, and you are at the edge. That is why the difference between 99.9% and 99.95% matters in a contract: doubling the nines behind the decimal halves your budget from about 44 minutes a month to under 22. If you want the arithmetic behind every figure in this table, we walk through it in how to calculate uptime percentage.
You will also see oddly specific numbers on real contracts. A vendor might quote 99.97%, which lands between three and four nines and buys roughly 2 hours 37 minutes of yearly downtime. Those in-between figures are not marketing noise. They usually come from a team that measured its own history and set the promise just below what it can consistently deliver. A precise number like 99.97% often signals more discipline than a round 99.9%, because someone did the math instead of copying a competitor.
Is 99.9% uptime good enough?
For most B2B SaaS, yes. If your product is a dashboard, a CRM, a project tool, or an internal API that people use during business hours, 99.9% is a promise you can defend and a level of reliability users will accept. An outage means a few retries and a support ticket, not a lawsuit. The 8 hours 46 minutes of annual budget gives your team enough slack to ship, patch, and recover like humans.
Where 99.9% falls short is anywhere downtime converts straight into money lost or harm done. Payment processing, health systems, and shared infrastructure that other companies build on top of cannot treat 44 minutes a month as acceptable, because during those 44 minutes someone cannot get paid, treated, or served. For those systems, a 44-minute monthly outage budget is a business risk, not a tolerance, and the expected answer is to move up a tier so the budget shrinks to minutes. Those systems reach for five nines and the automatic redundancy that comes with it. The right question is never "how high can the number go" but "what does an outage of this length actually cost me," a decision we work through in what is a good uptime percentage.
alertping
Measure your real uptime, do not estimate it
A 99.9% promise is only worth what you can prove. AlertPing checks as often as every 30 seconds from two regions and publishes a rolling uptime percentage, which is the resolution a 44-minute monthly budget actually needs.
Your check interval decides whether you can measure it
Here is the trap that catches most 99.9% claims: you cannot measure what you cannot see, and your check interval sets the smallest outage you can detect. A monitor that checks every 5 minutes reads your service as either up or down for the whole 5-minute block. It cannot tell a 30-second blip from a 4-minute one, and it can miss a short outage entirely if it happens between checks. Against a monthly budget of about 44 minutes, that coarse view is not good enough to resolve where you stand.
To measure 99.9% with any precision, checks need to run every 30 to 60 seconds, from outside your own network so the monitor sees what a customer sees. At that cadence a single missed check adds about a minute to the record instead of five, and the rolling percentage tracks the real budget closely enough to trust in a dispute. This is the core of SLA monitoring: the uptime figure is only as trustworthy as the interval behind it. Confirming each failure from more than one region keeps one flaky route from stamping false downtime onto a number that may carry the contractual obligations you have to track.
What "industry standard uptime" really is
When a vendor says it meets "industry standard uptime," they almost always mean 99.9%. It is the default line in cloud and SaaS contracts, the level below which customers push back and above which vendors get nervous about penalties. Major cloud providers publish tiered SLAs that often start around 99.9% for a single-instance service and climb toward 99.99% for multi-zone deployments, but the headline figure people reach for when they say "standard" is three nines.
Treat that phrase as a floor, not a badge. 99.9% is standard because it is the minimum most buyers will accept, which means matching it makes you ordinary, not exceptional. The teams that stand out are the ones that promise 99.9%, actually measure it with fast enough checks, and quietly run closer to 99.95% or better in practice. The number on the contract is a promise. The number your monitoring reports is the truth, and the only one worth building a reputation on.