What does a 99.9% SLA actually allow? Uptime, downtime, and your credit

"99.9% uptime" sounds like a rounding error away from perfect. It isn't. The gap between 99.9% and 99.99% is the difference between about 43 minutes of downtime a month and about 4 — a tenfold difference hiding in a single extra digit. Here's exactly how much downtime each SLA tier allows, when that downtime becomes a breach you can claim, and how to turn it into a dollar figure.

The short answer

A 99.9% uptime SLA ("three nines") allows roughly 43 minutes of downtime per calendar month — about 8 hours 46 minutes across a year. Any unplanned downtime beyond that budget in a month is an SLA breach, which typically entitles a paying customer to a service credit worth a percentage of that month's fee. Each additional nine cuts the allowance about tenfold: 99.99% permits roughly 4 minutes a month, while 99% permits over 7 hours. To know whether you're owed a credit, compare your provider's *measured* monthly uptime against its committed target — the size of the shortfall sets the credit tier.

The uptime cheat sheet: what each "nine" allows

Uptime targets are measured over a calendar month, so the allowed-downtime budget resets every month. Using a standard 30-day month:

Read down the monthly column and the jump is stark: moving from 99.9% to 99.99% shrinks your monthly downtime budget from about 43 minutes to about 4. That's why each nine is far harder — and more expensive — for a vendor to deliver than the last. On a daily basis, 99.9% works out to about 1 minute 26 seconds a day; 99.99% to under 9 seconds.

When does downtime become a breach?

An outage is an SLA breach when the provider's total monthly downtime exceeds the allowance for its committed target. Three things decide it:

• The measurement is monthly, not per-incident. A single 30-minute outage doesn't breach a 99.9% SLA on its own — the monthly budget is about 43 minutes — but a second outage the same month can push you over.
• The target is per service, not per account. Most vendors commit separately for each product or service tier, so measure the specific service that went down, not your whole bill.
• Scheduled maintenance usually doesn't count. Planned-maintenance windows are typically excluded from the uptime math, so only unplanned downtime eats your budget.

Once measured monthly uptime drops below target, the credit clause is triggered — but the credit is almost never automatic. You have to measure it and ask.

How much is the credit worth?

Most vendors publish a tiered schedule: the further below target uptime falls, the bigger the credit, expressed as a percentage of the affected service's monthly fee. AWS EC2's published structure is a representative example:

The ceiling and the middle band vary by vendor: Azure and most AWS services credit 25% for the middle tier; Google Cloud and Atlassian cap the top tier at 50%; Twilio pays a flat 10%; and Cloudflare uses a formula instead of tiers. So two vendors with the same 99.9% promise can owe very different amounts for the same outage — the tier table matters as much as the target. See the side-by-side in which SaaS vendors actually pay SLA credits.

How to calculate the SLA credit you're owed

1. Find the affected service's committed uptime target from its SLA (or its SLA guide) — commonly 99.9%, 99.95%, or 99.99%.
2. Measure the service's actual uptime for the calendar month: divide its available minutes by the total minutes in the month (excluding scheduled-maintenance minutes), then multiply by 100.
3. Confirm it's a breach by checking that measured uptime against the allowance in the cheat sheet above — if unplanned downtime exceeded the budget, the credit clause is triggered.
4. Match the shortfall to the vendor's credit tier to get the credit percentage — for example, 10% for a small miss.
5. Multiply that percentage by the affected service's monthly fee to get your credit; a vendor credit calculator does steps 2–5 for you from your spend and observed uptime.

What 99.9% looks like across real vendors

The vendors Ontracko monitors commit to a range of targets. Here's what a few of them actually promise, drawn from their published SLAs:

A 99.99% promise from Slack allows only about 4 minutes of monthly downtime before a credit is due; a 99.9% promise from a Premium Atlassian product allows about 43. Same idea, very different thresholds — always check the specific number, and read how to claim an Atlassian SLA credit for a worked vendor example. You can see which of these vendors has actually been missing lately on the reliability rankings.

Frequently asked questions

How much downtime does a 99.9% SLA allow?

About 43 minutes per calendar month, or roughly 8 hours 46 minutes per year, using a standard 30-day month. If a provider's unplanned monthly downtime exceeds that, it has breached the 99.9% SLA and typically owes a service credit for that month.

Is 99.9% uptime good?

It's the most common commercial SaaS commitment, but "three nines" still permits nearly 9 hours of downtime a year. For a system your business depends on continuously, 99.95% (about 22 minutes a month) or 99.99% (about 4 minutes a month) is meaningfully stronger — each extra nine is roughly a tenfold reduction in allowed downtime.

What's the difference between 99.9% and 99.99% uptime?

Roughly a factor of ten. 99.9% allows about 43 minutes of downtime a month; 99.99% allows about 4. Over a year that's about 8 hours 46 minutes versus about 53 minutes — a large practical gap hidden behind one extra digit.

Is an SLA breach automatic money back?

No. Credits are almost never applied automatically. You have to measure the monthly uptime shortfall, match it to the vendor's credit tier, and file a claim within the vendor's deadline. See what an SLA credit is and how to claim it.

How do I calculate the credit for a specific outage?

Measure the affected service's monthly uptime, find which credit tier the shortfall falls into, and multiply that percentage by the service's monthly fee. A credit calculator turns your spend and observed uptime into a dollar figure, and the SLA vs service credit explainer covers what form the payout takes.

Methodology & caveats

The downtime figures are arithmetic from a standard 30-day month (43,200 minutes) and a 365-day year (525,600 minutes); real months vary slightly in length, so treat them as close approximations. The credit tiers and vendor commitments are transcribed from the named vendors' published SLAs into Ontracko's profiles — percentages, caps, and covered services vary by plan and negotiated agreement, so confirm the exact clause in your own contract before filing. These are contractual availability *targets*, not a measure of any vendor's actual recent reliability — for that, see the reliability rankings and the SLA glossary.

*Ontracko monitors 144 SaaS & cloud vendors' public status feeds, detects when they miss their SLA, and recovers the credits you're owed. Monitor your vendors free — 8% only on recovered credits. See live reliability rankings or browse the SLA glossary.*