Uptime SLA Explained: What 99.9% Really Means
You've seen the numbers everywhere. "99.9% uptime guaranteed." "99.99% SLA." "Five nines of reliability." They sound impressive — so close to perfect that downtime seems like a non-issue.
But these percentages hide something important: even tiny fractions of a percent translate into real, measurable downtime. And depending on when that downtime happens, the difference between 99.9% and 99.99% can mean lost revenue, missed alerts, and frustrated customers.
Let's translate the marketing numbers into plain English.
What an SLA Actually Is
SLA stands for Service Level Agreement — a commitment, usually from a provider to a customer, about how reliable a service will be. The uptime portion of an SLA is expressed as a percentage: the fraction of time the service promises to be available.
An uptime SLA does two things:
- Sets expectations — how much downtime you can reasonably expect
- Defines consequences — often, if the provider misses the SLA, you get service credits or refunds
When you offer a service, your uptime is the promise you're implicitly (or explicitly) making to your users. When you consume a service, the SLA is the promise being made to you.
The Numbers Translated
Here's what each uptime level actually allows, in real time. This is the table worth bookmarking:
| Uptime SLA | Downtime per day | Downtime per month | Downtime per year |
|---|---|---|---|
| 99% ("two nines") | 14m 24s | ~7h 18m | ~3d 15h |
| 99.9% ("three nines") | 1m 26s | ~43m 49s | ~8h 46m |
| 99.95% | 43s | ~21m 54s | ~4h 23m |
| 99.99% ("four nines") | 8.6s | ~4m 23s | ~52m 35s |
| 99.999% ("five nines") | 0.86s | ~26s | ~5m 15s |
A few things jump out:
- 99% sounds high but is actually weak. It allows over 7 hours of downtime per month. For anything customer-facing, that's a lot.
- 99.9% is the realistic sweet spot for most small businesses — about 43 minutes of allowed downtime per month.
- Each additional nine is roughly a 10x improvement — and usually a large jump in cost and engineering complexity.
- Five nines (26 seconds/month) is extremely hard and typically reserved for critical infrastructure with significant redundancy investment.
Why the Percentage Doesn't Tell the Whole Story
Here's the catch: uptime percentage treats all downtime equally. But not all downtime is equal.
Consider two services, both at 99.9% uptime for the month (43 minutes of downtime):
- Service A had one 43-minute outage at 3 AM on a Sunday. Almost nobody noticed.
- Service B had 43 one-minute outages scattered throughout peak business hours. Hundreds of users hit errors, abandoned carts, and failed requests.
Same SLA. Wildly different real-world impact. This is exactly why an uptime report matters more than a single headline number — when and how downtime happens is often more important than the raw percentage.
SLA vs. Reality: What You Actually Measure
There's an important distinction between the SLA (the promise) and your measured uptime (what actually happened).
Your measured uptime depends heavily on your check interval:
- With 5-minute checks, a 2-minute outage might fall entirely between checks and never be recorded. Your measured uptime looks better than reality.
- With 1-minute checks, you catch far more short outages, so your measured number is more accurate — and often slightly lower.
This means a provider claiming "99.99% uptime" measured with infrequent internal checks may actually deliver less. Your own external monitoring — checking from outside their network — is the only way to know what you actually experience.
How to Choose Your Own Uptime Target
If you're setting an uptime goal for your own service, be realistic:
99.9% (three nines) is the right target for most small teams and SaaS products. It's achievable without massive infrastructure investment, and 43 minutes of monthly downtime budget gives you room for deploys, occasional incidents, and maintenance.
99.95%-99.99% makes sense if you're revenue-critical — an e-commerce checkout, a payment API, or a SaaS product where minutes of downtime cost real money. Reaching these levels requires redundancy, fast incident response, and tight monitoring.
Don't promise 100%. It's not achievable, and customers know it. A realistic, honestly-stated SLA builds more trust than an impossible promise.
What This Means for Monitoring
Whatever your target, you can't manage what you don't measure. To track uptime against an SLA, you need:
- Consistent monitoring — checks at a regular interval, 24/7
- Accurate incident records — every outage logged with start time, duration, and cause
- A check interval that matches your target — if you're aiming for 99.99% (4 minutes/month), 5-minute checks won't give you accurate enough data
- A way to communicate status — a public status page that shows your real uptime builds trust
The percentage on your dashboard is only as honest as the monitoring behind it. Frequent checks, complete incident history, and transparent reporting turn an SLA from a marketing number into a real, measurable commitment.
The Bottom Line
"99.9% uptime" isn't a near-perfect guarantee — it's a budget of about 43 minutes of downtime per month. Each additional nine cuts that roughly 10x but costs significantly more to achieve. And the percentage alone never tells you what matters most: whether downtime hit at 3 AM on a Sunday or during your busiest hour.
Understand the numbers, set a realistic target, measure honestly, and communicate transparently. That's worth more than any marketing claim of perfect uptime.
To learn how to interpret the data once you're tracking it, read How to Read an Uptime Report.