Uptrack

April 1, 2026

What is a good uptime SLA? 99.9% vs 99.99% explained

"We guarantee 99.9% uptime." Sounds impressive, right? That's almost 100%. Surely nothing bad can happen in that 0.1%.

Except 0.1% of a year is 8 hours and 46 minutes. That's an entire workday where your service could be completely unreachable, and you'd still be within SLA. When a customer asks "what's your uptime guarantee?" they hear "your service will basically always be up." What the number actually means is very different.

The nines: what each SLA actually means

Each additional "nine" in an uptime SLA represents a 10x reduction in allowed downtime. The jump from 99.9% to 99.99% isn't a small improvement. It's the difference between "we can go down for a few minutes every month" and "we can barely go down at all."

SLADowntime / yearDowntime / monthDowntime / week
99%3.65 days7.3 hours1.68 hours
99.5%1.83 days3.65 hours50.4 min
99.9%8.77 hours43.8 min10.1 min
99.95%4.38 hours21.9 min5.04 min
99.99%52.6 min4.38 min1.01 min
99.999%5.26 min26.3 sec6.05 sec

Most SaaS companies offer 99.9% (three nines). Major cloud providers like AWS and Google Cloud offer 99.95% to 99.99% on core services. Five nines (99.999%) is reserved for systems like telephone networks and financial infrastructure.

How to calculate your uptime SLA

The formula is straightforward. Track total time and subtract downtime.

SLA formula

Uptime % = ((Total minutes - Downtime minutes) / Total minutes) * 100

Example (one month):
  Total minutes:    43,200 (30 days)
  Downtime minutes: 45
  Uptime %:         ((43200 - 45) / 43200) * 100 = 99.896%

  Verdict: Missed 99.9% SLA by 1.8 minutes

The tricky part is defining "downtime." Some SLAs exclude scheduled maintenance windows, only count downtime during business hours, or require the customer to report the outage before the clock starts. Always read the fine print.

Watch out: Some providers calculate SLA based on "successful requests" instead of time-based availability. If 1 out of every 1,000 requests fails, that's 99.9% by their math, even if those failures happen in a concentrated 2-hour burst. Time-based SLAs are more honest.

Which SLA does your product need?

The answer depends on what happens when you go down. Not every product needs five nines. Most don't need four.

99.9% — Most SaaS products

If your product is a project management tool, CRM, analytics dashboard, or content platform, 99.9% is the industry standard. Your customers expect the product to work, but occasional brief outages are tolerable. You'll budget for about 44 minutes of downtime per month.

99.95% — Business-critical SaaS

If customers depend on your product for real-time operations (communication tools, e-commerce platforms, booking systems), you need a tighter target. 99.95% gives you about 22 minutes per month. Slack, for reference, targets this range.

99.99% — Infrastructure and payments

Payment processing, authentication services, and infrastructure APIs need four nines. Downtime means your customers' customers are affected. At 99.99%, you get about 4 minutes of downtime per month. This requires redundancy at every layer: multi-region, automated failover, zero- downtime deployments.

How monitoring interval affects SLA tracking

Your monitoring check interval directly impacts how accurately you can measure and respond to downtime. If you check every 5 minutes, your downtime measurement has a 5-minute margin of error in both directions.

5-minute checks

Your site goes down at 14:01. Your monitor last checked at 14:00 (passed). The next check at 14:05 detects the failure. You're alerted 4 minutes after the outage started. If you fix it at 14:08, the next passing check is at 14:10. Your measured downtime is 10 minutes, but actual downtime was 7 minutes.

30-second checks

Same scenario. Your site goes down at 14:01. Your monitor checks at 14:01 or 14:02 and detects the failure immediately. You're alerted within 60 seconds. You fix it at 14:08, and the next passing check confirms at 14:08 or 14:09. Your measured downtime is accurate to within 1 minute.

The math: With 30-second checks on Uptrack's free plan, you detect outages in under 60 seconds. With UptimeRobot's free plan (5-minute checks), the same outage takes up to 5 minutes to detect. That's the difference between a 99.9% SLA with 43 minutes of allowed monthly downtime and burning through 11% of your budget before you even know something is wrong.

How to actually hit your uptime target

An SLA is a promise. Monitoring is how you keep it. Here are the engineering practices that separate teams who hit their SLA from teams who pay refunds.

Deploy with zero-downtime strategies -- Rolling deployments, blue-green, or canary releases. Every deployment that takes your site down for 30 seconds chips away at your SLA budget. At 99.99%, you only get 4.38 minutes per month total.
Monitor faster than your SLA requires -- If you promise 99.9% uptime (43 minutes/month of allowed downtime), you cannot afford to detect issues with 5-minute checks. By the time you detect and fix, half your budget is gone. Use 1-minute or faster checks.
Automate incident response -- The fastest fix is the one that happens without human intervention. Auto-restart crashed processes, auto-scale on load spikes, auto-failover to standby databases. Human response time is the bottleneck.
Track your SLA budget in real time -- Know how much downtime you have left this month. If you've already used 30 of your 43 allowed minutes, postpone that risky deployment until next month.
Use consecutive-check confirmation -- False alerts waste time. If your team spends 15 minutes investigating a false alarm, that's 15 minutes they're not shipping. Consecutive-check confirmation ensures you only respond to real issues.

FAQ

Does scheduled maintenance count against SLA?

It depends on the SLA agreement. Most SaaS providers exclude pre-announced maintenance windows from SLA calculations. But your customers still experience downtime during maintenance. The best approach is zero-downtime deployments so maintenance windows become irrelevant.

What happens when you breach your SLA?

Typically, SLA breaches result in service credits — a percentage of the customer's bill credited back. AWS, for example, offers 10% credit for availability below 99.99% and 30% below 99.0%. The financial penalty incentivizes providers to invest in reliability.

Can I offer 99.99% uptime on a single server?

Realistically, no. A single server means any hardware failure, kernel update, or network issue takes you down completely. Four nines requires redundancy: multiple servers, load balancing, database replication, and automated failover. A single server can reasonably achieve 99.5% to 99.9%.

Track Your Uptime SLA for Free

5 monitors free — all at 30-second checks. No credit card required.

Start Monitoring Free