SLA Reality Check
What each uptime SLA actually allows
And the real-world dollar cost it represents at $50K MRR with 2% churn impact.
| Uptime SLA | Allowed yearly downtime | Typical annual cost (estimate) |
|---|---|---|
| 99% | 3d 15h 39m | $50,000+ |
| 99.5% | 1d 19h 49m | $25,000+ |
| 99.9% | 8h 45m 36s | $5,000+ |
| 99.95% | 4h 22m 48s | $2,500+ |
| 99.99% | 52m 33s | $500+ |
| 99.999% | 5m 15s | $50+ |
The Detection Problem
How most teams find out about downtime
Faster detection is the cheapest way to reduce the cost of an outage. Every minute matters.
1. Customer support tickets
Average detection time: 30-90 minutes. By the time the first ticket arrives, you've already lost dozens of conversions and have an angry user base.
2. Twitter mentions
Average detection time: 60-120 minutes. Your VP of Marketing screenshots a tweet and forwards it to engineering. Bad day for everyone.
3. Real-time monitoring
Average detection time: under 60 seconds. An automated notification reaches your team as soon as error rates spike. Fix it before customers notice.
FAQ
Frequently asked questions
How much does API downtime really cost?
It depends on your business model, but the Gartner average across industries is $5,600 per minute of downtime. For SaaS specifically, the cost has two parts: direct revenue loss prorated from your MRR, and churn impact from customers who cancel after a visible incident. Use the calculator above to estimate your specific number.
What is a realistic churn rate caused by an outage?
For most B2B SaaS, a single visible outage causes 1-3% churn from affected customers. For payment APIs, e-commerce checkouts, or critical infrastructure the rate can hit 5-10%. The longer the outage and the more customers it affects, the higher the rate.
How do I detect downtime faster?
Use a monitoring tool that combines external uptime checks (pings from outside your infrastructure) with internal execution monitoring (metrics from inside your server). External pings catch DNS, CDN, and platform outages. Internal monitoring catches slow database queries and silent failures. Route notifications to the channel your on-call team actually watches (Slack, email, SMS or a paging tool) so a spike in error rates reaches a person within minutes.
How can I reduce the cost of API downtime?
Three ways: (1) detect failures faster, every minute saved reduces both direct loss and churn impact, (2) communicate proactively with customers during incidents, which dramatically reduces churn vs. silent failures, (3) invest in redundancy for critical paths like checkout and authentication.