What an SLA actually contains
An SLA stops the conversation about "we'll do our best" and replaces it with numbers. A serious one defines:
- What is being measured : pages loaded, transactions completed, API endpoints responding, not "the platform is up."
- The target : typically a percentage (99.9 %, 99.95 %, 99.99 %) for availability, plus latency targets (p95 latency under 500 ms).
- How it's measured : from where, by what tool, over what time window. "Average over the month" is very different from "5-minute rolling window."
- Exclusions : scheduled maintenance, force majeure, customer-caused outages.
- Remedies : what happens when the target is missed: service credits, fee reduction, escalation rights.
Without all five, the SLA is decorative.
The "three nines" trap
Marketing pages love 99.9 % uptime because it sounds reassuring. In practice, three nines allow:
- 8 hours 45 minutes of downtime per year.
- 43 minutes per month.
- 1.4 minutes per day.
99.99 % ("four nines") cuts that to 52 minutes a year, which usually requires multi-region failover, hot standbys and a paging team. The cost difference between three and four nines can be 10× the engineering investment, so the right question is not "how many nines do we want?" but "what's the cost of an hour of downtime to the business, and what nines does that justify?"
Response time vs. resolution time
Two SLA metrics that get confused:
- Response time : how long until a human acknowledges your incident ticket.
- Resolution time : how long until the incident is fixed.
Most providers commit hard on response, soft on resolution. That's reasonable: resolution depends on the nature of the bug. But it means a "15-minute SLA" usually means "you'll hear from us within 15 minutes," not "we'll have it fixed in 15 minutes." Read the contract.
What Hidora SLAs commit to
Our SLA Expert offering is built around three commitments: response under 15 minutes for P1 incidents, ISO 27001 certified procedures, and on-call coverage that does not depend on permanent access to your infrastructure (useful when regulatory rules forbid that). For full operational coverage including resolution and capacity work, see Managed Services.