What an IT SLA should actually promise
Read the service level agreement you are about to sign and find the number that matters. In most of them, it is not there.
Response time is not resolution time
A four-hour response time means someone will acknowledge your ticket within four hours. It says nothing about when your accounting system will work again. Providers quote response times because they are easy to hit and easy to measure.
What you actually want to know:
- How long until a person is working on it, not just aware of it
- How long until service is restored, by severity
- What happens when that target is missed
Severity has to be defined by impact
A good agreement grades incidents by what has stopped, not by how loudly someone complained. Three or four levels are enough:
- Critical — the business cannot trade. Everyone stops until it is fixed.
- High — a department or a core system is down.
- Normal — one person is blocked, or something is degraded.
- Low — a request, not a fault.
Vague grading is how a genuine outage ends up in the same queue as a mouse that needs replacing.
The clauses worth arguing over
- Coverage hours. If you trade on Saturdays and cover ends on Friday, the number is decorative.
- Escalation. Who is called, at what point, and by whom.
- Reporting. Monthly figures on targets met and missed, published without you asking.
- Exit. Your documentation, credentials and configuration handed over on request.
That last one is the real test. A provider confident in their work has no reason to make leaving difficult.
See how we structure IT support agreements.