The five parts of an SLA you can meet
1. Severity levels defined with examples, not adjectives
“Critical,” “high” and “medium” mean nothing on their own. What works is defining each level and giving it an example from the customer’s own operation:
| Severity | Definition | Example |
|---|---|---|
| Critical | Service down with no workaround | The walk-in cooler is not cooling |
| High | Service degraded, affecting production | One of the two production lines is down |
| Medium | Affects work, but there is a workaround | One of four units is out of service |
| Low | No operational impact | A question or a feature request |
With concrete examples, the argument over whether something is critical takes thirty seconds.
2. Two separate clocks
First response time and resolution time. Mixing them up is the most common cause of disputes. If the SLA says “4 hours” without saying which, the customer will read resolution and your team will read response.
3. Coverage hours and holidays
“4 hours” does not mean the same thing 24/7 as it does during business hours. Decide whether the clock runs in calendar hours or business hours. And which holiday calendar applies when the customer is in another state or country.
If you change your business hours, write down when. The earlier history will be read with different rules.
4. When the clock stops
When the ball is in the customer’s court: waiting for information, a maintenance window or approval of a quote. If it is not written into the contract, it does not exist, and all that time counts against you.
The honest flip side: pauses get abused too. If a large share of your tickets sits in “waiting on customer,” that label is hiding stalled work.
5. Who provides the data
It is the part that almost never gets negotiated and the one that causes the most trouble. The sensible setup is that the data comes from the system automatically and the customer can check it whenever they want. The report has to be the same for both sides.
How to really measure it
Use the median and the 90th percentile, not the average. An average of 3 hours can hide one ticket in ten taking three days. And those are exactly the ones that cause complaints and put the renewal at risk.
Example
Ten tickets resolved in 1, 1, 2, 2, 2, 3, 3, 4, 5 and 27 hours. The average is 5 hours, and it looks like a good month. The median is 2.5 hours: the typical ticket goes well. But the slowest one took 27 hours. That is the one that needs an explanation, and the one the customer remembers.
Always measure all three: compliance, reopen rate and satisfaction. Compliance on its own can be gamed by closing before resolving. The other two give it away.
The three clauses that cause conflict
- Penalties without a measurement method. If there is a penalty, the calculation must be in the contract, not in a conversation.
- “99.9% availability” without defining what counts as downtime or what is excluded.
- Commitments that depend on third parties. You cannot promise a deadline that depends on a manufacturer shipping you a part. That gets agreed separately.
The report that wins the meeting
The one that shows compliance by severity and how it has moved over the last six months. And each missed ticket, one by one, with its reason.
Showing your misses seems counterproductive, and it is the opposite. A report with the month’s three misses and their explanation builds more trust than one that says 100% and nobody believes.
Before you sign your next SLA, check that your system records when the ticket came in, when it was answered and when it was resolved. If it does not, you are signing something you cannot prove.
Last updated: