SLA management: setting service levels you can actually meet

SLA management is the work of agreeing service level targets, measuring performance against them, and dealing with breaches before the customer raises them. The agreement itself is the easy part — most disputes come from what the clock counts. If your target says four hours and the customer counts from when they emailed while you count from when the ticket was triaged, both sides can report honestly and disagree completely.

What SLA management has to define

• Priority levels with concrete examples, so classification is not argued case by case.
• Response target and resolution target per priority, stated separately.
• Service hours: business hours or 24/7, in which timezone, and which public holidays.
• When the clock starts, and exactly what pauses it — waiting on the customer, third-party vendors, scheduled maintenance.
• Measurement method and the report both sides will look at.
• What happens on breach: escalation path, service credits, and who is notified.

Setting targets you can meet

Derive targets from your actual performance distribution rather than from what sounds impressive. Look at the last six months by priority, take the 90th percentile, and set the target there or slightly better. A target met 95% of the time creates trust and a manageable exception process; a target met 60% of the time creates a monthly argument and a credit negotiation. Where a customer demands tighter targets, price the cost of meeting them — staffing outside business hours is the usual driver.

Pausing the clock while waiting for the customer is standard and fair, but it is also the most abused mechanism in SLA management. Log every pause with a timestamp and the request that triggered it, or breach reports become unarguable only because they are unverifiable.

Targets mean nothing until the conversations live somewhere countable. Customer support software supplies the queue, the states and the response measurement that an agreed target is reported against.

Running the process

1. Classify every ticket on arrival, and audit a sample of classifications monthly — priority inflation and deflation both distort the numbers.
2. Alert before breach, not after: a warning at 75% of the target is what prevents the breach.
3. Review breaches weekly with causes, not just counts.
4. Report monthly in the same format every time, including the exclusions and pauses.
5. Revisit targets annually against actual volumes and staffing.

Ettex Records suits the tracking layer for teams without a dedicated service desk: one row per ticket with priority, received time, response and resolution timestamps, pause periods and breach flag, plus a monthly view by priority. Because pauses are recorded as data rather than as comments, the report can be reproduced when a customer questions it, alongside the underlying service level agreement.