SLA policy starter

About this policy

This is a working service-level agreement, not a legal exhibit. Fill in the bracketed values, delete what doesn't apply, and publish only the targets you can actually staff. An SLA the team misses every week erodes trust faster than having no SLA at all.

Replace [Company], the timezone, and the hours with your own before you ship it.

Priority definitions

Set the targets after you agree what each priority means. Vague definitions are where SLAs quietly break — everything becomes a P1 the moment a customer is annoyed.

One rule that saves arguments: priority is set by impact, not by who is shouting. Anyone can raise a priority; a downgrade needs a reason noted on the ticket.

First-response targets

First response means the first human, substantive reply — not an autoresponder. An automated "we got your ticket" message does not stop this clock.

Resolution targets

Resolution means the customer's problem is actually fixed and they've confirmed it, or you've delivered everything within your control. A ticket that's closed and then reopened did not meet this target.

Update cadence for open tickets

The most common complaint isn't slow fixes — it's silence. Commit to a next-update rhythm so nothing goes dark, even when a fix is slow.

Business hours

Define these precisely — every target above is measured against them, not the wall clock.

• Standard hours: [Mon–Fri, 09:00–17:00], [Timezone — e.g. CET / America/New_York]
• Weekends & public holidays: [closed / P1 only]. Published holiday calendar: [link]
• After-hours coverage: [P1 only, via on-call] — [pager / phone rota]. Tickets raised after hours start their clock at the next business-hour open.
• How business-hour math works: a 2-business-hour target on a ticket that arrives at 16:30 is due at 10:30 the next morning, not 18:30 that evening.

State your timezone explicitly. "2 hours" means nothing to a customer in another region without it.

What pauses the clock

Be strict here. Every status that stops a timer is a place the timer will get stopped to make a number look good.

Pauses the resolution clock:

• Awaiting customer — you've asked a specific question and genuinely can't proceed without their reply
• Scheduled with customer — a call or maintenance window is booked for a future time
• Blocked by a named third party — waiting on a vendor or upstream provider, logged with a reference

Does not pause any clock:

• An automated acknowledgement or "ticket received" email
• Flipping to "pending" the instant a ticket lands, before anyone has read it
• Internal handoffs, escalations, or "waiting on engineering" — that time is yours, not the customer's
• "Awaiting customer" when your last message didn't actually ask them anything

Guardrail: the first-response clock never pauses. It runs from arrival to the first human reply, full stop.

Measuring the right thing

A green first-response number sitting on top of an unresolved ticket is a false positive. Watch the metrics the customer actually feels:

• Time to resolution as the headline number, paired with reopen rate. A ticket reopened twice didn't resolve in 4 hours — it resolved in 3 days and annoyed someone on the way.
• % of tickets meeting the resolution SLA, reported next to first-response — never first-response on its own.
• CSAT on resolved tickets, so speed that leaves people unhappy still shows up.
• Break every figure down by priority, so being fast on P4s can't hide misses on P1s.

And avoid the trap in reverse: don't reward a metric a macro or a bot can satisfy without ever reading the ticket.

Adapt the targets to real capacity — read this before publishing

The numbers above are a starting point, not a target to copy. A promise you can't staff is a forecast of missed SLAs and burned-out agents, not a service level.

Pressure-test before you commit to any target:

• Take last quarter's ticket volume by priority and channel — could you have hit these targets with the headcount you actually had, not your ideal roster?
• Model your worst realistic day (a Monday spike, someone off sick), not the average. SLAs are judged on bad days.
• If you'd miss a target in a normal week, loosen the target — don't quietly hope. Publish something you can hit roughly 95% of the time.
• Confirm after-hours P1 coverage is a real rota with named people, not an aspiration.
• Re-check every quarter as volume and headcount move. An SLA is a living commitment, not a launch artifact.

Better to publish a slower target you consistently meet than a fast one you miss a third of the time. Customers forgive a clearly stated wait; they don't forgive a broken promise.

Publish & maintain checklist

• Priority definitions agreed with the team and written down
• Targets pressure-tested against real historical volume and staffing
• Business hours, timezone, and holiday calendar stated explicitly
• Clock-pause rules configured in the help desk to match this policy
• Reporting shows resolution time and reopen rate, not first-response alone
• An owner is assigned and a quarterly review date is on the calendar