The SLA is the document your client actually experiences. They'll never re-read the contract, but they feel the SLA every time they open a ticket. Here's a service level agreement template that sets expectations you can defend — and survive.
Template, not legal advice
1. What an SLA Actually Does
An SLA does three jobs: it converts "fast" into numbers, it defines what does not count against you, and it states the consequence of a miss so a bad week doesn't become a contract dispute. Notice what's not on that list: marketing. An SLA stuffed with 5-minute responses and 99.999% uptime you can't staff is a liability document, not a sales document.
2. The Template, Section by Section
1. Priority Matrix, Response, and Resolution Targets
| Priority | Definition | Response | Resolution target | Coverage |
|---|---|---|---|---|
| P1 — Critical | Business stopped: site down, server down, security incident in progress | 15 min | 4 hours | 24/7 |
| P2 — High | Major function impaired: line-of-business app down for a team, degraded network | 1 hour | 8 business hours | Business hours + on-call |
| P3 — Normal | Single user impaired with a workaround available | 4 business hours | 2 business days | Business hours |
| P4 — Low | Requests, questions, scheduled changes | 1 business day | Scheduled | Business hours |
"Response" means a qualified technician engaged and working — not an auto-acknowledgement. "Resolution target" is a target, not a guarantee; only response carries credits by default.
2. Hours of Coverage
Business hours defined precisely (e.g., 7:00 am – 6:00 pm local, Monday–Friday, excluding listed holidays), the after-hours channel for P1s, and the rate for after-hours work on anything below P1. If you sell 24/7 help desk as an upgrade, this is where the tier splits.
3. Measurement
Response time is measured from ticket creation in your PSA (with phone-reported issues logged within a stated number of minutes) to first technician action recorded. Uptime, if committed, is measured monthly per covered system by your monitoring platform, excluding agreed maintenance windows. Name the system of record — disputes end when the data source is agreed in advance.
4. Exclusions
Misses don't count when caused by: client changes made against written advice, systems the client declined to remediate or retire, third-party outages (ISP, SaaS vendor, power), force majeure, or issues on assets outside the covered environment. This list does more to protect your margins than any other section.
5. Service Credits
Per verified P1 response miss: 5% of that month's monthly fee, capped at 25% of the monthly fee across all misses in a month. Credits are claimed in writing within 30 days, applied to the next invoice, and are the exclusive remedy for SLA misses. Chronic failure (3 consecutive months at the cap) opens the termination-for-cause path in the MSA — that's the honest escape valve that makes the cap fair.
6. Review Cadence
The SLA is reviewed at the quarterly business review and may be amended by mutual written consent without re-executing the MSA. This is why the SLA lives as a schedule, not contract body text.
3. Setting Targets You Can Hit
Work backward from staffing, not forward from marketing. Pull 90 days of PSA data and find your actual P90 response time per priority — the number you hit 9 times out of 10, including Friday afternoons and the week two techs were out. Commit to that, not your best week. A published 15-minute P1 response only works if someone is genuinely on-call with an escalation chain behind them.
Sell the delta
4. Common SLA Mistakes
- Guaranteeing resolution times on P1s — you can't schedule how long a fix takes; commit to response and communication cadence instead.
- Uncapped credits, which turn one bad month into a free month and an SLA into an insurance policy you underwrote for free.
- No measurement source of record, so every miss becomes an argument about whose clock counts.
- One SLA for every client size — a 10-seat office and a 200-seat firm should not carry identical commitments at identical price points.
- Copying an enterprise SLA with 99.999% uptime language across systems you don't control.
Key Takeaways
Commit to response, target resolution, measure from a named system of record, exclude what you don't control, and cap the credits. Attach it as a schedule to the managed services agreement and present it inside the proposal as proof you run a real operation.