SLAs for MSPs: A Practical Introduction
An SLA — Service Level Agreement — is simply a written commitment to how fast you'll respond to and fix a client's IT problem, and it's one of the few pieces of an MSP relationship a client can actually measure. Everything else about managed services is a little intangible: uptime they don't see, patching they don't notice, security work that only proves itself when nothing goes wrong. An SLA gives them a number to hold you to — and gives you a number to hold yourself to.
For MSPs, SLAs matter for a second reason that has nothing to do with the client: they force discipline internally. A response target with a clock running behind it changes how a Service Desk actually triages, escalates and reports — in a way a vague promise to "get to it soon" never does.
SLA or no SLA: what actually changes
Not every client is on an SLA, and that's a legitimate business choice, not an oversight. Plenty of MSPs run a mix: clients on a managed services agreement with a formal SLA baked in, and pay-as-you-go or ad hoc clients who are served best-effort, in the order tickets land.
The practical differences show up in three places:
- The clock. An SLA-covered ticket has a response-due and resolution-due time attached the moment it's logged, calculated from its priority. A best-effort ticket has no due time — it's just worked in queue order (or by whoever's free).
- What "on time" means. With an SLA, a ticket can be breached — a visible, reportable event. Without one, there's no such thing as a breach, because there was never a target to miss.
- Status is not the same as the clock. It's worth being precise here, because clients often conflate the two: ticket status (Open, In Progress, Waiting on Client, Resolved, Closed) exists whether or not an SLA applies — it tracks where the work stands. The SLA clock is a separate thing layered on top of status: it only runs while the ticket is in a status that counts as "the MSP's turn to act," and it pauses the moment the ball is back in the client's court (waiting on a reply, waiting on a part, waiting on access). A ticket can sit in "Waiting on Client" for three days without ever breaching, because the clock paused the instant it moved there.
That distinction — status describes the ticket, the SLA clock describes your obligation — is the single most common source of confusion when a client asks "why hasn't this breached yet?" or "why does the clock say something different to the ticket status?"
The anatomy of an SLA
Strip away the jargon and every SLA is built from the same four parts.
Response time — how long before a human acknowledges the ticket. This is almost always the client-facing headline number, because it's the first proof someone is actually there.
Resolution time — how long before the issue is actually fixed. This is harder to promise (some problems genuinely take longer than others) so it's usually set looser than response time, and often scoped to "workaround in place" rather than "root cause fixed."
Priority tiers — not every ticket deserves the same urgency, so targets are set per severity. A typical four-tier scheme:
| Priority | Example | Response target | Resolution target | Coverage |
|---|---|---|---|---|
| P1 — Critical | Server down, whole office offline | 15–30 min | 2–4 hrs | 24/7 |
| P2 — High | One department can't work | 1 hr | 8 hrs (same business day) | Business hours |
| P3 — Medium | Single user impacted, has a workaround | 4 hrs | 1–2 business days | Business hours |
| P4 — Low | Cosmetic issue, how-to question | 1 business day | 3–5 business days | Business hours |
Business hours — the calendar the clock actually runs against. A P3 logged at 5pm Friday with an 8-hour resolution target doesn't breach over the weekend; the clock simply stops ticking outside the defined hours and days, then resumes Monday morning. Critical issues are the usual exception, running 24/7 regardless.
Put together, those four parts are what makes an SLA a clock, not just a sentence in a contract: priority decides which target applies, business hours decide when the clock is allowed to run, and the ticket's status decides whether it's currently running or paused.
How clients really see SLAs
Most clients never read the SLA document after the day they signed it. What they actually experience, and judge you on, is narrower than the whole agreement:
- The first reply. Clients notice response time far more than resolution time. An acknowledgement within the target — even just "we've got this, looking at it now" — does most of the reassurance work, regardless of how long the fix itself takes.
- Being told, not left to ask. A client who gets a status update before they have to chase one reads that as the SLA working. A client who has to email twice to ask "any update?" reads the exact same resolution time as the SLA failing — even inside target.
- Consistency over speed. A client who gets a 2-hour fix one week and a 2-day fix the next for a similar issue, with no explanation of the difference, loses more trust than one who consistently waits a day but always knows what to expect.
- The invisible pause. Clients rarely notice that the clock pauses while waiting on them — until a report shows the ticket was "open" for three days and they assume that's three days of inaction. This is one of the most common support disputes, and it's almost always a communication gap, not an SLA failure: it's worth explicitly telling the client when the ball is in their court.
- The report, not the promise. Clients with a formal SLA increasingly expect to see the numbers — a monthly or quarterly compliance report showing percentage met per tier. That report is often what justifies the contract renewal far more than the original SLA clause ever did.
The upshot: an SLA is less a legal shield and more a communication contract. MSPs that treat it that way — proactive updates, clear status, visible reporting — get more credit for the same underlying response times than those that just quietly meet the number.
Setting up SLAs: a practical walkthrough
Setting up SLAs well is less about picking impressive numbers and more about picking numbers you can actually hit, every time, without heroics. A rough order of operations:
- Define priority, not just urgency. Write down what makes something a P1 versus a P3 in concrete terms ("P1 = more than 5 users down, or any revenue-generating system down"), not gut feel. Engineers and the Triage Queue need a rule, not a vibe — otherwise every client thinks their issue is P1.
- Set business hours first, targets second. Decide your coverage windows (standard business hours vs. 24/7 for critical) before you set the time targets — the hours a target runs against change what's actually achievable.
- Set response and resolution targets per tier. Start conservative. A target you consistently beat builds trust; a target you miss even occasionally erodes it far more than a slightly longer target ever would.
- Decide where the policy lives, and in what order it resolves. Most MSPs end up needing SLA policy at more than one level: a sensible default for the whole business, overridable per client (because a managed client's contract may promise tighter targets than a block-hours client's), and sometimes per board or queue (a security-incidents board often needs stricter targets than a general-requests board). When more than one policy could apply to the same ticket, decide the resolution order in advance — typically the most specific policy wins: board-level, then the client's own agreement, then the tenant-wide default.
- Define exactly what pauses the clock. "Waiting on Client," "Waiting on Third Party," "Waiting on Parts" should all pause it. Be explicit about this with clients up front — it's the single most-asked-about mechanic once SLAs go live.
- Build in breach visibility before you need it. A ticket approaching its target should flag internally well before it actually breaches, so engineers can act rather than react. Decide who gets the alert and when (e.g. at 75% of the target elapsed).
- Report it. Whatever the targets are, track percentage met per priority tier and review it monthly. This is both an internal health check and, for clients who want it, the proof the SLA is real.
- Review the targets periodically — don't set and forget. A target that looked fine at 50 tickets a month can quietly become unworkable at 300. Revisit targets alongside headcount and ticket volume, not just once a year on autopilot.
The common failure mode isn't picking the wrong numbers — it's skipping steps 1 and 5. Without a clear priority definition, every ticket argues its way into the fastest tier. Without a clear pause policy, every delayed ticket looks like a breach that wasn't one.
The takeaway
An SLA isn't a number you publish once and forget — it's a running clock, tied to priority and business hours, paused and resumed by ticket status, and reported on to prove it's real. Get the mechanics right (clear priority rules, an explicit pause policy, visible breach warnings) and the client-facing side mostly takes care of itself: clients forgive a lot when they're told what's happening and rarely forgive silence, whatever the actual numbers say.
If you're setting this up for the first time, the practical place to start is priority definitions and pause rules — everything else in an SLA policy builds on those two decisions.
How this maps to MSP Pulse
If you run your desk on MSP Pulse, the mechanics above line up with the product like this:
- Policy levels and order. SLA policy can live at three levels: a tenant-wide default, a billing agreement, and a service desk board. When more than one could apply, the most specific wins: board, then the client's agreement, then the default. You manage the policies in Settings → SLA.
- Best-effort clients. A billing agreement can be marked Best-effort / SLA exempt, so its tickets carry no deadlines and can never breach. A board can likewise have no SLA at all.
- Priority and business hours. Targets are set per priority. P1 runs 24/7 by default, while P2 to P4 run against the business hours and business days you define in Settings → SLA.
- The clock follows status. The clock pauses automatically when a ticket moves to Waiting or Awaiting Information, and resumes when it goes back to Assigned or In Progress. Each ticket shows its response and resolution deadlines, and they turn red when missed.
- Frozen at the point of setting. A ticket's SLA is stamped when it's raised, and changing the billing agreement afterwards doesn't move its existing deadlines.
Two items from the walkthrough, an early warning as a ticket nears its target and a compliance-percentage report per tier, are not part of the SLA engine today, so for now treat steps 6 and 7 as practices to run alongside it.