SLA design for small support teams: four tiers, business hours, and one honest clock

Service-level agreements have an enterprise reputation: long contracts, penalty clauses, a dashboard nobody looks at. For a support team of five handling refunds, reschedules and "where is my order", that version is irrelevant. But the underlying idea — a promise about how fast you respond, and a clock that shows whether you kept it — is the single most useful discipline a small team can adopt.

Here is how to set it up so that it helps rather than nags.

Start with priority, not with time

An SLA is a promise per priority level. So the first decision is what "priority" means in your business, and the honest answer is usually four tiers:

• Urgent — the customer cannot use what they paid for, right now. A boarding pass that will not scan at the gate. A payment taken twice.
• High — money or a commitment is at stake today. A refund that has not arrived, a reschedule for tomorrow's sailing.
• Medium — a question with a deadline this week. An invoice correction, a group enquiry.
• Low — everything else. Feedback, general questions, "do you have parking".

Resist the urge to add a fifth tier. Four is enough to route by, and few enough that agents can assign them without a flowchart.

Set targets you can meet on a bad Tuesday

Now attach a first-response target and a resolution target to each tier. The mistake is to set them aspirationally. Set them for the week when two people are on leave.

A reasonable starting point for a small team on business hours:

Tier	First response	Resolution
Urgent	15 minutes	2 hours
High	1 hour	8 business hours
Medium	4 hours	2 business days
Low	1 business day	5 business days

You will tighten these later. Tightening a target you have been meeting is a celebration. Loosening one you have been missing is a quiet admission — and you will have trained customers to expect the old number.

Run the clock against business hours

An SLA clock that runs through the night breaches everything by morning and teaches the team to ignore red. Define business hours per desk — including the Saturday half-day if you have one — and pause the clock outside them. Pause it too when you are waiting on the customer; a ticket that has been "waiting on customer" for three days is not a breach, it is a follow-up.

The clock should also start on the customer's message, not on the agent's first look. If a WhatsApp message sits unread for forty minutes, that is forty minutes of response time, and the report should say so.

Escalate on the clock, not on complaints

The point of the clock is that it acts before a customer has to. A simple escalation ladder:

1. At 75% of the response target — the ticket goes amber for the owner.
2. At breach — the ticket goes red and the team lead is notified.
3. At 150% of the target — it moves to the team lead's own queue.
4. At 2× the target — the manager sees it on their dashboard by name.

Four levels, all automatic. Nobody has to notice a ticket is late; the system does, and it tells the right person at the right moment. Most teams find that level 3 almost never triggers once level 1 is visible — the amber does the work.

What the report should show

A weekly SLA report for a small team needs exactly three views:

• Compliance by tier — what percentage of urgent, high, medium and low tickets met their targets. Anything under 90% on urgent or high is the week's conversation.
• Compliance by category — refunds versus reschedules versus boarding. This is where you discover that one category is consistently late because it needs someone from finance, and fix the routing.
• Breaches by hour of day — the 4pm dip when the counter gets busy shows up immediately. That is a staffing decision, not a performance one.

Skip per-agent SLA numbers in the first month. Until routing and staffing are stable, they mostly measure who got the worst shift.

Where AI fits, and where it does not

Two things AI does well here: assign the priority on arrival (so the clock starts on the right tier without an agent reading it first), and draft the reply so the first response is fast. Two things it should not do: change a priority a human set, or close a ticket to stop a clock. The SLA is a promise to the customer; the AI helps keep it, it does not get to redefine it.

Request a demo of Service CRM →

The one-week setup

Four tiers, one table of targets, business hours per desk, a four-step ladder, three report views. All of that is configuration in any serious helpdesk and should take a week including the arguing. After that, the clock runs, and for the first time the team can say "we respond to urgent tickets within 15 minutes" and have a number behind it — which is worth more to a customer than any feature on the product page.