Customer service SLAs explained: what to promise and how to keep it

"We'll get back to you as soon as possible" is not a promise. It's a shrug. Customers read it as "sometime", and your team reads it as "when we get to it". Neither side knows whether things are going well.

A service level agreement turns that into something concrete: we reply to emails within four business hours, chats within two minutes, urgent payment problems within one hour. Customers know what to expect. Your team knows what good looks like. And you find out quickly when something is slipping.

This guide explains what an SLA is, which targets to set, how business hours change the maths, and how a small team keeps its promises. Mid-August is a good time to set them, while summer holidays show you what happens to response times when a few people are away. That gives you a couple of months to adjust before the busy end of the year.

Why "as soon as possible" isn't enough

Without clear targets, the same problems show up in almost every team:

• Oldest tickets wait longest. Agents pick the easy, recent ones and the hard ones sit.
• Urgent issues queue behind trivial ones. A double charge waits behind a question about gift wrapping.
• Nobody notices until customers complain. Slow weeks are discovered in reviews, not in reports.
• Every channel feels the same. A chat customer waits as long as an email customer, and leaves.
• Promises don't match reality. The website says "we reply within an hour" and the team works nine to five.

An SLA fixes this by defining what "fast enough" means, and making it visible before it's missed.

The three times that matter

Measure	What it covers	Why it matters
First reply time	From the customer's first message to your first real reply	It's the wait customers feel most
Next reply time	From each customer reply to your next answer	It keeps conversations moving after the first reply
Resolution time	From the first message to the issue being solved	It shows whether problems actually get fixed

Auto-replies don't count as a first reply. An AI answer that actually solves the question does, as long as the customer didn't need to come back.

Step 1: Set targets per channel

Customers expect different speeds on different channels. Start with targets like these and adjust to your team:

Channel	First reply target	Resolution target	Notes
Live chat	Under 2 minutes while online	Same conversation, where possible	AI agent covers out of hours
Social messages	Within 1–2 hours in business hours	Same day	Some platforms limit how long you can reply, so don't let messages sit
Email and contact form	Within 4–8 business hours	1–2 business days	Auto-reply with the expected time
Phone or callback	Answer or call back within 1 business hour	Same day	Only if you offer it

Pick targets you can hit nine times out of ten in a normal week. Then tighten them. A missed target every Monday trains everyone to ignore it.

Step 2: Add priorities

Not every message is equally urgent. Three levels are enough for most teams:

• Urgent: payment problems, a service that's down, a customer who can't use what they paid for. First reply within an hour.
• Normal: most questions, orders, returns, how-to questions. Your standard target.
• Low: feedback, feature requests, general questions with no deadline. A longer target is fine.

Set priority automatically where you can, from keywords ("charged twice", "can't log in"), the channel, the form the customer used, or the customer's plan.

Step 3: Decide on business hours

SLAs usually run in business hours, so a ticket that arrives at 7pm on Friday starts counting at 9am on Monday. That's fair, but only if customers know your hours.

• Publish your hours on the contact page and in the chat widget.
• Set an auto-reply outside hours that says when you'll answer and links to the most useful help articles.
• Let an AI agent answer common questions out of hours, so many customers don't need to wait at all.
• Plan for holidays. Add them to your schedule, or targets will "break" on days nobody is working.

Step 4: Get warned before a breach

An SLA you only check at the end of the month is a report, not a tool. Make it part of the daily work:

• Sort the queue by time left, not by arrival time.
• Show how long each ticket has left before its target.
• Alert the team when a ticket is close to its target, and escalate to a lead if it passes.
• Review breaches weekly and look for the cause: staffing, a confusing policy, a missing help article, a slow third party.

Step 5: Keep SLAs realistic as volume changes

• Plan for peaks. Sales, launches and holidays bring more messages. Either add hours or temporarily relax low-priority targets, and say so.
• Use automation for first replies that help. An instant AI answer that solves the question beats a fast "we've received your message".
• Watch next reply time. Teams often hit first-reply targets and then let conversations sit.
• Don't game the numbers. Replying "looking into it" to stop the clock satisfies the report, not the customer.

SLAs for business customers

If you sell to companies, some will ask for a formal SLA in their contract. Keep it simple and honest:

• define support hours, channels and priority levels in plain language;
• promise response times, not resolution times, unless you control the whole fix;
• agree what happens if you miss a target (a review, a service credit) before it happens;
• report on it regularly, and use the same help desk data your team works from.

Common mistakes

• One target for every channel. Chat and email customers expect different speeds.
• Calendar hours without staff to match. Your team isn't working at 3am, so either measure business hours or cover nights with AI.
• Only measuring first reply. The second wait is where many customers give up.
• Setting targets you can't keep. Credibility with customers and the team goes fast.
• No owner. Someone should look at the breach list every day.

How to set up your first SLA this week

1. Day 1: pull last month's median first reply and resolution times per channel. That's your starting point.
2. Day 2: set targets slightly better than today's median, per channel, in business hours.
3. Day 3: define three priorities and the rules that set them automatically.
4. Day 4: configure the targets and breach warnings in your help desk, and publish your support hours.
5. Day 5: switch on an AI agent or a helpful auto-reply for out of hours, then review breaches every Friday.

If you want customers covered around the clock without a night shift, read how to offer 24/7 support with a small team. And for the rest of the numbers worth tracking, see the customer service metrics that matter.