SLAs

What an SLA is here

An SLA policy is a promise about time, written down so the system can check it for you. Each policy answers three questions: which conversations does this apply to, what are we measuring, and what is the limit.

You find them under Settings, then SLAs. A policy runs on a schedule, records a breach when the limit is crossed, and builds a daily attainment figure you can look at over time.

What you can measure

Each policy measures exactly one metric.

• First Response Time ⚡ How quickly the team replies to a new customer message. The default threshold is 30 minutes.
• Time to Resolution 🏁 How long the whole case takes from arrival to close. The default threshold is 24 hours.

If you want to hold both, create two policies. That is normal and it is the intended shape.

What counts as a first response

This is worth being precise about, because it is where most disagreements about a number come from. A valid first response is an outbound message that is genuinely a reply.

• Internal notes do not count. Talking to your colleagues is not answering the customer.
• Auto-generated messages do not count. An automatic acknowledgment is not a response, and letting it count would make every SLA trivially green.
• Failed deliveries do not count. A reply that bounced never reached anyone.
• AI replies do count. From the customer's side an answer is an answer. If the AI handled it in forty seconds, you hit your target.

Scoping a policy

The Scope section decides which conversations a policy watches. You add filters on four dimensions:

• Channel 📨 Email, chat, phone and the rest.
• Market 🌍 One country or brand. See Markets.
• Priority 🔺 The urgency the conversation was given.
• Tag 🏷 Anything you tag by hand or automatically. See Tags.

Within a single filter, any of the selected values will match. Across different filters, all of them must match. So "Channel is chat" plus "Market is Sweden" means Swedish chat only, while picking both chat and email in one filter means either.

Leave the scope empty and the policy applies to every conversation. That is a perfectly good first policy.

The business hours pause

Turn on Pause deadline outside business hours and the clock stops when you are closed. A message that arrives at 17:30 on a Friday does not quietly burn through your two-hour target over the weekend.

This needs business hours to exist first, under Settings → Account → Business Hours. If a market has its own hours, those are used for conversations in that market automatically. See Business Hours.

Most teams want this on. The exception is a promise you have actually made in public, like "we reply within 24 hours, always". If the promise ignores your opening hours, the measurement should too. 

What is not measured

Some traffic is excluded on purpose, so your attainment reflects real customer work:

• Playground and demo conversations.
• Test cases.
• Conversations imported or backfilled from a previous helpdesk, since their timestamps describe someone else's response times.

A policy also only measures conversations from the point it was created onwards. Creating a policy today does not retroactively fail last month.

Where breaches show up

Open a policy and you get three things:

• Overview with the metric, the threshold in plain language, and whether the business hours pause is on.
• Attainment trend, the daily percentage over the last 30 days, plus an overall figure.
• Recent breaches, the individual conversations where the target was missed.

There is also a Scan now button if you have just changed something and do not want to wait for the next scheduled check.

Note: Breaches are recorded and shown here. They do not currently send an email to anyone. Treat the policy page as something you look at, rather than something that will chase you.

Choosing a threshold

Start from what you do today, not from what you wish you did. Set the threshold slightly tighter than your current typical response time. A target you miss half the time produces a page full of red that everyone learns to ignore, and an ignored SLA is worse than no SLA.

Tighten it later. Moving a target from four hours to two once you are consistently hitting four is a good day. Starting at two and never hitting it teaches your team that the number is decoration.

Scope before you loosen. If one channel is dragging the average down, the answer is usually a separate policy for that channel with its own realistic target, not a weaker target for everything.

Good habits

Start with one policy and no scope. First response time across everything. Live with it for a couple of weeks and see what it tells you before you build a structure.

Add scope when you find a real difference. A market with two agents and a market with twelve do not belong under the same target. Neither do chat and email.

Read the breaches, not only the percentage. The percentage tells you how often. The breach list tells you which kind of case keeps running over, and that is the one you can fix.

Deactivate rather than delete. A policy has an Active toggle. Turning it off keeps its history, which you will want if you ever turn it back on.