SLA policies
A service level agreement turns "we'll get back to you soon" into a number you have committed to and can prove. For enterprise customers it is often contractual. For everyone else it is the difference between knowing you are slow and finding out from a complaint.
What an SLA is for
An SLA policy does three things:
- Sets an expectation you and the customer both understand.
- Warns your team before a target is missed, while there is still time to act.
- Records what happened, so you can see whether you are meeting your commitments and where you are not.
The third is the one that changes behaviour. Most teams discover, on first measurement, that their average is fine and their worst cases are terrible — and it is the worst cases that lose customers.
The three targets
- Name
First response time- Type
- how long until a human replies
- Description
From when the customer writes to when an agent responds for the first time. The number customers actually experience, and the one worth being strictest about.
- Name
Next response time- Type
- how long between replies
- Description
Applied to every subsequent reply, not just the first. This is what stops a conversation being answered promptly and then going quiet for three days — the most common way a good response time hides a bad experience.
- Name
Resolution time- Type
- how long until it is done
- Description
From opening to resolved. Realistic targets here are much longer than people expect, because resolution often depends on somebody outside the support team.
You do not have to set all three. Many teams start with first response time only, add next response time once that is stable, and treat resolution time as a reporting measure rather than a commitment.
Business hours
A policy can be measured only during business hours, which is almost always what you want.
Without it, a message arriving at 11pm on Friday is eating your response target all weekend, and your numbers describe your opening hours rather than your performance. With it, the clock pauses overnight and resumes when you open.
Business hours are configured per channel, so a policy applied to a 24/7 channel and one applied to an office-hours channel behave correctly without needing separate policies.
Applying a policy
Create policies under Settings → SLA, then apply them:
- To a channel, so everything arriving there is measured the same way.
- Through a workflow, which is where the real value is — apply a stricter policy based on the conversation, not just where it came from.
That second option is what makes SLAs useful commercially. A workflow that applies your two-hour policy when the contact's plan attribute is enterprise, and your standard policy otherwise, lets you sell a genuine service level rather than a promise.
You can run several policies at once: a standard one for everyone, a tighter one for priority customers, a tighter one still for anything marked urgent.
What happens on a breach
As a conversation approaches its target it is flagged in the inbox, so an agent can see what is about to go wrong while they can still prevent it. That warning is the point — a breach notification after the fact is a record, not a tool.
Once a target is missed the conversation is marked as breached, and the breach is recorded against the conversation, the agent and the team for reporting.
Pair this with a workflow for anything that must not be missed: when a high-priority conversation is approaching breach, notify the team lead by email.
Setting targets you can meet
The most common mistake is setting an aspirational target, missing it constantly, and quietly stopping looking at the report. A breached SLA everyone ignores is worse than none.
- Measure before you commit. Run for two weeks with no policy and look at your actual times in reports. Set the first target somewhere you already hit most of the time.
- Set it against the worst case, not the average. If your average first response is twenty minutes but a tenth of conversations wait six hours, your customers' experience is the six hours.
- Different channels deserve different targets. Live chat is minutes. Email is hours. WhatsApp sits in between, and customers expect it to be closer to chat than to email.
- Tighten gradually. Move the target when you are comfortably beating it, not before.
- Only promise externally what you meet internally, with room to spare. An SLA in a contract is a liability; an SLA on a dashboard is a tool.
Reporting
The SLA report shows how many conversations met each target and which breached, sliced by agent, team, channel and period.
Read it for patterns rather than for individual failures. Breaches clustered on one weekday mean a staffing gap. Breaches clustered on one channel mean routing or capacity. Breaches spread evenly across everyone mean the target is wrong, not the team.