SLA · Service Level Agreement
What is an SLA in customer service?
An SLA sets out how fast you respond and how often you meet that. The agreement is only worth something when the timer measures what your customer experiences.
In short
An SLA, in full Service Level Agreement, is the agreement on how quickly you respond to and resolve a customer question, and in what percentage of cases you meet that time. A complete SLA therefore always has three parts: a time, a percentage and the office hours during which the clock runs.
Calculation
SLA = response time + attainment percentage + office hours
"Within 4 hours" is not an SLA. "90 percent of questions within 4 working hours, Monday to Friday, 9 to 5" is one: you can measure it, and a customer knows what to expect.
Why a stricter SLA often works out worse
An SLA of one hour that you meet 70 percent of the time is worse than a four-hour SLA you meet nearly always. In the first case your promise creates an expectation you break three times out of ten, and those three customers send a chaser. That email is new work you caused yourself.
So an SLA measures predictability rather than speed. Choose the time you still meet on your busiest day, not the time that's possible on a quiet Tuesday.
When should the timer stop?
As soon as you put a question back to the customer, the ball is with them. If the timer keeps running, you're measuring how fast your customer replies, not how fast you work. On questions that need an order number or a photo, that quickly adds up to days.
So pause the timer on a wait-for-customer status and resume it when the customer replies. The same goes for outside office hours: a question arriving at 11pm when you open at 9am shouldn't count as ten hours of waiting.
An SLA per type of question
One response time for everything means a complaint about a damaged delivery gets the same urgency as a question about opening hours. So work with a short list: for example two hours for complaints and payment issues, eight working hours for the rest, and a separate agreement for B2B customers with a contract.
Keep the number of categories small. More than about four becomes unworkable, because someone has to get every question into the right category before the clock runs correctly.
Where it goes wrong
Only measuring the first response
An automatic acknowledgement within a minute meets any response-time SLA, while the customer still knows nothing. So alongside the first substantive reply, also measure the time until the question is genuinely handled. Those two numbers together tell the story.
Letting the clock run at the weekend
If you're closed on Saturday, the timer should stop. If it doesn't, your Monday morning is structurally red and nobody steers on the number any more because everyone knows it isn't right.
Promising an SLA that isn't written anywhere
If your response time is on your contact page, it's a promise to every visitor. Check whether that text still matches what you achieve today; an outdated promise on your website produces exactly the disappointed customers your service suffers most from.
Writing an SLA you actually meet
First measure what you do now
Take last month's response time and look at the slowest ten percent, not the average. That's where your real promise sits.
Pick the time you meet on your busiest day
Not the time possible on a quiet Tuesday. A promise you break at peak produces chaser emails exactly when you can least handle them.
Attach a percentage
Ninety percent within four working hours is measurable. Within four hours with no percentage is a wish.
Define office hours and pause rules
When does the clock run, and when does it stop. Without those two you measure waiting time that isn't yours.
After a month, check it holds
If you meet it easily it's too loose and promises too little. If you just miss it, change the work rather than the promise.
What to do this week
Look up which response time you promise on your website. Put it next to the percentage you actually achieved last month. If those two differ, adjust the promise to what you deliver, even if that's a longer time. An honest eight hours beats a broken two hours.
What this looks like in a system
In Cuego an SLA hangs on the ticket rather than the mailbox, with office hours and public holidays built in. The timer pauses automatically when a ticket is set to wait-for-customer and resumes on a reply. Approaching breaches become visible in the queue before they go red, and an escalation can drop a task or an alert with a colleague.
Ticket systemRelated terms
Backlog
A backlog is the pile of open questions. The number says little; the age of the oldest says everything.
Read onFirst Response Time
First Response Time measures how long a customer waits for your first substantive answer. Automatic acknowledgements make the number look good without the customer knowing anything.
Read onQueue
A queue is the list of conversations waiting on someone. How you sort that list determines which customers you systematically help too late.
Read on