Service level agreement and SLA response times

SLA response time is the reply half of that promise, the longest a customer should wait to hear from you, such as 4 business hours for a bug that blocks them. Resolution time is the other half, how long until the problem is fixed. To check you kept the promise, divide the conversations that met the target by all the ones that met it or missed it. If 126 of last month’s 144 got their first reply in time, your SLA compliance was 87.5%.

When a customer asks for “your SLA”, check which promise they mean. Some want support response times, and some want an uptime commitment, such as the product being up 99.9% of the time, which allows about 43 minutes of downtime in a 30-day month. Atlassian keeps the two apart. Its service level agreement covers uptime, and its support offerings give response times it “will use commercially reasonable efforts to meet”. This entry is about the support kind, and what you can promise without a support team.

On your own, any reply time you promise only holds while you’re at your desk, and customers write at night and at weekends too. Deacon is an AI customer support agent that answers them from your own docs in seconds, at any hour, and passes you the questions only a person can answer. It has a free plan. The helpdesk pages and support policies quoted on this page were read on 10 November 2026.

Response time and resolution time targets

A response target times the wait for a reply, and a resolution target times the wait for the fix. Any honest reply meets a response target, including “I can see the bug and I’m working on it”, so it’s the one you control. A resolution target depends on the problem, and a bug fix can take a week whatever the SLA said.

The first response time entry covers measuring the first wait, with published figures on how fast companies reply, and average resolution time covers the second. A helpdesk can start a clock on two more waits as well, and each tool has its own names for them.

A resolution clock needs a rule for the days you spend waiting on the customer. One that keeps running while they go quiet for a week is a target you’ll miss without doing anything wrong. Zendesk’s total resolution time includes the time a ticket sits pending, waiting on the customer, and its requester wait time leaves that out. Intercom can pause its SLAs while a conversation is snoozed or waiting on the customer. Help Scout counts a conversation as resolved once it’s closed with at least one reply from your team.

Decide too whether an automated reply counts. An autoreply saying the message arrived tells the customer nothing they didn’t know, yet Zendesk’s SLA help says its reply targets are met by a trigger that sends a public autoreply, and autoreplies come only with its Copilot add-on. Intercom’s own support policy lists Fin AI Agent among the ways it gives support around the clock, then sets separate targets for its support team’s first response. If you use an AI agent, write your SLA the same way, with the AI’s answer at any hour and a person’s reply within a stated time when the AI can’t help.

SLA response times two SaaS companies publish

Atlassian’s support offerings give its Standard cloud plan these initial response times, with support 9 hours a day, Monday to Friday.

Intercom’s support policy, in force since 25 June 2024, gives customers on its Essential, Advanced and Expert plans a target first response of 1 hour when “a major feature has stopped working” and 4 hours for anything else. Customers on its older plans get an attempt at a reply within one business day. Both parts of the policy say “We do not promise or guarantee any specific response time”, and Intercom keeps the right to raise or lower a request’s priority.

Both companies define a priority by what the problem does to the customer. Neither gives a resolution time, and both call their times targets.

An example SLA a small SaaS could keep

This is an example for one founder answering support, not a benchmark. It assumes support hours of 9:00 to 17:00, Monday to Friday, in your time zone.

Like both published policies above, it has no resolution column. When the fix is yours to ship, promise updates on urgent problems instead of a time for the fix.

Set each number from your own history. Take your first reply times for the last 8 weeks, in business hours. For each week, sort the waits from shortest to longest and read the one 9 in 10 of the way down, which is the 36th if you had 40 conversations, then take the highest of the 8. If that’s 5 business hours, a 4-hour promise will break, so promise 6.

Keep the priority your call, as Intercom does, so the urgent row stays for real emergencies. Setting it as each message arrives is ticket triage. In Zendesk, a ticket with a blank priority gets no SLA at all, whatever its other conditions, so its SLA help suggests a trigger that sets one when the ticket is created.

Business hours or calendar hours in an SLA

The same reply can keep the promise or break it, depending on whether your clock stops at night and at weekends. A customer writes at 15:00 on a Friday, and you reply at 10:30 on Monday. On a 9:00 to 17:00 weekday clock, that’s 2 hours on Friday and 1 hour 30 minutes on Monday, so a 4-business-hour target is met with 30 minutes to spare. In calendar hours, which count every hour of the week, the customer waited 67 hours 30 minutes, and a 24-hour target was missed by 43 hours 30 minutes.

Most helpdesks make you choose, each in its own place. Zendesk asks for each priority, Help Scout for each target and Freshdesk for each policy. HubSpot’s goals apply at all times or during hours you set, with a time zone. Intercom counts your office hours, and gives the example of a 15-minute first response target on a 9:00 to 18:00 day, where a message at 17:50 is due at 9:05 the next working day.

Be careful with targets written in days. Freshdesk is changing a business-hours target of 1 day from the length of one business day to 24 business hours, in a phased release during 2026. On a 9:00 to 17:00 weekday schedule, that turns a next-day promise into three working days. Freshdesk says targets set in days “must be changed to hours”, so enter a 1-business-day promise as 8 hours.

Put the hours, the days, your time zone and the holidays you close on in the SLA itself. If you’re in London, a customer in Sydney writes during your night, and a business-hours promise is only fair if they can see when your hours are. Measure on the clock your SLA promises, and keep the calendar wait beside it, since that’s what the customer sat through.

How to calculate SLA compliance

Leave out conversations whose clock is still running, since they haven’t met or missed anything yet. Zendesk’s % Achieved SLA tickets works the same way in its metrics reference, dividing achieved tickets by achieved plus breached ones. A ticket counts as achieved there only if it met every target on it, so one late update makes the whole ticket a miss.

Say October brought 150 conversations with a first reply target. At the end of the month 6 were still waiting with time left, so they count in November, when their clocks stop. Of the other 144, 126 got their first reply in time and 18 didn’t, so compliance was 126 ÷ (126 + 18) × 100 = 87.5%.

The month’s 87.5% looks good, but 3 of the 8 urgent customers waited longer than you promised. Work out compliance for each priority, and read every urgent miss.

If you answer from Gmail without a ticketing system, a spreadsheet does the job. Put the time each message arrived in column A and the time of your first reply in column B, as dates with times. This formula in C2 gives the business hours between them on a 9:00 to 17:00 weekday clock, in Google Sheets or Excel. Fill it down the column.

=((NETWORKDAYS(A2,B2)-1)*(TIME(17,0,0)-TIME(9,0,0))+IF(NETWORKDAYS(B2,B2),MEDIAN(MOD(B2,1),TIME(17,0,0),TIME(9,0,0)),TIME(17,0,0))-MEDIAN(NETWORKDAYS(A2,A2)*MOD(A2,1),TIME(17,0,0),TIME(9,0,0)))*24

For the Friday message answered at 10:30 on Monday, it gives 3.5. Change the 17 and the 9 for your own hours. To skip your holidays, list their dates in a range such as H2:H20 and add it to each NETWORKDAYS as a third input, as in NETWORKDAYS(A2,B2,$H$2:$H$20). Then =IF(C2<=4,"Met","Missed") in D2 checks a 4-hour target, and =COUNTIF(D:D,"Met")/(COUNTIF(D:D,"Met")+COUNTIF(D:D,"Missed")) gives your compliance as a fraction to format as a percentage.

Catching a target before it’s missed

Without a helpdesk, you can check the inbox at fixed times instead of waiting for a warning. Three checks a day, at 9:00, 13:00 and 16:30, keep every message inside a 4-business-hour first reply if you answer at each one, because no message waits more than 4 business hours for the next check. A 2-hour urgent target needs something that interrupts you, such as an outage address that rings your phone.

A helpdesk can warn you while a reply would still keep the promise. This is what each one does as a target gets close, by its own help pages.

Helpdesk automation has the prices of the Zendesk, Freshdesk and Help Scout plans that carry SLA deadlines.

What a breach should trigger

A missed target should trigger a reply straight away, with what you know and when you’ll write next. Keep a line on each miss, with its priority, how late it was and why, and read the list once a month. Repeated misses on the same Normal question mean its answer belongs in your docs.

Money changes hands only if a contract says so, and Atlassian’s puts it on uptime. Its SLA credits 10% of the month’s fee for the product when Premium uptime falls below 99.9%, 25% below 99.0% and 50% below 95.0%, against a future invoice. The customer has to claim within 15 days of the month’s end, and the SLA calls credits their “exclusive remedy”. In its own example, 60 minutes of downtime in a 30-day month is 99.86% uptime, which earns the 10%. Its support offerings mention no credits for response times.

Should you publish an SLA without a support team?

Yes, as response targets. Save guarantees and credits for a customer who pays for them in a contract, and have a lawyer read that contract. A page that says when you reply tells a customer when to expect you, and it gives a buyer’s procurement form something to point to.

Publish what you’d keep on a bad week, and name the hours. This is a starting point to paste on your support page. Fill in the brackets, and delete the AI line if you don’t use one.

Support hours: Monday to Friday, 9:00 to 17:00 [time zone], except [country] public holidays.

[Product]'s AI assistant answers questions about [Product] at any hour, in seconds. When it can't help, or you'd rather talk to a person, we reply within these times during support hours.

Urgent, [Product] is down, you can't log in or a payment failed: 2 business hours, then an update every 4 business hours until it's fixed.
Broken for you, with no workaround: 4 business hours.
Questions, billing and bugs with a workaround: 1 business day.
Feature requests and feedback: 2 business days.

We set each message's priority when it arrives. These times are targets we work to. They aren't guaranteed and don't come with service credits.

Keeping an SLA with Deacon

An SLA is easier to keep when fewer messages need you. Deacon answers your customers from your own docs in seconds, at any hour, so a how-to question asked at midnight is answered long before your support hours start. Once 30 questions are in, More detail on Analytics › Widget puts the times people asked against working hours, which shows how much of your support falls outside any hours you’d promise. When your docs don’t cover a question, Deacon tells the visitor it doesn’t have the answer, without guessing at one, and asks for their email. It always declines a question about a visitor’s own account or invoice, whatever your docs say, since only you can look into it.

The questions that need you reach you fast. When a visitor leaves their email or asks for a person, the conversation shows in the Deacon dashboard at once, and if nobody on your team has seen it there, the owners get an email about three minutes later. People waiting for a reply come first in What needs you on Analytics › Widget. The chat then stays open for your reply for three days, which covers a weekend. Your reply shows in it under your name, and if the visitor hasn’t read it there a minute later, it goes to the email address they left, if they left one.

Each answer you write leaves fewer questions waiting on you. Press Answer this on a question your docs didn’t cover, write the reply, and Deacon answers from it straight away. The next person to ask can get it in seconds instead of waiting on your SLA, and Deacon gets better with every answer you write. Topics group the questions and count the people behind each one, so you can see which answers to write first and what customers want you to build next.

Deacon doesn’t time your replies or warn you before a target passes, so keep your SLA in your helpdesk or the spreadsheet above. Human handoff follows a conversation from Deacon to your reply, and you can start on the free plan without a card.