First response time, and whether a bot’s reply counts

HubSpot’s ticket properties call it time to first response.

Calculating first response time from two timestamps

The clock starts when a customer’s message reaches you and stops at the first reply your team sends. Support teams often answer from a helpdesk, a shared inbox built for support, where each customer’s request is called a ticket.

A small online shop got seven emails in one week and answered all of them.

The seven waits add up to 2,042 minutes, so the average is 4 hours and 52 minutes. The Saturday email alone accounts for 1,560 of those minutes. The other six average 1 hour and 20 minutes.

The median is the middle wait once the seven are sorted from shortest to longest. Here that is ticket 5, at 65 minutes, and it would still be 65 minutes if the Saturday email had waited a week.

Helpdesks differ on which one they show, with Help Scout’s email report and Freshdesk’s analytics showing the average. The median is what you get from Gorgias’s metrics and from Zendesk’s report recipe for first reply time, and Intercom’s reporting can show either.

If you answer customers from Gmail, with no helpdesk, you have to work it out yourself. A week of emails is enough. Write down when each one arrived and when you first replied, and leave out any conversation you started yourself. In a spreadsheet, subtract the arrival time from the reply time, then take the AVERAGE and the MEDIAN of that column.

Whether a bot’s reply counts

An automatic email or a bot can reply seconds after every message. Most of the tools below count only a reply from a person.

A person’s reply can be empty too. Some guides accept a personal reply saying the request is being looked into, and Shopify’s guide to first response time counts one. Others want a reply that deals with the question, as Gorgias’s guide to response times does. Zendesk’s own metric stops at the first public comment from an agent, whatever it says.

Business hours or calendar hours

Calendar hours count every hour, including nights and weekends, and business hours count only the time you are open. Use calendar hours if you only track one, because that is how long the customer waited, as Help Scout’s guide to reply times says.

The tools split on this, with calendar hours the default in Zendesk’s ready-made reports and in Gorgias’s metrics. Help Scout’s reports let you switch to business hours. The advice splits too, because Zendesk’s blog on first reply time says to leave out a weekend when you are closed.

A customer writes in at 4:50 pm on a Friday and gets a reply at 9:40 am on Monday. The calendar wait is 64 hours and 50 minutes. A nine-to-five weekday clock counts 10 minutes on Friday and 40 on Monday, 50 minutes in all.

How fast customers expect a first reply

Zendesk’s guide to support metrics gives 24 hours for email and an hour for social media as example targets. An hour on email meets most people’s expectations, and in Toister’s 2020 survey it was fast enough for 88% of them.

Expectations keep rising, and in Zendesk’s 2026 trends research 88% of customers said they expect faster responses than a year before. We checked these figures on 19 September 2026.

Cutting the wait for a first reply

Start with when your slow tickets arrive. If they arrive in the evenings or at weekends, as the shop’s Saturday email did, someone needs to check the inbox then. A single check on Saturday afternoon would have cut that 26-hour wait to a few hours.

You can also answer the questions customers ask again and again before they become tickets. This is called ticket deflection. A help centre, or a bot that answers from your documentation, takes those questions out of the queue and frees your team for the rest.

Deacon is a bot like that for your website. It answers from your own help pages and asks for the visitor’s email address when it cannot answer.

An automatic “we got your message” email tells the customer someone will reply. Most of the tools in the table above do not count it, so it will not lower the number.

How it differs from response time and resolution time

Four ways the first-reply clock goes wrong

Someone on your team can reply in two minutes with a stock answer that does not help. Intercom’s blog on support metrics calls the number easy to game. It is a poor target for any one person. Track first contact resolution beside it to see whether the fast replies solved the problem.

Help Scout’s email report and Intercom’s responsiveness report only time conversations that got a reply. Unanswered emails drop out of the average. Leaving the hardest ones unanswered would make it look better, and a count of the tickets nobody answered shows whether that is happening.

Say a customer replies to a thread from last month. Intercom’s reporting measures the wait from the thread’s original start time, weeks before the new message.

A helpdesk only counts replies sent from inside it. If someone answers a Gorgias ticket straight from Gmail, the reply does not count. HubSpot’s ticket properties miss a reply sent from another inbox in the same way.