First response time: benchmarks and how to cut yours

By the Onsites AI team · Last updated August 28, 2026 · 4-minute read

First response time is the support metric buyers actually experience — and they experience it against the clock of the channel, not the clock of your team. A chat answered in five minutes is a ghosted chat; an email answered in two hours is a delightful email; a WhatsApp message answered tomorrow morning is a buyer already talking to a competitor. This guide sets the 2026 benchmarks per channel, shows how to measure the number honestly (medians, channel-clocked, first-reply-not-first-autoresponder), and walks the five queue habits that cut time-to-first-reply without a single hire — staffing shape, routing, drafts, the morning list, and expectation-setting (the channel guide's published hours). Every benchmark here is achievable by a three-person desk on a free tier; none requires monitoring, only measurement.

The benchmarks, channel by channel

Live chat: under two minutes, wall clock — chat is a conversation, and the buyer is sitting at the other end; beyond two minutes the chat's "still there?" message buys nothing. A desk that cannot staff chat continuously should publish its hours and hide the widget outside them (the live chat guide works the staffing arithmetic). WhatsApp and IM channels: 15–30 minutes in waking hours — these channels feel personal, and the buyer checking twice in an hour expects an answer twice; the channel playbooks carry the tone. Email: one to two business hours as the 2026 bar (an hour is the competitive edge, four hours is the old world); an after-hours auto-reply that states the morning is trust, not delay. Social DMs: under four hours — the only channel where the wait is public, and public waiting is marketing for your competitors. The pattern in all four: the channel sets the clock; the desk's job is to be honest about which clocks it keeps.

Measuring it honestly

Three rules keep the number true. Median, never mean: one thread opened Friday and answered Monday triples the mean and barely moves the median; the median is what the typical buyer felt. The channel's clock, business hours only where the channel is business-hours: an 11 pm email answered at 9:08 am is a 4-minute first response, not an 10-hour failure; chat and IM measure wall clock, email and DMs measure desk hours. First human reply, not first autoresponder: "we received your message" is a receipt, not a response — count the first reply that moves the thread (answers, offers a time, asks a question); the receipt-only desk inflates its honesty. Track it per channel, weekly, as a trend — the KPI guide's scoreboard carries it — and against your own history before you measure it against the industry: a three-person desk cutting 5 hours to 90 minutes has moved a cliff, whatever any benchmark says.

The five habits that cut it without hiring

One — shape the hours to the peaks: volume has a shape (the busiest-hour KPI); one overlap shift moved into the peak hour moves median first response more than any tool — measure the peak, then place the people. Two — route, don't triage: the inquiry sitting in a shared inbox waiting for "who wants this?" is the desk's most expensive minute; assignments by channel, load or account (the routing guide) make ownership instant. Three — draft instantly, review in seconds: the copilot drafts the first reply from the order or the handbook; the human edits on a phone and sends — a 4-minute email first response from a two-person desk, at credit prices that keep the line in single digits. Four — the morning list: the oldest five threads, first coffee — nothing else moves backlog age and the response tail as cheaply. Five — publish the hours you keep: the away-reply stating 9–5 with a morning guarantee converts silent waiting into trusted waiting; buyers forgive clocks they can read (the satisfaction they rate later reflects the honesty). Five habits, zero hires — the cliff is climbed by systems.

The tail: where small desks actually lose

The median hides the failure mode: the 10% of threads answered a day late, each one a review in progress. Watch the tail: the 90th percentile of first response — a desk whose median is 40 minutes and whose tail is 20 hours is a desk whose mornings are organized and whose afternoons are noise. Fix the tail structurally: the morning list owns the tail's overnight half; the routing rule that a thread unclaimed in 15 minutes escalates to anyone owns the midday half; the auto-reply that states when the thread will be answered owns the honest remainder. The tail's special case — the saga: the 60-message thread nobody wants is precisely where first response matters most for the relationship (the escalation matrix gives it an owner and a clock). The tail is where small desks earn their reputation disproportionately: the buyer with no other choice forgives nothing, and the one whose thread got answered first thing in the morning tells everyone.

A worked month

One small desk's before and after. A two-person team, 600 conversations monthly across email (70%), chat (20%), IM (10%): median first response — 4.5 hours email, 6 minutes chat, 2 hours IM; tail (90th) — 22 hours email. The two changes: routing and the morning list in week one (median email — 2.1 hours, tail — 9 hours); drafts with phone-review in week two (median email — 55 minutes, tail — 4 hours). The team's staff count: unchanged. The costs: routing and lists are configuration; the drafts' credits for a month of first replies — under $10 (the published table). The reviews mentioning response time by name: three in the following month, two of them unprompted. The benchmarks are not the point of the exercise — the point is the buyer's felt clock per channel — but the desk that meets them does so by systems, not heroics, and the worked month is what that looks like in practice, numbers and all: a small team holding clocks its size has no business holding, and holding them without overtime.