A good cold email reply rate depends on what you divide by

Ask what a good cold email reply rate is and the honest answer starts with a question of its own: replies per what? The two largest recent datasets put the average at 0.45 percent and 3.43 percent, and one campaign can produce either kind of number. Once you measure your own results the way a benchmark does, the comparison becomes simple. On Instantly's definition, about 3.4 percent is average, 5.5 percent is top quartile and anything above 10.7 percent puts you in the top tenth of senders.

Two averages, seven times apart

Belkins' 2026 study covers 7,530,489 cold emails its team sent for clients during 2025, which drew 34,393 replies. It reports an average reply rate of 0.45 percent, defined as unique replies divided by emails sent, excluding auto-replies and bounce notifications. Belkins says openly that this looks dramatically lower than its earlier studies, because those divided replies by the number of people who opened. It calls the new denominator stricter and more honest.

Instantly's 2026 Cold Email Benchmark Report, drawn from billions of interactions across thousands of workspaces during 2025, puts the average at 3.43 percent. Its definition is all replies received, follow-up responses included, divided by total emails sent. The report does not say whether out-of-office replies are counted, although it notes that Friday brings the most auto-replies of the week. Many pages that quote a single cold email average repeat this figure; Woodpecker's statistics page and lemlist's guide both lead with 3.43 percent.

Even inside one company the definition moves. Instantly's own guide to reply rate benchmarks divides replies by delivered emails rather than sent ones, calls 5 to 10 percent good for B2B teams, and describes Belkins' figures as clustering in the mid single digits, which matched Belkins' earlier, opener-based method. lemlist's guide to cold email response rates also divides by delivered emails and counts any response as a reply, an out-of-office message included.

One campaign, six reply rates

Here is an invented but ordinary campaign. You write to 100 prospects with a four step sequence that stops when someone replies. Two addresses bounce on the first email. Four people reply to the first email and four to the second. Of those eight replies, three are out-of-office messages, two are a polite no and three are interested. Everyone else receives all four emails. That comes to 374 emails sent and 372 delivered, and your tracker shows opens from 40 people.

• All replies divided by emails sent, Instantly's definition: 8 of 374, or 2.1 percent.
• All replies divided by emails delivered, lemlist's definition: 8 of 372, or 2.2 percent.
• Human replies divided by emails sent, Belkins' current definition: 5 of 374, or 1.3 percent.
• People who replied, out of the 98 people the campaign reached: 8 of 98, or 8.2 percent.
• People who replied themselves, leaving out the auto-replies: 5 of 98, or 5.1 percent.
• Human replies divided by people who opened, Belkins' earlier method: 5 of 40, or 12.5 percent.

Every line is correct arithmetic about the same eight emails. The per person figures come out nearly four times the per email ones here because most people received all four steps; a longer sequence widens that gap and a single email closes it. The number your pipeline actually runs on is not in the list at all. Three interested people out of 98 is about 3.1 percent.

A reply rate without its denominator is a number without a unit.

Measure yours so it compares fairly

• For your own decisions, count people. Someone who answers your third email is one reply from one person, and a per email rate makes a longer sequence look worse even when it produces more conversations.
• Take bounces out of the bottom and auto-replies out of the top. Neither tells you whether the message worked.
• Keep a separate count of interested replies. A polite no is useful to know, but it is not pipeline.
• To compare with a published average, convert first. Divide your human replies by every email you sent and you have Belkins' measure; divide all replies by every email sent and you have Instantly's.
• Wait for enough sends before judging. With 98 people in a campaign, a single reply moves a per person rate by about one point.

So what is good? Measured Instantly's way, its report puts 3.43 percent at the average, 5.5 percent at the top quartile and 10.7 percent at the top tenth. Measured Belkins' way, 0.45 percent was the average across its 2025 campaigns, ranging from 0.22 percent at companies with 10,000 or more employees to 0.72 percent at companies with ten or fewer. If you write individual, researched emails from Gmail, neither dataset is made of email like yours, and your own per person rate over a few months will tell you more than either.

How Fast Mail Tracker counts a reply

In Fast Mail Tracker, a campaign's reply rate is per person: the share of the people the campaign was sent to who have replied, with each person counted once however many messages they send back. A reply is recorded when the extension sees a message from someone other than you in the Gmail thread. The tracker reports what is in the thread and does not guess at intent, so an out-of-office answer from a recipient counts as a reply too. That makes the honest next step the one this article recommends: open the replies, set the automatic ones aside and count the interested people yourself. Each campaign shows its open, click and reply rates side by side, and on cold email the open rate stays the weakest of the three.

Fast Mail Tracker is in early access while Google's security review completes, so places are limited; to see per person reply rates on your own campaigns, use the install button below.