Cold Email Reply Rate Benchmarks, and Why Most Are Misleading

Published cold email reply rate benchmarks range from 0.45% (Belkins, 2025 client campaigns) to 8.5% (Backlinko and Pitchbox, 2019). Most of that gap is not performance. It comes from what each study divides by: emails sent, prospects, or prospects who opened, and whether auto-replies count. Before comparing yourself to any number, match its formula.

This post lists five benchmarks we fetched on 1 October 2026, quotes only the figures that appeared on each page, and shows how one campaign can produce a reply rate anywhere from 1.5% to 12.9% depending on the formula.

The benchmarks people quote, side by side

The commonly cited figures come from sending platforms, an outreach agency and a link-building tool, and no two use the same definition. Here is what each page actually says.

Source	Headline figure	Sample stated	How "reply rate" is defined	Main problem
Instantly, Cold Email Benchmark Report 2026	3.43% average; top quartile 5.5%+; top 10% 10.7%+	"billions of cold email interactions across thousands of active workspaces", 1 Jan to 18 Dec 2025	"all replies received (including follow-up responses) divided by total emails sent"	Auto-reply handling not stated; per-send, so longer sequences dilute it
Belkins, B2B cold email response rates (2026 study)	0.45% average	7,530,489 emails, 34,393 replies, Jan to Dec 2025	Unique replies divided by emails sent, excluding auto-replies and bounce notifications	One agency's client campaigns; earlier editions used a different denominator
Backlinko and Pitchbox email outreach study	"Only 8.5% of outreach emails receive a response"	12 million emails; page last updated April 2019	Not set out on the page (methods are in a separate PDF)	Link-building, guest post and PR outreach, not sales; seven years old
lemlist, how many follow-ups	4.5% for email 1; 22.37% across 10 emails	Not stated on the page	Not stated on the page	No sample size, period or definition published alongside the figure
Woodpecker, cold email statistics	"Average cold email reply rate: 3.43% platform-wide"	Title says "over 20M cold emails"	Not stated	The 3.43% is attributed to Instantly's report, not Woodpecker's own data

Belkins' arithmetic checks out: 34,393 divided by 7,530,489 is 0.457%. That is the only benchmark in the set where we could reproduce the headline from numbers on the same page.

Problem 1: the denominator decides the answer

The single largest source of disagreement is what goes underneath the line. There are three common choices, and they produce very different numbers from identical activity.

• Per email sent. Every step of a sequence counts as a send. Instantly and Belkins both use this. A four-step sequence to 1,000 people is roughly 4,000 sends, so the rate falls as you add follow-ups even if more people reply.
• Per prospect. One row per person contacted. This answers "what share of the people I emailed wrote back", which is usually the question you care about.
• Per opener. Replies divided by people who opened. Belkins says its earlier studies did this: "In prior studies, reply rates were calculated against the number of unique recipients who opened an email." Its 2026 page then states: "A 5% reply rate against openers and a 0.45% reply rate against total sends can describe the same campaign; they're just measuring different things."

That last sentence is the most honest line in any benchmark report we read, and it should be printed above every other one. It also means any year-on-year comparison between Belkins' 2026 figure and its earlier editions compares two different formulas, not two different years of performance.

The per-opener denominator has a second flaw. Open tracking relies on a pixel loading, and some mail clients and security scanners load images automatically without a person reading anything. The open count is unreliable, so a rate divided by it inherits that noise. We cover why opens mislead in more detail in writing a sunset policy.

Problem 2: auto-replies and "remove me"

A reply rate that counts out-of-office messages is partly measuring holiday season. Belkins excludes auto-replies and bounce notifications explicitly. Instantly's report defines reply rate as "all replies received" and does not say whether automatic replies are filtered out. The other pages do not say either way.

The same problem applies at the other end. "Please remove me" is a reply. So is "wrong person, try Sam". A raw reply rate puts both alongside "yes, Tuesday works". Instantly's own blog benchmarks page recommends tracking "Positive, referral, objection, not now, unsubscribe, out of office" as separate categories, which is right, but its headline benchmarks are not broken down that way.

None of the five sources publishes a positive reply rate with a stated definition and sample. If you see a "positive reply rate benchmark" elsewhere, ask who decided what counted as positive.

Problem 3: who is in the sample

Every one of these datasets comes from people who chose to use a particular product or agency, and that shapes the result.

• Platform data (Instantly, Woodpecker) covers that platform's active customers. Accounts that burned their domains and churned in March are not around in December to drag the average down, and power users who send the most volume weigh the most.
• Agency data (Belkins) covers campaigns run by professionals for clients who paid for them. That is a different population from a founder sending their first sequence. Belkins' own breakdown shows how much the mix matters: 0.72% for companies of 0 to 10 employees against 0.22% for companies with 10,000 or more.
• Tool data (Backlinko and Pitchbox) covers link-building and content outreach: guest posts, infographics, resource pages, roundups. Asking a blogger to link to an article is not the same request as asking a VP for a sales call.

There is also the plain incentive. Every source here sells cold email software or cold email services. That does not make the numbers wrong, but it is a reason to prefer the ones that publish their definition and sample, which in this set means Belkins and, partly, Instantly.

Problem 4: citations that loop

Some widely shared figures are one study quoted many times, not many studies agreeing. Woodpecker's page title advertises "over 20M cold emails", yet its headline reply rate is Instantly's 3.43%. Instantly's blog in turn cites Backlinko's 8.5% from a link-building study last updated in 2019. Search for "cold email reply rate" and you will find the same handful of numbers repeated across dozens of pages, often without the definitions.

If three articles agree on a number, check whether they are three sources or one.

The same campaign, six different reply rates

To show how much the formula matters, here is one hypothetical campaign measured six ways. The numbers are invented for illustration, not data from any customer.

Assumptions:

• 1,000 prospects, four-step sequence, the tool stops a prospect when a human replies.
• 30 addresses hard bounce on step 1 (an unverified list) and are removed.
• 58 human replies in total: 30 after step 1, 15 after step 2, 8 after step 3, 5 after step 4.
• Of those 58, 18 are interested. The rest are referrals, "not now", objections and removal requests.
• 20 automatic out-of-office replies arrive; those prospects stay in the sequence.
• The tracking pixel reports 450 opens.

Sends: 1,000 + 940 + 925 + 917 = 3,782 emails.

Formula	Calculation	Reply rate
Human replies per email sent (Belkins-style)	58 / 3,782	1.53%
All replies incl. auto per email sent	78 / 3,782	2.06%
Interested replies per delivered prospect	18 / 970	1.86%
Human replies per delivered prospect	58 / 970	5.98%
All replies incl. auto per prospect	78 / 1,000	7.80%
Human replies per opener	58 / 450	12.9%

Same emails, same people, same replies. Depending on the formula, this campaign is "below average", "solid" or "best in class" against the tiers on Instantly's blog ("5-10% is solid for B2B. 10-15% is excellent").

Notice the bounces as well. Thirty dead addresses are in the per-send denominator but could never have replied. A list that is not verified lowers every per-send rate and hurts sender reputation at the same time, which is the subject of why cold emails bounce.

How to measure your own reply rate properly

Use per-prospect rates with automatic replies removed, and keep the raw counts so anyone can recompute them. These steps work in any sending tool that exports per-contact status.

1. Export at prospect level, one row per person, with the final status and the step they reached.
2. Remove hard bounces from the denominator and report them separately as a bounce rate. They tell you about list quality, not message quality.
3. Classify every reply by hand or with a reviewed rule: interested, referral, not now, objection, removal request, out of office or auto-reply.
4. Drop auto-replies from the reply count. Keep them in a separate column. An out-of-office message confirms the mailbox is live, which is useful for hygiene.
5. Report three numbers: human replies per delivered prospect, interested replies per delivered prospect, and removal requests per delivered prospect.
6. Report the step breakdown separately as replies per email sent at each step, so you can see where a sequence stops paying for itself. Our post on follow-up sequences uses this.
7. Write the definition next to the number every time you share it. "6% reply rate" means nothing; "6% human replies per delivered prospect, 4-step sequence, n=970" means something.

Which benchmark to compare against

Pick the one whose formula matches yours, and treat the rest as background.

• If you measure per email sent with auto-replies removed, Belkins' 0.45% for 2025 is the closest like-for-like figure with a published sample, keeping in mind that it is one agency's campaigns.
• If you measure per email sent and are unsure about auto-replies, Instantly's 3.43% and its quartile cut-offs (5.5% and 10.7%) are the nearest match.
• If you measure per prospect, none of these five pages gives a directly comparable number. Your own history is the better baseline: compare this month's campaign with last month's, measured the same way.
• If someone quotes 8.5%, check whether they know it is link-building outreach from a study last updated in 2019.

What to take from this

A reply rate is only meaningful with its formula attached. Before you conclude that your campaign is failing or thriving, rewrite the benchmark and your own number in the same terms: same denominator, same treatment of auto-replies, bounces removed. Most of the time the "gap" either shrinks or flips.

The cheapest improvement to measured reply rate is also the least glamorous: stop putting addresses that cannot reply into the denominator. You can check a sample of your list with the free email verifier before the next send.