The short version
- Yes, it still works. Four large studies published in 2026 all record replies. Their average or median reply rates sit between 0.45% and 3.7%.
- Much of that spread is about counting. Reports divide replies by emails sent, emails delivered or people contacted, and some count out-of-office messages as replies.
- Worldwide reply rate: 5.33% reply rate across 421,456 people contacted for 33 clients in 2026, 55% above the 3.43% industry average (source: Instantly 2026 Cold Email Benchmark Report). We only count real human replies, not out-of-office.
- What works: a verified list of the right decision-makers, an offer that fits their job, a few emails per inbox per day and correct authentication.
- What does not: enterprise-only targets, consumers, a vague offer and a need for leads this week.
What do the 2026 cold email benchmarks say?
They say cold email still gets replies, but a typical campaign gets few of them and the best get many more. Here are four large public studies published in 2026, next to our own data.
| Study | Sample | Period | Typical reply rate |
|---|---|---|---|
| Saleshandy | 53.1 million emails, 60,000 sequences | January to June 2026 | 3.7% average |
| Woodpecker | 56,614 campaigns | Last 90 days of 2025 | 1.5% median |
| Smartlead | 850 million+ emails | Not stated (published 25 September 2026) | 0.74% median |
| Belkins | 7,530,489 emails, 34,393 replies | January to December 2025 | 0.45% average |
| Newlead's own outbound | 371,590 emails to 143,319 decision-makers | Late 2023 to 10 August 2026 | 1.89% |
The best campaigns do far better. Saleshandy's top campaigns pass 10%, and Smartlead's top 10% start at 2.63% of people contacted. That gap between typical and top matters more than any average, and the sections below cover what creates it.
Lower numbers in some 2026 reports do not mean cold email stopped working. Belkins' 2026 average is lower than in its older studies, and Belkins says that is “not because cold email fell off a cliff overnight” but because it now divides replies by emails sent instead of by people who opened.
How should you read a reply-rate claim?
Check what is divided by what before you compare two numbers. The same campaign can show very different reply rates depending on the base and on which replies count.
| How replies are counted | Used by | What it does to the number |
|---|---|---|
| Unique replies divided by emails sent, auto-replies and bounce notifications excluded | Belkins, 0.45% average | The strictest base. Every follow-up adds to it, so longer sequences tend to score lower |
| Replies divided by people contacted, positive or not | Smartlead, 0.74% median | One rate per person, however many emails each one got |
| Unique human replies divided by unique people contacted, auto-replies excluded | Newlead, 1.89% | Same base as Smartlead. Out-of-office and other auto-replies are left out |
| Not stated (a median across campaigns) | Woodpecker, 1.5% median | Useful as a reference, hard to compare directly |
| Total replies divided by emails delivered | Saleshandy, 3.7% average | Counts total replies, not unique people, so it reads higher than a unique count would |
| Any response, “even an out-of-office message” | lemlist guidance (not a study) | Automatic replies push the rate up |
| Replies divided by people who opened | Older Belkins studies | Much higher numbers that do not compare with the rows above |
Three more checks before you compare:
- Per email or per person? A sequence sends each person several emails. We sent 2.59 emails per person in our own outbound, so the same results measured per email would show a much lower rate.
- Average or median? A few very strong senders pull an average up. A median shows the middle campaign, so it sits lower when a few senders do very well.
- Replies or opportunities? A reply rate counts a “no thanks” too. What pays is a live sales conversation, and in our own outbound we contacted 202 people for each one.
What we see in our own campaigns
Cold email works for us, but results swing with the market, the segment and the copy. These numbers come from Newlead's own outbound, the campaigns we run to win our own clients. They are not client results. The full data and method are in our cold email benchmarks 2026.
From late 2023 to 10 August 2026 we sent 371,590 emails to 143,319 unique decision-makers in 24 campaigns, 82.5% of the volume in 2026. Of those people, 1.89% replied, counting unique human replies and excluding auto-replies. The bounce rate was 0.62%. We contacted 202 people for every opportunity, meaning a reply positive enough to log as a live sales conversation.
5.33% reply rate across 421,456 people contacted for 33 clients in 2026, 55% above the 3.43% industry average (source: Instantly 2026 Cold Email Benchmark Report). We only count real human replies, not out-of-office.
| Market | People contacted | Opportunity rate |
|---|---|---|
| Sweden | 11,511 | 1.46% |
| United Kingdom | 11,726 | 0.97% |
| United States | 47,424 | 0.31% |
5.33% reply rate across 421,456 people contacted for 33 clients in 2026, 55% above the 3.43% industry average (source: Instantly 2026 Cold Email Benchmark Report). We only count real human replies, not out-of-office.
Most replies come from the first email
In our largest campaign, 117,475 emails to financial services firms in the US and Canada, the first email brought 60.0% of all replies. The second brought 19.2% and the third 20.8%. The opening email decides most of the result, but the follow-ups still added two in five replies.
The copy moves results
In the same campaign, two versions of the first email got reply rates of 0.67% over 43,513 sends and 0.25% over 12,998 sends. That is a 2.6x difference between two first emails in one campaign.
What separates cold email that works from cold email that fails?
Four things: the right list, an offer that fits the reader's job, low volume per inbox and clean authentication. Each is backed by the data or the mailbox rules below.
A verified list of the right people
Send to decision-makers who plausibly have the problem you solve, not to everyone with a matching title. High bounce rates go with fewer replies too. In Woodpecker's data, campaigns that kept bounces under 2% got a 1.6% median reply rate, against 1.0% for campaigns bouncing over 20%.
We verify every address before it enters a sequence and rebuild lists monthly instead of reusing them. Across our own outbound, 0.62% of emails bounced.
An offer that fits the reader's job
In a Gartner survey of 632 B2B buyers, 73% said they actively avoid suppliers who send irrelevant outreach. Write about one problem the reader is responsible for, say what changes for them, and keep it short enough to read in a few seconds. Our A/B test above shows how much the first email alone can change.
A few emails per inbox per day
Lower volume per inbox goes with more replies. Woodpecker found a 2.0% median reply rate for mailboxes sending 1-20 emails a day, against 0.5% for mailboxes sending 101-200. A typical Newlead setup sends about four emails per inbox per day, spread over 100-250 inboxes on 4-6 dedicated sending domains, separate from the client's main domain. More in how many cold emails per day per inbox.
Authentication that meets mailbox rules
The big mailbox providers tightened their rules in 2024 and 2025. Since February 2024, Gmail requires every sender to personal Gmail accounts to set up SPF or DKIM, valid forward and reverse DNS and TLS, and to keep the spam rate below 0.3%. Google advises staying below 0.10%. Senders of close to 5,000 messages a day to personal Gmail accounts must also use SPF, DKIM and DMARC and offer one-click unsubscribe on marketing mail, and Gmail stepped up enforcement in November 2025.
Outlook.com has required SPF, DKIM and DMARC from domains sending over 5,000 emails a day since 5 May 2025. Google's rules cover personal Gmail accounts, not Google Workspace business accounts, and a typical setup like ours sends far fewer than 5,000 emails a day in total. Authentication and a low spam rate are still the baseline. New domains also need time: we warm dedicated sending domains for 14 days before the first emails go out. Rules last checked 3 October 2026.
This is how we run our outbound lead generation services: verified lists, copy tested in several variants, dedicated sending domains and inboxes, low volume per inbox, and every interested lead delivered to your Slack with name, company, email and phone number.
When does cold email not work?
Cold email is a poor fit when your buyers are only large enterprises or consumers, when the offer is vague, or when you need leads this week.
Your buyers are only large enterprises
In our own outbound, selling our own service, enterprise targets took far more effort. US enterprise accounting firms replied at 0.44%: 22 replies and 2 opportunities from 5,035 people contacted, or 2,518 contacts per opportunity. The rest of our US accounting campaigns took 284. In Sweden, enterprises took 135 contacts per opportunity, against 47 for B2B services firms.
| Target (our own outbound) | People contacted per opportunity |
|---|---|
| Small and mid-market, Nordics | 40 to 50 |
| Small and mid-market, UK | About 100 |
| Small and mid-market, US | 280 to 370 |
| Professional services, US | About 610 |
| Enterprise, US | 2,500+ |
These numbers show how each group answered our own outreach, selling our own service. They are not the rates your prospects will give you. The pattern is the point: in both Sweden and the US, enterprises took several times more contacts per opportunity than small and mid-sized companies. For enterprise-only targets, email alone is a weak main channel. Compare the options in cold email vs LinkedIn outreach.
Your buyers are consumers
Cold email is a B2B channel. In the UK and the EU, marketing email to private individuals needs prior consent, apart from a narrow exception for existing customers (PECR regulation 22, ePrivacy Directive Article 13). Sweden's Marketing Act (section 19) says the same for natural persons. For business recipients, whether cold email is legal depends on the country. This is not legal advice. Last checked 3 October 2026.
Your offer is vague
If you cannot say in two sentences who you help, with what and why it matters now, more sending will not fix it. Our A/B test showed a 2.6x gap between two first emails in the same campaign. A weak message wastes a good list, so fix the offer first and scale after.
You need leads this week
New sending domains need 14 days of warm-up before the first emails go out, so cold email is not a same-week fix. See how long cold email takes to work.
Sources
- Newlead, Cold email benchmarks 2026 (own outbound data, published 10 August 2026)
- Saleshandy, Cold email statistics (published 7 June 2026)
- Woodpecker, Cold email benchmarks (updated 5 August 2026)
- Smartlead, Cold email reply rate (published 25 September 2026)
- Belkins, Cold email response rates (updated 26 June 2026)
- lemlist, Cold email response rate (updated 28 September 2026)
- Gartner, B2B buyer survey press release (25 June 2025)
- Google, Email sender guidelines
- Google, Email sender guidelines FAQ
- Microsoft, Outlook.com postmaster announcements
- UK Privacy and Electronic Communications Regulations, regulation 22
- EU ePrivacy Directive 2002/58/EC, Article 13
- Swedish Marketing Act (2008:486), section 19
Last checked 3 October 2026.