Cold email statistics: reply rates traced to the original data

This page is for SDRs, founders and sales leaders who want to know whether a cold email campaign is working, or who need a realistic target before they start one. Every figure below comes from the original study or vendor report, opened in October 2026, with its sample, period and definition. Nearly all cold email data is vendor research, because only sending platforms and outreach agencies see millions of emails, so each source is labelled and each figure comes with what it divides by what.

Cold email reply rate benchmarks at a glance

Read the table by its last column. Smartlead and Hunter count people; Belkins and Instantly count emails. Hunter's 4.5% average mixes sales with digital PR (13%), headhunting (7.5%) and marketing (6.2%) campaigns, which is why its sales-only figure is lower. The 2019 Backlinko figure is still the most quoted cold email statistic, but it measured site owners answering link-building and guest post requests, not buyers answering sales emails.

The most transparent measure is Smartlead's: the median sender gets one reply per 135 contacts, the top quarter one per 70 and the top tenth one per 38. A reply is not a meeting. Smartlead says its data "measures replies, not meetings or revenue", and none of these sources publishes a meeting rate from the same data.

Why published cold email statistics disagree

Four things explain most of the spread:

• The denominator. A three-step sequence to 1,000 people sends close to 3,000 emails. Thirty replies is 3% per contact but about 1% per email sent.
• What counts as a reply. Belkins excludes auto-replies and bounce notices. Most other sources don't say, and in many sending tools an out-of-office message or "please remove me" counts as a reply.
• Who is in the sample. A self-serve platform includes every new sender on a fresh domain. An agency dataset reflects one team's process. Hunter's includes PR and recruiting, which get more replies than sales.
• Average or median. A few very good senders pull an average up. Smartlead publishes a median and percentiles, which is closer to what a typical team sees.

Definitions change inside one vendor too. Belkins' earlier studies used unique openers as the denominator. Its 2025 data divides by emails sent, with open tracking switched off, so its 0.45% can't be compared with its older, higher figures.

What moves reply rates in the data

These findings come from the same vendor datasets. They're correlations across many senders, not controlled experiments, so treat them as hypotheses to test on your own list.

Three patterns hold across sources. Smaller, tighter lists beat big sends. The first email does most of the work, and each follow-up earns less than the one before, but the second and third still add replies. Shorter emails do better. The open tracking gap may partly reflect who uses tracking, but Belkins switched it off after the tracking pixel hurt its deliverability in 2024, which points the same way.

List size is the finding most teams can act on, and it moves the work upstream. A 30-account send only works if those 30 accounts are a strong fit, which is a research problem more than a writing one. The guide to building a prospect list covers the method. Bebop does the matching from a plain-language description and explains why each company fits, so a small, sharp send doesn't cost a day of research.

Open rates no longer tell you much

Hunter's 2025 data puts the average cold email open rate at 30%. Treat any open rate with caution. Apple's Mail Privacy Protection, announced in June 2021, "stops senders from using invisible pixels to collect information about the user" and prevents senders from knowing when Apple Mail users open an email. And as Belkins found, the tracking pixel itself can hurt deliverability. Judge cold email on replies, positive replies and meetings. For marketing email, where opens are still reported, see the email open rate benchmarks.

Deliverability rules that cap your numbers

Since February 2024, the largest inbox providers have enforced sender rules. Mail from a domain that breaks them lands in spam or is rejected, and no copy fixes that.

A cold list has no opt-in, so every recipient can click "report spam". At 0.3%, three reports in every 1,000 emails puts a domain over Gmail's limit. Hunter's 2025 average bounce rate was 3.6%. Verify addresses before you send and keep each mailbox's daily volume modest.

The legal floor in the US and UK

This is general information, not legal advice.

• United States: CAN-SPAM covers B2B email. The FTC's guide says "the law makes no exception for business-to-business email." Use accurate headers and an honest subject line, identify the message as an ad, include a valid postal address and an opt-out, and honor opt-outs within 10 business days. The FTC guide lists penalties of up to $53,088 for each email that breaks the law.
• United Kingdom: PECR treats companies and sole traders differently. The ICO says you can email corporate bodies without prior consent, and that it's good practice to keep a do-not-email list of businesses that object. Sole traders and some partnerships count as individuals, so you need their specific consent. Emailing a named employee at their own work address also brings in data protection law.

How to measure your own cold email numbers

Your own numbers beat any benchmark, because they reflect your list, your offer and your sending domain. Track each campaign from one start date, with fixed definitions:

Illustrative example. The company, campaign and numbers are invented to show the arithmetic.

A seller at Northwind Analytics (northwind.example) emails 400 operations leaders in a three-step sequence.

• 12 addresses bounce, so 388 contacts are delivered. Bounce rate: 12 / 400 = 3.0%.
• 21 people reply. Five replies are out-of-office messages, so 16 are real. Reply rate per contact: 16 / 388 = 4.1%.
• People who reply leave the sequence, so the tool sends 1,129 emails in all. Per email sent, the same 16 replies are 16 / 1,129 = 1.4%. Same campaign, two "reply rates".
• Six replies are interested (6 / 388 = 1.5%), and three become meetings (3 / 388 = 0.8%).
• Contacts per meeting: 388 / 3 = about 129.

To book ten meetings a month at that rate, the seller needs about 1,290 delivered contacts a month, or a better meeting rate. The numbers point at targeting and the offer, not the subject line.

How many contacts before a test means anything

Reply rates are small, so chance swings them a lot. To tell a 2% reply rate from a 3% one with the usual standard (95% confidence, 80% power, a two-proportion test), you need about 3,800 delivered contacts in each version. With 200 contacts per version, 4 replies against 6 is noise. If your volume is low, test big changes, such as a different audience or offer, rather than small ones like a subject line, and judge over several weeks.

Campaign: [name]    Dates: [start] to [end]
Audience: [who, and why they fit]

Contacts uploaded: [ ]
Bounced: [ ]    Bounce rate (bounced / sent): [ ]%
Contacts delivered: [ ]

Real replies (no auto-replies): [ ]
Reply rate per contact (replies / delivered): [ ]%
Positive replies: [ ]    Positive reply rate: [ ]%
Meetings held: [ ]    Meeting rate: [ ]%
Contacts per meeting (delivered / meetings): [ ]

Replies by step:  1: [ ]   2: [ ]   3: [ ]
Spam rate (Postmaster Tools): [ ]%

One thing to change next time: [ ]

Where to look when the numbers are low

Common misreadings

• Comparing a per-contact rate with a per-email rate. Convert one before you compare. The same campaign can show 4.1% and 1.4%.
• Quoting 8.5% as a sales benchmark. It came from 2019 link-building outreach.
• Optimizing for opens. Opens are unreliable, and tracking them can cost you replies.
• Calling a test after 200 sends. The difference is almost always noise at that size.
• Treating a reply as a win. Track positive replies and meetings, because those are what the business pays for.

Next step: write the first email with the step-by-step guide to writing a cold email, then plan the sequence with the cold email follow-up templates.