The raw counts
Everything below is a row count from 29 December 2025 to 8 October 2026: volume, not rates. Don't divide one row by another. The email rows include win-back mail to a company's own former users, which isn't cold, and sends from before 10 June 2026, when reply logging started. The LinkedIn request rows include failed attempts. The rates on this page come from the cleaned cuts linked below the table.
| Event | Channel | Count |
|---|---|---|
| Emails sent, win-back included | 49,848 | |
| Replies logged, auto-replies included | 443 | |
| Connection requests logged, failed attempts included | 1,539 | |
| Connection requests withdrawn | 1,826 | |
| Messages sent to connections | 286 | |
| Profile views generated | 1,219 | |
| Booking-link clicks | 139 | |
| Total logged events | Both | 55,598 |
Three deeper cuts of the same log, each with its own method and limits: cold email follow-ups, reply rate by sequence step, LinkedIn outreach, every invitation followed to its outcome, and B2B email verification and list decay. If you want to see where these numbers sit next to everyone else's, I put every public vendor report side by side, with its sample and its definition of a reply, in every public cold email benchmark, compared. For the European market around it (buying online, cloud and AI adoption, the outbound rules), see European B2B sales statistics.
What the numbers actually say
Same replies, two honest denominators: 4.2% per person, 2.07% per email.
269 of the 6,411 people we cold emailed wrote back: 4.2%. Those same 269 replies over the 12,973 emails it took, first emails and follow-ups together, is 2.07%, and the first email alone got 3.07%. Both are true. They divide by different things, which is exactly what a quoted 3% to 5% usually doesn't tell you. Ours includes the segments that died and the lists that turned out to be wrong, and leaves out what flatters a rate: bounces, out-of-office replies, win-back mail and every send from before replies were logged. If a vendor quotes you a rate without telling you what it's divided by, they've told you nothing. The method and the per-step numbers are on the follow-up statistics page.
LinkedIn converts far better per touch, and you can't buy more of it.
21.1% of invitations were accepted: 149 of 706 sent between 7 July 2026 and 15 September 2026, each given at least 35 days. A cold list accepted 18.6%. People who had engaged with one of my posts first accepted 65.8%, a small group of 38, so treat that one as a direction, not a benchmark. Once connected, people talk: of the 263 who accepted an invitation anywhere in the log, 43.3% replied at some point, against 4.2% of the people we emailed. The channel is also capped: LinkedIn limits invitations per week, and pushing past that gets the account restricted. The one that converts best is the one you can't buy more of, which is exactly why the two run together. Every invitation, followed to its outcome, is on the LinkedIn outreach page.
Withdrawn invitations are a feature, not waste.
449 of those 706 invitations went unanswered and were withdrawn, more than were accepted. That's normal, and withdrawing them is the point: LinkedIn caps how many invitations can sit pending, so leaving dead ones open silently throttles everything behind them. Most outreach setups never withdraw anything and then wonder why volume quietly collapses after a few months.
Plan the list size backwards from the reply rate.
At one reply per 48 emails, and with only a fraction of replies becoming meetings, ten meetings a month from email alone takes about 3,200 emails a month if 30% of replies are interested and half of those book. Those two are my planning assumptions, not measured. That's a sending-infrastructure problem before it is a copy problem, which is the part most founders discover after their main domain is already burned. The cold email domain checker shows whether a domain is set up for it before you find out the hard way.
Method, and what this is not
- Source. Every send, reply, connection, withdrawal and click is written to an event row at the moment it happens. The volume table is a straight count of those rows.
- Windows. Volume: 29 December 2025 to 8 October 2026. Cold email rates: emails sent 10 June 2026 to 6 October 2026, replies counted to 20 October 2026. LinkedIn rates: invitations sent 7 July 2026 to 15 September 2026.
- Excluded from the rates. Bounced emails, out-of-office replies, win-back mail to a company's own former users, sends from before reply logging started, and one company whose replies are handled outside this log. That's why the email rates cover 3 companies, not 4.
- Scope. 4 companies selling B2B services and software, mostly into the Netherlands, Belgium and the wider EU.
- Not a survey. Nobody was asked to estimate anything.
- Small n on the client side. 4 companies is enough to be real and not enough to be universal. Your ICP, offer and list quality will move these numbers more than any subject line will.
- No client is identifiable. These are aggregate counts. No company, campaign or individual is named or derivable from them.