What this means for you
189 of 269 human replies answered the first email: 3.07% of 6,155 delivered first emails, one reply per 33. Rewrite that one before you add steps.
29.7% of the replies here answered a follow-up, from people who had let the first email pass. Go past two and the return gets thin: the second follow-up already needed 131 sends per reply.
90.3% of replies came within three days of the email they answered. No answer after a week means the next email is the one that might work, not more waiting.
112 bounced sends are out of every number here. Counted as touches, they make a sequence look longer and a step look weaker than anything a human actually saw.
The numbers that matter
How fair is it to hold 3.07% against the big studies? Belkins' 2026 follow-up study (7,530,489 emails from its own agency campaigns in 2025) puts the first email at 0.59%, so ours is about five times higher. The definitions are close but not identical. Belkins' response rate study on the same emails defines "Reply rate = unique replies ÷ emails sent, excluding auto-replies and bounce notifications": auto-replies out, like ours, but bounced emails stay in its denominator, where we divide by delivered emails, and it doesn't say how a reply is credited to a step. Those differences move a rate by a few percent of itself, not by five times. The rest is the sample: 3 companies with tightly picked, mostly Dutch lists, against a large, US-heavy agency portfolio. Platform data lands in between: Smartlead's State of Cold Email 2026 has the first email at 1.18% per contact, Woodpecker's cold email benchmarks a 1.0% median. The shape agrees everywhere: Instantly's 2026 report says "58% of all replies are generated from step one", we have 70.3%. Every other public cold email report I could find is side by side in every public cold email benchmark, compared.
The short answer
Follow-ups are worth sending, and each one is worth about half the one before it. The first email earned 3.07% replies, the first follow-up 1.45%, the second 0.76%. Together the two follow-ups brought in 29.7% of all replies, so a one-email campaign leaves almost a third of its answers on the table. After the third email the return gets thin: it took 131 sends to get one reply, against 33 for the first.
Replies by sequence step
Raw counts. A reply is counted once per address and pinned to the last email that address got before it replied.
| Step | Emails sent | Replies | Reply rate | Share of all replies |
|---|---|---|---|---|
| First email | 6,155 | 189 | 3.07% | 70.3% |
| Follow-up 1 | 3,924 | 57 | 1.45% | 21.2% |
| Follow-up 2 | 2,623 | 20 | 0.76% | 7.4% |
| Follow-up 3 | 271 | 3 | 1.11% | 1.1% |
| All steps | 12,973 | 269 | 2.07% | 100% |
The fourth step only ran in one company's sequence and went out 271 times. Three replies is not a result, so I'm not drawing anything from that row. Every other sequence in the data stops at three emails.
Days between steps
The sequences waited a median of 5 days before the first follow-up (eight in ten between 3 and 7.1 days) and 7 days before the second (between 5 and 8). That's not enough spread, and 57 and 20 replies are not enough volume, to say whether a shorter or longer gap works better. So there's no chart here. What the data does say: 90.3% of replies arrived within three days of the email they answered, and 94.1% within fourteen. If someone hasn't answered in a week, the next email is the one that might get them.
Bounces are not touches
112 emails in this window bounced, 86 of them first emails. They're out of every number above. A bounced email reached nobody, so counting it as a touch makes a sequence look longer than anything a human actually saw, and it drags the reply rate down for a reason that has nothing to do with the copy. If your tool reports "touches", check whether bounces are in there.
How this compares to what vendors publish
Belkins' sales follow-up statistics (updated 26 June 2026) cover 7,530,489 sent emails over 12 months. They found that "the first email in a sequence delivers the highest per-step reply rate (0.59%)" and that "follow-up emails collectively account for 58.6% of all replies", counting steps 2 through 6.
Woodpecker's follow-up statistics (updated 21 January 2025) say that without a follow-up "you can reach 9% reply rate on average" and that with at least one, "your average reply rate goes around 13%". The page doesn't say how many emails or campaigns that's based on.
The shape agrees with ours: the first email is the best single email, and follow-ups still carry a big part of the replies. The levels differ, and the denominator explains most of it. Belkins counts up to six steps, so more of its replies can land after step 1 than in our mostly three-step sequences. Its per-email rate is lower than ours because 7.5 million emails across many senders is a much wider net than 12,973 emails to tightly picked lists. Woodpecker's 9% and 13% are reply rates per campaign, a different unit from a reply rate per email, and without a sample size they can't be checked.
What I would do with this
I send three emails and stop. The first follow-up goes about five days after the first email, the second about a week later. The numbers back that: the second follow-up still earned one reply per 131 sends, which is worth it on a list you built carefully, and past that I'd rather spend the next email on someone new than on a person who has ignored three. If you want more on how the emails fit next to LinkedIn and calls, the multichannel cadence template and how many touches it takes to book a meeting go further, and the cold email templates show what the emails themselves look like.
Method
- Source. Every send and every reply is written to an event row as it happens. Sends come from the email platform's own statistics, with the sequence step on each row.
- Window. Emails sent 10 June 2026 to 6 October 2026. Replies counted up to 20 October 2026, so every email had at least two weeks to be answered. The window starts on 10 June 2026 because that is the first day replies were logged.
- Scope. 3 B2B companies, 6,411 addresses, all B2B prospect lists.
- What counts as a reply. A human answer. Of 490 logged replies, 137 were automatic (out-of-office, left the company, mailbox no longer read) and removed, by matching the reply text and the reply classifier's own out-of-office label. Each address counts once, at its first human reply, so "not interested" and "please remove me" are replies too.
- Attribution. Each reply goes to the last delivered email sent to that address before the reply arrived.
- Excluded. 112 bounced sends; 2,582 emails from win-back campaigns to a company's own former users, which are not cold email; 815 emails from one company whose replies are handled outside this log; 42 replies from addresses that got no delivered email in the window.
Limitations
- Mixed lists, no test. Different companies, markets and offers, and nobody was randomly assigned a sequence length. People who reply early drop out of later steps, so later steps go to a harder crowd by design.
- The copy changed. Sequences were rewritten during the window, so this is several versions of the emails, not one.
- Step 4 is thin. 271 sends from a single company. Ignore it.
- Our own outreach is in here. One of the 3 companies is Sell Successfully itself.
- Shorter window than the hub. The cold outreach benchmarks page counts sends from December 2025. Replies are only logged from 10 June 2026, so this page uses only the window where both halves exist, and its reply rates are not comparable with the hub's headline figure.