# Which Follow-Up Gets the Most Replies? Reply Data by Sequence Step

> Send 4 emails, 3-6 days apart. In Ken's 2026 data, the 4th email drew human replies at the opener's rate, and 59% of replies came after a follow-up.

Oct 31, 2026 · Cristian Frunze · 13 min · Guide

Tags: how many follow-up emails to send, cold email follow-up, reply rate by sequence step, cold email sequence, follow-up email timing, cold email data

![A manager behind a mail heap misses the squashed yellow envelope at the bottom and reaches for the matching one a hand sets on top.](/images/blog/cold-email-follow-up-reply-rate-by-step-cover.webp)

Open the step report in most cold email tools and the first email wins by a mile. Ours did too. 52% of our replies answered email one. Then we took out the out-of-office replies, and that share fell to 41%.

**Send four emails, the opener plus three follow-ups, three to six days apart.** Counting only replies from real people, our fourth email drew 0.99x the first email's replies per email sent, and 59% of the people who wrote back did it after a follow-up.

**The first email gets the most replies because it reaches the most people, not because it's the best email.**

The numbers below come from every email we sent to contacts first emailed between February 25 and August 3, 2026, across 17 workspaces (separate sender accounts on our platform), set next to seven published studies.

## Which email in a sequence gets the most replies?

By raw count, the first email wins. Per email sent, the first and the fourth email are close to a tie.

The first email collected 41% of first human replies. But it also went to every contact. By the fourth email, the audience had shrunk by 72%, because people replied, bounced, got pulled, or sat in a campaign with only three emails. Fewer emails out means fewer replies back, even if each email works just as well.

So the fair test is replies per email sent, measured against email 1 inside the same workspace. The numbers in brackets are 95% confidence intervals, the range the true value very likely sits in.

| Email | Emails sent, as a share of email 1 | Share of human replies | Human replies per email vs email 1 (95% CI) | All replies per email vs email 1, incl. out-of-office |
|---|---|---|---|---|
| 1 | 100% | 41% | 1.00 | 1.00 |
| 2 | 77% | 22% | 0.71 (0.58-0.86) | 0.52 |
| 3 | 61% | 20% | 0.85 (0.69-1.04) | 0.47 |
| 4 | 28% | 12% | 0.99 (0.76-1.28) | 0.57 |
| 5 and later | 9% | 4% | Too few to measure | 0.46 |

*Ken sending data, contacts first emailed Feb 25 to Aug 3, 2026, 17 workspaces. Ratios compare each email with email 1 inside the same workspace.*

The second email is the weak spot, at about 0.7x the opener. Then the curve climbs back. The third email's range includes 1.0, and the fourth sits right on it. If follow-ups were just fading echoes of the first email, those numbers would keep falling. They don't.

Email 4 also has the harder crowd. It only reaches people who ignored three emails in a row. If anyone should be hard to reach, it's them, and they still answered at the opener's rate.

![Line chart: counting only human replies, Ken's fourth email drew replies per email at 0.99 times the first email, while Ken's all-replies line and the published Belkins and lemlist series fall to about half](/images/blog/cold-email-follow-up-reply-rate-by-step-1.svg)
*Source: Ken sending data, contacts first emailed Feb 25 to Aug 3, 2026, 17 workspaces. Ken lines are within-workspace ratios with 95% confidence intervals. Belkins and lemlist per-step rates as published, checked October 2026.*

## Why do most studies say the first email does all the work?

Most studies don't say whether they count auto-replies, and most report shares instead of rates. Both make the first email look stronger than it is.

Out-of-office replies pile onto the first email. A typical one arrives about 25 seconds after the email lands, and 55% of our auto-replies fired on the opener. In this mostly summer window, about three in four people who replied at all sent only an auto-reply. Count those, and the opener gets a lot of free credit. With robots included, our follow-ups drew only about half the opener's replies per email, between 0.46x and 0.57x.

Our own first-email share moved 11 points on that one choice. Here's what the published studies say:

| Source (checked October 2026) | Data | Counts out-of-office? | First email's share of replies | Per email after that | Recommended length |
|---|---|---|---|---|---|
| [BusySeed](https://v7.busyseed.com/research/where-cold-email-replies-actually-come-from) | 1,857 replies, Dec 2025 to Aug 2026 | Not stated. Counts "any inbound message" | 61.7% | Not measured | None given |
| [Instantly 2026 benchmark](https://instantly.ai/cold-email-benchmark-report-2026) | "Billions" of interactions, 2025 | Not stated. Counts "all replies received" | 58% | Not published per step | 4-7 emails, 3-4 days apart |
| [Saleshandy](https://saleshandy.com/blog/cold-email-statistics/) | 30.3M emails, Jul to Oct 2026 | Not stated. Tags them as their own class | 43%, labeled both all and positive replies | 0.80% at follow-up 1, down to 0.37% at follow-up 5+ | 4-6 follow-ups over 20-21 days |
| [Belkins 2026 study](https://belkins.io/blog/sales-follow-up-statistics) | 7.5M emails, 2025 | Not stated | 41.4% | Halves by email 3, then flat | 3-5 emails |
| [lemlist](https://lemlist.com/blog/how-many-cold-email-follow-ups) | "Millions" of campaigns | Not stated | Not published | About a third of email 1 by email 5, flat to email 14 | 4-9 follow-ups |
| [Hunter](https://hunter.io/the-state-of-cold-email) | 31M emails, 2025 | Not stated | Not published | Not published | 3 emails total |
| [Woodpecker](https://woodpecker.co/blog/follow-up-statistics/) | Not stated | Not stated | Not published | First follow-up about 40% above the opener | 2-3 follow-ups |
| Ken, this article | Contacts first emailed Feb to Aug 2026, 17 workspaces | Shown both ways | 41% humans only, 52% with auto-replies | Email 4 at 0.99x email 1, humans only | 4 emails, 3-6 days apart |

None of these studies says outright whether out-of-office replies are in its count. But look at how the two highest ones define a reply. BusySeed counts "any inbound message a recipient sent back in the thread, regardless of sentiment." Instantly divides "all replies received" by emails sent. Both cast the widest possible net, and both sit closer to our robots-included 52% than to our human-only 41%.

The two in the low 40s are murkier. Belkins doesn't define a reply. Saleshandy calls its 57% follow-up share "all replies" in one place and "positive replies" in another, and a positive reply can't be a robot. I can't prove the definition explains the gap. But it's the first thing I'd ask about any study that says follow-ups don't matter.

![Bar chart: published studies credit the first email with 41% to 62% of replies, and Ken's own share moves from 52% with auto-replies to 41% without them](/images/blog/cold-email-follow-up-reply-rate-by-step-2.svg)
*Sources: Ken sending data, contacts first emailed Feb 25 to Aug 3, 2026, 17 workspaces; BusySeed, Instantly, Saleshandy and Belkins as published, checked October 2026. Lines on the Ken and BusySeed bars show 95% confidence intervals.*

To judge your own sequence, compare each step to your own email 1, per email sent, humans only. We go deeper on [why every cold email reply rate benchmark disagrees](https://ken.so/blog/cold-email-reply-rate).

## How many follow-up emails should you send?

Send three follow-ups, so four emails in total. Add a fifth only if it says something new.

The published advice runs from three emails in total at Hunter to ten at the top of lemlist's range. Put the six recommendations side by side and the middle of the pack is four to six emails. Our data says the low end leaves conversations behind. In campaigns built with four or more emails, here's how human replies piled up:

| Stop after | Share of human replies you'd have kept (95% CI) |
|---|---|
| Email 1 | 37% (33-42%) |
| Email 2 | 56% (52-61%) |
| Email 3 | 76% (72-80%) |
| Email 4 | 94% (91-96%) |

*Ken sending data, campaigns with 4+ steps, 10 workspaces, contacts first emailed Feb to Aug 2026. Shares are of the human replies those campaigns received.*

Stop at two emails and you give up 44% of the human replies those campaigns earned. Some of those people might have replied anyway. But people reply within hours of the email that moves them, so I'd count most of that 44% as real losses.

Where does it stop paying? Few of our sequences ran past four emails, so our data can't say. The published curves flatten rather than crash. lemlist's per-email rate settles near a third of email 1 from email 5 all the way to email 14. Belkins' sits near half from email 3 to email 6. Belkins also found that email 3 alone booked 35.6% of the meetings that came from email.

So a fifth email still gets replies. The real question is what it costs.

## Do more follow-ups cost you anything?

Extra follow-ups can cost you remove-me requests, and spam complaints that never show up as replies.

Per email sent, our third and later emails drew about as many human replies as the opener, at 0.90x with a range of 0.75 to 1.07. They drew about as many yeses, at 0.83x with a range of 0.56 to 1.23. But they drew twice the "remove me" replies, at 1.97x with a range of 1.34 to 2.89.

That rise came mostly from one large sender. Without it, the gap shrank to a small one, on too few replies to call. BusySeed saw the reverse, with 87% of its remove-me and hostile replies landing on the first email, though on only 31 replies. So I read our number as a warning, not a law. A late follow-up doesn't annoy people by default. A bad one annoys them fast.

The yeses held up. In every study I found that splits out positive replies, about half came after a follow-up. It was 47% at BusySeed, 52% at Ken (range 44% to 61%) and 57% at Saleshandy. Belkins found emails 3 to 5 booked 53.5% of its email meetings.

**A remove-me request is the polite version of a spam complaint.** Some people who want out don't write back. They click the spam button, which never shows up in your reply data. Google's [sender guidelines](https://support.google.com/a/answer/81126), checked October 2026, ask you to keep spam complaints below 0.1% and never reach 0.3%. That's one complaint per 1,000 emails. Extra follow-ups to people who've already ignored you three times are where that budget goes.

## How long should you wait between follow-ups?

Wait three to six days after the previous email.

We compared like with like, meaning the same sender, the same email number and the same send weekday. Follow-ups sent three to six days after the previous email drew 1.44x the human replies of follow-ups sent within three days, with a range of 1.14 to 1.81. Without our largest sender it was 1.31x, with a range of 0.98 to 1.76, so the effect shrinks but doesn't flip. Waiting more than six days didn't add anything we could measure, at 1.02x.

Timing gives a likely reason. A human reply lands a median of 1.8 hours after the email that prompted it, and 95% land within 72 hours. So each email needs about three days to finish collecting its replies. Send the next one sooner and you talk over the last one.

Seen from the first email, the typical replier answered 3.3 days in, right after email 2 went out at a median of 3.1 days. Most repliers don't answer the first email. They answer a nudge.

A simple schedule that fits is day 1, day 4, day 8 and day 12. That's four emails in under two weeks, close to Instantly's advice of three to four days between touches.

![Forest plot: follow-ups sent 3 to 6 days after the previous email drew 1.44 times the human replies of those sent within 3 days, and emails 3 and later drew about as many replies as the opener but 1.97 times the remove-me requests](/images/blog/cold-email-follow-up-reply-rate-by-step-3.svg)
*Source: Ken sending data, contacts first emailed Feb 25 to Aug 3, 2026, 17 workspaces. Human replies only. Lines show 95% confidence intervals. Hollow dots repeat the comparison without our largest sender.*

## What should each follow-up say?

Each follow-up should say something the last email didn't. That's the whole rule. Here's how I'd apply it.

1. **Make email 2 more than a bump.** It's the weakest email in our data, at 0.71x the opener, and it's where the drop starts in Belkins' and lemlist's numbers. My guess is that it's the email people put the least thought into, often a "just bumping this" two days later. Wait three days and add one useful thing.
2. **Keep follow-ups in the same thread.** Instantly found a step 2 written to feel like a reply beat formal follow-ups by about 30%. About 97% of our follow-ups went out under the first email's subject line, so we can't test new-thread follow-ups ourselves.
3. **Give emails 3 and 4 a new angle.** Try a different problem, a short proof point, or a smaller ask. Saleshandy says its strongest follow-ups "introduced a new insight, use case, or conversation angle rather than repeating the original message."
4. **Switch channels after the last email.** Belkins goes further. When three emails got no reply, it suggests a LinkedIn connection request before any fourth email. Our data says email 4 still earns its place, but the instinct is sound. We run [LinkedIn and cold email as one sequence](https://ken.so/blog/linkedin-vs-cold-email) for that reason.

At Ken, human copywriters build the frameworks, AI fills them with real data about each prospect, an AI quality check reviews every email, and a person approves before anything sends. That's mostly so the fourth email reads like it was written for that person, not like a reminder.

## How we measured this

- **Data.** Contacts whose first email went out between February 25 and August 3, 2026, with replies captured to August 24, across 17 sending workspaces. We count every email those contacts got in that window.
- **Steps and credit.** "Email 3" means the third email that contact got in that campaign. A reply counts toward the last email sent before it. Our sending platform's own step tag agreed with that rule on 98% of human replies. Each contact counts once, at their first reply.
- **Humans vs robots.** The "all replies" numbers include auto-replies and out-of-office. Everything else is human replies only.
- **Comparisons.** Ratios compare emails within the same workspace, so a sender with long sequences can't skew the later steps. Ranges are 95% confidence intervals.
- **Checks.** Counting only replies within 72 hours of each email, using only campaigns with four or more steps, and using only summer contacts all gave the same shape. One workspace holds 44% of human replies, and we re-ran the main claims without it.
- **Limits.** Most contacts were first emailed in June and July, when out-of-office replies are common. Positive and remove-me labels don't cover every reply. Email 5 and later is too thin to measure for humans.

## Frequently asked questions

### Should a follow-up be a reply in the same thread or a new email?

Inside one sequence, reply in the same thread, so your earlier email sits right there for context. Start a new thread only when you come back weeks later with a new reason.

### What reply rate should each follow-up get?

There's no fair outside number for a single step, because each benchmark divides by something different. Use your own first email as the bar. If one step sits far below the steps around it, rewrite that step before you cut it.

### What should I do with out-of-office replies during a sequence?

Pause the sequence and pick it up a day or two after their return date. Sending follow-ups into an empty inbox wastes your best emails, as we explain in [why out-of-office replies hurt your deliverability](https://ken.so/blog/out-of-office-replies-are-quietly-killing-your-cold-email-deliverability).

### Should the sequence stop when someone replies?

Yes, on any human reply, including a "no." A person who answered is now a conversation, not a sequence step. Our guide to [handling cold email replies](https://ken.so/blog/how-to-handle-cold-email-replies) covers what to send next.

### How long should a whole sequence take?

About two weeks for four emails at three-to-four-day gaps. Saleshandy's best-performing sequences ran 20 to 21 days with more steps. Past that, switch channels or stop until something changes.

## The short version

Send four emails, three to six days apart, and make every follow-up add something new. Per email, the late ones work about as well as the first. They just reach fewer people. Ken runs cold email and LinkedIn outreach for B2B companies, done for you, on infrastructure we own. If you want a sequence built this way, [your first campaign is free](https://ken.so/free-campaign).