# How Long Should a Cold Email Be? Reply Rates by Word Count

> Aim for 50-125 words. In our own first emails, length stopped mattering once we compared emails inside the same campaign. The data, charted.

Nov 01, 2026 · Cristian Frunze · 11 min · Guide

Tags: how long should a cold email be, cold email length, cold email word count, cold email reply rate, cold email copywriting, cold email data

![A founder on a bench snips his trousers into shorts after watching a champion sprinter in short shorts win, as if shorts made runners fast.](/images/blog/cold-email-length-reply-rate-cover.webp)

Sort our first emails by length and longer looks better. Openers of 150+ words drew 1.35x the human replies of 100-149 word ones. Inside the same campaign, that edge shrinks to 0.97x.

**Aim for 50 to 125 words in a first cold email, then stop tuning length. The campaign sets the reply rate, and word count just tags along.**

That's the short version. Below, I line up nine published studies, show why they can't agree, and take our own first emails apart, campaign by campaign.

## What do the big studies say about cold email length?

Every study names a sweet spot. Put them side by side and the sweet spots barely overlap.

| Study (checked October 2026) | Data | "Best" length | What it measured |
|---|---|---|---|
| [Boomerang](https://blog.boomerangapp.com/2016/02/7-tips-for-getting-more-responses-to-your-emails-with-data/), 2016 | 40M+ emails users asked Boomerang to watch for replies, not only cold | 50-125 words | Replies to any kind of email |
| [Yang et al., SIGIR](https://www.microsoft.com/en-us/research/uploads/prod/2017/04/sigir17a.pdf), 2017 | 938,035 emails inside one tech company | Around 40 words | Internal work email, peer-reviewed |
| [Yesware](https://www.yesware.com/blog/analyzed-data-sales-templates/), 2020 | 262,518 sales templates | 75 words | Templates, not single sends |
| [Gong](https://www.gong.io/blog/cold-email-stats), 2021, updated 2026 | "Millions" of cold emails | 30-150 words beat 30 or fewer | Meetings booked |
| [Lavender](https://lavender.ai/blog/best-length-cold-email), 2023 | Own users, size not stated | 25-50 words | First emails |
| [Hunter](https://hunter.io/blog/cold-email-word-count), 2024 | 34M cold emails, 2022-2024 | 20-39 words | Average reply rate |
| [Sales.co](https://howlongshouldacoldemailbe.com/blog/optimal-cold-email-length-data/), 2026 | 3.1M cold emails, 2025 | 75-100 words | Reply rate |
| [Instantly](https://instantly.ai/cold-email-benchmark-report-2026), 2026 | "Billions" of interactions, 2025 | Under 80 words | What top senders do, no length buckets |
| [Smartlead](https://www.smartlead.ai/benchmarks/ideal-cold-email-length), 2026 | 850M+ emails, first half of 2026 | Under 50 words | Replies per contact, first emails only |

The "best" lengths run from 20 words to 150. The studies don't even agree on how much length matters. In the Sales.co data, 75-100 word emails replied at 3.8% and 300+ word emails at 0.8%, a 4.8x gap. In Hunter's data, the best bucket beat the worst by 1.3x, 4.5% against 3.4%, and the author wrote that "the difference isn't spectacular." Smartlead's middle looks like the mirror image of Sales.co's. Its 100-149 word emails were the weakest group at 1.03%, below both shorter and longer ones.

![Three large datasets show three different shapes for reply rate by first-email length](/images/blog/cold-email-length-reply-rate-1.svg)

*Sources: Sales.co (3.1M cold emails, 2025), Smartlead (850M+ first emails, H1 2026), Hunter (34M cold emails, 2022-2024), all checked October 2026. Each line is divided by its own average bucket, so only the shapes compare.*

One thing does hold up. Every dataset that split out very long emails, 300 to 500+ words, saw them do worse. That's Boomerang, the SIGIR paper and Sales.co. Below 300 words, the studies point in different directions.

## Why do cold email length studies disagree?

Cold email length studies disagree because they compare different emails, sent by different people, to different lists, with different offers. Length rides along with all of that.

The clearest example comes from the one peer-reviewed study on the list. Researchers at Microsoft and UMass Amherst found that work emails with long subject lines got far fewer replies. The reply rate fell from 11.47% for one-word subjects to 1.10% at 29 words. Then they read the emails. Many of the long subjects belonged to announcements and machine-generated bug reports, the kind of email nobody needs to answer. The length didn't cause the silence. The type of email did.

Cold email has the same problem:

- **Top senders write short, and they do everything else well too.** Instantly's report says its "elite performers average fewer than 80 words per first-touch email." The same report says those senders write hyper-relevant subject lines, ask for one thing and lead with the problem. That describes good senders. It doesn't show what happens when you cut your own email from 140 words to 80.
- **Volume and care travel with length.** Yesware found that "templates sent at a high volume consistently yielded lower reply rates," while personal, low-volume templates did best. A 40-word email to 50 hand-picked people and a 40-word email blasted to 50,000 are not the same email.
- **The authors know it.** Hunter's analyst wrote that his results don't mean "if you adjust the email copy you're currently using to match this word count range, your reply rate will improve."

**A word-count study mostly measures who writes short emails, not what short emails do.**

## What happens when you compare emails inside the same campaign?

When you compare emails inside the same campaign, the length effect disappears. That's the main finding from our data.

We looked at every first email we sent between February 25 and August 3, 2026, across 52 campaigns in 16 workspaces. A workspace is one company's account on our platform. We counted only replies from real people and left out out-of-office and other auto-replies. Word count is everything the recipient sees, greeting and signature included.

Then we compared longer and shorter emails three ways: all emails in one pile, inside the same workspace, and inside the same campaign. The numbers in parentheses are 95% confidence intervals, the range the true value very likely sits in.

| Comparison | All emails pooled | Same workspace | Same campaign |
|---|---|---|---|
| 100+ words vs under 100 | 1.44x (1.13-1.83) | 1.49x (1.04-2.15) | 0.99x (0.62-1.59) |
| 150+ words vs 100-149 | 1.35x (1.11-1.65) | 1.70x (1.38-2.09) | 0.97x (0.73-1.29) |

*Source: Kenso sending data, first emails sent Feb 25 to Aug 3, 2026, 52 campaigns in 16 workspaces, human replies only. Each ratio is the longer group's reply rate divided by the shorter group's.*

In the pile, longer wins. Inside one workspace, longer wins by even more. Inside one campaign, it's a tie.

![Pooled by length, human replies swing almost 3x; compared inside the same campaign, the line is flat](/images/blog/cold-email-length-reply-rate-2.svg)

*Source: Kenso sending data, first emails sent Feb 25 to Aug 3, 2026, 52 campaigns in 16 workspaces. Bars show 95% confidence intervals. Left: 10-word bins with 30+ human replies and 5,000+ first emails each. Right: each email's length minus its campaign's average, replies vs. that campaign's normal rate.*

The left panel gives the game away. Emails of 90-99 words replied at 2.22x the average. Emails of 120-129 words replied at 0.77x. That's almost a 3x swing for a 30-word difference. Real effects don't zig-zag like that. Templates do. Each bin holds mostly a handful of campaigns, and it inherits their results.

The right panel compares each email with its own campaign. An email 15+ words longer than its campaign's average drew 0.88x that campaign's normal reply rate. One 15+ words shorter drew 1.03x. Every interval includes 1.0, so none of these gaps stands out from random noise.

We checked two other explanations. Company size isn't it. Smaller companies do reply more, and our contacts at 1-10 person companies replied nearly twice as often as those at 51-200 person companies. But long and short campaigns went to a similar company-size mix, and adjusting for size made the pooled gap bigger, 1.62x, not smaller. The season isn't it either. Summer-only sends gave the same 0.97x inside campaigns.

The data has two limits. Inside one campaign, most emails sit within 10 or 15 words of the average, so it answers a narrow question. Does a personalized line that adds 10-20 words cost replies? We saw no sign of it, though a small cost could hide inside those intervals. It can't tell you whether an 80-word template beats a 160-word one for the same list. And only 4% of our first emails were under 75 words, so we can't say anything about very short emails.

## How long are real B2B cold emails?

Real B2B first emails in our data run longer than the advice. The median was 130 words counting greeting and signature, and the middle half sat between 115 and 142 words. Only 18% came in under 100.

![Each campaign has its own length. Campaign averages run from 74 to 161 words, while most emails inside a campaign sit within 10 words of its average](/images/blog/cold-email-length-reply-rate-3.svg)

*Source: Kenso sending data, first emails sent Feb 25 to Aug 3, 2026, the 33 campaigns with 1,000+ first emails each.*

The chart shows why pooled studies get fooled. Campaign averages run from 74 to 161 words. Inside a campaign, 6 in 10 emails sit within 10 words of the average. Across all 52 campaigns, 70% of the variation in length is between campaigns. Another 5% comes from different versions of the copy inside one campaign. The last quarter is what changes from one email to the next inside the same version, mostly personalization like company names and custom first lines.

So whoever writes the template sets the length, and after that it barely moves. At Kenso, human copywriters build the framework, AI fills it with real prospect data, an AI quality check reviews every email, and a human approves before the first send. I wrote more about that process in [how we write and QA cold email at scale](https://ken.so/blog/how-to-write-and-qa-cold-email-at-scale). The framework sets the length. Personalization moves it by a sentence, and in our data that sentence didn't measurably cost replies.

**Length is a fingerprint of the campaign, not a lever on the reply rate.**

## So what length should you write?

Write the shortest email that still makes the case, and expect it to land somewhere between 50 and 125 words. Here's how I'd do it:

1. **Start from jobs, not a word count.** Every sentence should do one of four jobs: why this person, why now, what you offer and the proof, and what to do next. Delete the sentences that do none of them.
2. **Spend extra words on them, not on you.** A specific line about the prospect's business is the best use of 15 words, and in our data it didn't cost replies. A paragraph about your company's history is the worst use.
3. **Don't pad and don't chop.** Don't stretch a sharp 60-word email to look serious. Don't cut a strong 140-word email to hit a number from a blog post.
4. **Watch the far end.** Past about 300 words, every dataset that measured it saw replies drop. If your first email is that long, it's probably two emails.

If your reply rate is bad, length is rarely the reason. Check the list and the offer first. Our guide on [what a good cold email reply rate looks like](https://ken.so/blog/cold-email-reply-rate) explains why published rates vary so much, and [how AI email personalization works](https://ken.so/blog/ai-email-personalization) covers the lines worth spending words on.

## How do you find the right length for your own list?

The only fair test is a split test inside one campaign. Split one list at random, keep the offer and the sender the same, and send two versions of different length. The catch is that length effects are small, so you need a lot of contacts to see one.

Here's how many contacts each version needs so a real lift shows up 8 times out of 10, at the usual 95% confidence:

| Your reply rate | To spot a 20% lift | To spot a 50% lift | To spot 2x |
|---|---|---|---|
| 1% | 42,700 | 7,750 | 2,320 |
| 2% | 21,100 | 3,830 | 1,140 |
| 4% | 10,300 | 1,860 | 550 |

*Standard sample-size math for comparing two reply rates, two-sided test. Calculated by Kenso.*

Most gaps between neighboring length buckets in the published studies are under 40%, and many are under 10%. Hunter's best bucket beat its own average by about 20%. So a fair length test needs thousands of contacts per version, often tens of thousands.

We went looking for that test in our own data and barely found one. Campaigns in our data ran 22 A/B splits on the first email with at least 300 contacts per version. In 16 of them, the longest and shortest versions sat within 10 words of each other, so they tested wording, not length. Only 5 differed by 15 words or more, and those drew too few human replies to name a winner. The typical version went to about 500 contacts. At a 2% reply rate, that's less than a seventh of what you need to spot even a 50% lift.

I don't think skipping length tests is a mistake. If you email 2,000 people a month at a 2% reply rate, a test that can catch a 20% lift takes about 21 months. Spend your tests on the list and the offer, where I'd bet on bigger differences. Our guide to [cold email A/B testing](https://ken.so/blog/cold-email-ab-testing) covers what to test first.

## Frequently asked questions

### Does the email signature count toward the word count?

In our numbers, yes, because the recipient sees it. Most published studies don't say what they counted, which is one more reason their sweet spots don't line up. The signature barely changes inside a campaign, so it doesn't skew our same-campaign comparison.

### Should follow-up emails be shorter than the first email?

The advice conflicts. Woodpecker [suggests 50-75 words](https://woodpecker.co/blog/how-long-should-a-cold-email-be-the-ultimate-guide/) for follow-ups, while [Lavender reports](https://lavender.ai/blog/best-length-cold-email) that Gong found follow-ups with 4+ sentences booked 15x more meetings than shorter ones. We didn't measure it, because our follow-ups quote the earlier email, which makes their word counts unreliable.

### Is 200 words too long for a cold email?

Not automatically, but the evidence is mixed. Smartlead's 200+ word bucket replied at 1.18%, above its 100-149 bucket at 1.03%, and Hunter's 200+ bucket sat at 3.7%, right at its average. Sales.co's 200-300 word emails replied at 1.4%, well below its 3.8% peak. The clear, repeated drop starts past 300 words.

### Does reading level matter more than length?

It might matter as much. Boomerang found emails written at a 3rd-grade reading level got 36% more responses than college-level ones, 53% against 39%. That's the same kind of pooled comparison as the length studies, so I'd treat it as a good habit rather than a law. Plain words cost you nothing.

### How long should a cold email subject line be?

Short. In Boomerang's data, subject lines of 3-4 words got the most responses, and response rates fell slowly as subjects got longer. As with body length, the subject matters less than who you send it to and why.

## The bottom line

Write 50 to 125 words, cut every sentence that doesn't do a job, then put your energy into the list and the offer, because that's where reply rates get decided. Kenso runs cold email and LinkedIn outreach for B2B companies, done for you, on infrastructure we own. If you want to see a campaign built that way, [your first campaign is free](https://ken.so/free-campaign).