How long a cold email should be

Anywhere between about twenty and two hundred words, and the reason that range is so wide is that the best-documented dataset on the question shows almost no difference across it.

You probably arrived here with a draft open and a word counter reading something over 125, because that is the number every page on this search result gives you. Before you cut, look at what happens to reply rates when people do not cut.

Hunter published the full word-count distribution for 34 million cold emails sent through its sequencing tool between 2022 and 2024.

The best band and the worst band are next to each other. Twenty to thirty-nine words replies at 4.5%, forty to fifty-nine at 3.4%, and everything from sixty words to over two hundred sits between 3.5% and 4.1%. Across a tenfold difference in length the entire spread is 1.1 percentage points, and it does not move in one direction.

That shape has a name in every other field. It is noise.

One thing Hunter does not publish, and this article should hold itself to the standard it holds everyone else to: the number of emails in each band. The 34 million is the total. If the longest band is thin and the shortest is enormous, the flatness is less solid than it looks. What can be checked is that the ordering is not monotonic in either direction, which no amount of uneven band sizes produces on its own.

Hunter's author, Ziemek Bućko, closes the piece this way: "If you can make your email shorter but retain the information, try it, and see if it performs better. Otherwise, don't worry too much about the word count of your cold email." Hunter sells cold email sequences. A flat curve is not a finding that helps them sell anything, which is a reason to take it more seriously rather than less.

What that table is counting, including the parts that cut against it

Reply rate here is replies received divided by emails sent. Hunter states plainly that the sentiment of those replies is not accounted for, so a "please remove me" counts the same as a "tell me more."

Two filters were applied before the averages were computed, and Hunter publishes both. Campaigns with fewer than fifty emails were removed. So were campaigns with a reply rate above 30% or below 0.5%, on the stated reasoning that the top ones often are not going to genuinely cold recipients and the bottom ones ignore basic practice.

That second filter is defensible and it costs something. It removes exactly the campaigns a founder most wants to study, the ones that worked unreasonably well. What is left is the broad middle of paying customers of a sending tool, which is a real population and not a random sample of cold email.

The four numbers you will see, and what each one is measuring

The rules on this search result disagree because they are counting different things and none of them says so on the page.

A reply rate that ignores sentiment, a reply rate that folds in responses to follow-ups, and a rate that counts only positive replies are not three answers to one question. They are three different questions.

The Instantly line deserves its own sentence because it is the one most often repeated as an instruction. Instantly says elite performers average fewer than eighty words. That is a description of who the top senders are. Those same senders have cleaner lists, warmer domains and better offers, and nobody has run the test where length is the only thing that changes. No source located anywhere in this research ran one.

Where that range comes from

It comes from a single blog post, published by Boomerang on 12 February 2016 and written by Alex Moore. Boomerang makes an email follow-up tool for Gmail, and the post analyzes emails its own customers had asked it to remind them about. The sentence everyone quotes, in full, is "The sweet spot for email length is between 50-125 words, all of which yielded response rates above 50%."

Two things about that post matter for your draft.

Read the second half of that sentence. Response rates above 50%, falling to 44% at 500 words, about 35% past 2,500 words, and 11% for an email with a subject line and no body. Cold email, in every dataset that publishes a denominator, runs between 0.5% and 5%. Hunter's whole table lives between 3.4% and 4.5%. Boomerang's population was people emailing colleagues, clients and, in the post's own worked examples, a local pizza place. Half of them replied because they already knew the sender.

And Boomerang never states how many emails the length analysis covered. The "40 million" figure attached to it everywhere is the number of emails Boomerang customers asked it to chase, not a sample that was studied.

One publisher, Artisan, does raise the problem in passing, noting that Boomerang never says whether those responses came from warm subscribers, cold prospects or internal email. That is the right question, and the base rates answer it.

The variable that does move the number

Gong, on a page reporting on more than 28 million cold emails, states that pitching reduces reply rates by as much as 57%. It attaches no separate sample to that figure and does not publish how it defines a reply, which is the same complaint this article makes of everyone else. Take the direction and the rough size, not the second digit.

Hold that against the length table. The entire spread of Hunter's word-count distribution is 1.1 percentage points on a base of about 4%. One content decision, whether the email pitches or not, moves the same metric by more than half. Gong also reports that its top reps get 4.2 times more replies than average reps, which is the same descriptive shape as Instantly's elite senders and carries the same caveat: it says who does well, not what to change.

Who you send to sits in the same category. Belkins, across 7,530,489 emails in 2025, found companies of zero to ten people replying at 0.72% against 0.22% at ten thousand and above, and founders replying at 0.57% against VPs at 0.32%. We worked through those numbers in who actually replies to cold email. Those are ratios of two and three to one, on the same metric where length delivers a ratio of about 1.3 to one at its most generous.

Length is the thing people optimize because it is the thing that is easy to count.

The rule nobody has data for

Several pages on this query give you word counts by seniority: shorter for executives, longer for practitioners, a tier for each. The numbers vary by page and none of them carries a source.

Gong has both variables. It analyzed more than a million executive sales cycles and reports how executives respond, and separately reports what happens to reply rates past a hundred words. It does not cross-tabulate the two, and no published dataset located anywhere in this research does either. Every seniority-tiered word-count rule in this search result is somebody's house style presented as data.

That does not make the advice wrong. It makes it advice, and you should treat it the way you treat any other opinion about your buyers, which is to test it against the ones you actually email.

What to do with the draft you have open

Write the email at whatever length says the specific thing you have to say, then cut the sentences that are not doing work. If that lands at 60 words, send 60 words. If the checkable detail that makes your email worth answering needs 130, send 130. The evidence does not penalise you for it.

What to cut first, in order, is everything Hunter's table does not measure and Gong's does. Cut the pitch, and the test for that is mechanical rather than aesthetic: read each sentence and ask whether it is about them or about you. A sentence naming their company, their role, their hiring, their product, or something they published is about them. A sentence naming your product, your funding, your customer count, your award, or what you help companies do is about you. Cut every sentence in the second group. Then cut the second ask. Then look at whether the person is right, because on the published numbers that decision is worth several times what the word count is worth.

If you want a rule to hold in your head, take Gong's rather than a word count: three to four sentences, no pitch. It is the only length guidance in this whole set that comes attached to a mechanism instead of a percentage.

Then send fewer, to people you have checked. Our piece on how many cold emails a day is actually safe covers the ceiling on the other side of that decision.

All figures in this article come from each publisher's own page: Hunter, Boomerang, Gong, Instantly, Lavender, Snov.io and Belkins, read on 3 September 2026. Where a number is attributed to a source that does not print it, this article says so rather than repeating it.