How to write a cold email: structure, format, and best practices

A cold email is the first message you send to someone who has no prior relationship with you and hasn't asked to hear from you. Writing one that gets a reply is less about clever tricks and more about a handful of decisions: what you open with, how long the email runs, what you ask for, and how you close it.

This guide covers what a cold email actually is, a useful structure for drafting one, a step-by-step writing process, and what selected benchmark reports say about length, openers, and closings. It draws mainly on vendor-published 2025/2026 datasets from Instantly, Hunter.io, Gong, Woodpecker, and Boomerang. These sources use different populations, definitions, and methods; some cross-reference the same underlying benchmarks, so the figures are directional rather than universal averages.

Key takeaways

• A cold email is a first message to someone with no prior relationship with you, sent with a specific, relevant reason. It is not automatically spam, and it is not automatically compliant either; applicable rules still apply.
• There is no universal reply-rate word count: Hunter reports strong bands at 61 to 80 and 181 to 200 words, while Gong reports that 100 words or fewer and 3 to 4 sentences perform best in its dataset (Hunter.io, 2026; Gong, 2025/2026).
• Hunter reports higher open and reply rates for messages with two custom attributes, but open rates are noisy and all of these figures are vendor benchmarks rather than guarantees (Hunter.io, 2026).
• A single, low-friction ask outperforms a hard pitch. Gong found that pitch-heavy language can cut reply rates by up to 57% (Gong, 2025).
• Follow-ups are worth testing, but cadence is not universal. Instantly reports that 42% of replies came from follow-ups, while Woodpecker repeats and cross-references that figure alongside other sources; treat 4 to 7 touchpoints as a test range, not an independent consensus (Instantly, 2026; Woodpecker, 2026).

A cold email is a first message sent to someone with no prior relationship, no referral connecting you, and no previous interaction, aimed at starting a specific conversation. The word "cold" describes the relationship, not the tone. A cold email can read as warm and specific and still be cold email, because the recipient has never heard from you before.

That's the meaning behind terms like "cold email meaning" or "what is a cold email": it's a relationship category, not a quality judgment. A cold email can be well-researched and genuinely useful to the recipient, or it can be a mass-blasted template. Both are cold email in the technical sense, but only one of them tends to get a reply.

Cold email is a relationship and outreach category, not a legal category. A relevant, targeted message may be less disruptive than a bulk message, but personalization does not automatically make it compliant. Rules such as the FTC's CAN-SPAM requirements in the U.S. and GDPR, ePrivacy, and national requirements in the EU can still apply depending on the recipient, channel, audience, and jurisdiction. For a broader look at how cold email fits alongside LinkedIn and cold calling as a category, see what cold outreach means and how the channels compare.

The exact rules depend on where the sender and recipient are located, what the message promotes, and which channel you use. For U.S. commercial email, start with the FTC's CAN-SPAM guidance. In the EU, direct marketing involving personal data can require GDPR transparency and a right to object, while ePrivacy and national rules may add restrictions (European Commission guidance). Treat these as a safety checklist, not a substitute for legal advice:

• Identify the sender accurately and use a subject line that reflects the message.
• Include the disclosures, business address, and opt-out mechanism required for the message and jurisdiction. Keep suppression records and do not continue after an objection.
• Use a business-relevant, non-sensitive detail. Do not use health, financial, political, employment, family, or other sensitive information merely because it is publicly visible.
• Check the rules that apply to email, social messages, and phone outreach separately. A rule that applies to U.S. commercial email does not automatically authorize LinkedIn messages or calls in another country.

A useful cold email can be built from six parts, although a good message may combine or omit some of them depending on the audience and ask. Skipping or padding the wrong part is usually where a draft goes wrong.

1. Subject line. Specific and honest about the message. Two to four words can be a useful test for opens, while other reply-rate datasets favor roughly 5 to 6 words or 36 to 50 characters. Do not treat one length as universal; see the data on which cold email subject lines actually get opened.
2. Opener. One sentence that proves you did the research. No greeting-and-introduction paragraph before it.
3. Value or context. One or two sentences connecting the opener to a reason the recipient should care, without listing every feature or credential you have.
4. Proof, if it fits. A specific result, number, or example relevant to a business like theirs. Optional, and only if it's short.
5. Call to action. One clear, low-friction ask. Not two asks, not a vague "let me know if you're interested."
6. Sign-off. A short, warm closing line and your name. No lengthy signature block.

That's a practical cold email format, in order. Many underperforming drafts either skip the opener and jump straight to the pitch, or cram proof, credentials, and a case study into a message that would be clearer if it were shorter.

Here's the process that produces the structure above, in the order it actually gets written.

1. Pick one recipient and one reason. Before writing a word, know who this specific person is and why this specific email is relevant to them right now. If the reason would apply equally to a hundred other companies, it's not specific enough yet.
2. Find one business-relevant detail that supports the reason for writing. A product launch, a change visible on the company's site, or another non-sensitive business signal can become the opener. Public availability does not make sensitive personal information appropriate to use.
3. Write the subject line last, not first. It's easier to write a short, specific subject line once you already know what the email says.
4. Draft the body in three sentences. Opener, then value, then the ask. Add proof only if there's a genuinely relevant, short example. Resist the urge to explain everything about what you do.
5. Cut it down. Read it back and remove any sentence that doesn't change whether the recipient replies. This is usually where an email goes from 200 words to 90.
6. Check the ask is singular and low-friction. One question, ideally answerable with a yes or no, not a request to "hop on a call to discuss further" with no specifics.
7. Send from an authenticated domain and monitor reputation. SPF, DKIM, and DMARC help receiving providers verify the sender, but authentication and domain history do not guarantee inbox placement. Requirements vary by provider and volume; unwanted mail, high complaint rates, poor list quality, and deceptive content can still result in spam placement or rejection. Writing and deliverability are related but separate problems.

This is the practical answer to "how to cold email" or "how to send a cold email": most of the work happens in steps 1 through 6, while step 7 influences whether any of that writing gets seen.

The opener gives the recipient the context for the email, and many drafts weaken it by starting with a greeting and a self-introduction instead of the reason for writing.

Hunter's 2026 report, based on 31 million emails sent by its users in 2025, reports a 40.2% open rate for subject lines with two custom attributes versus 35.4% with one, and a 5.6% reply rate for body copy with two custom attributes versus 3.6% without personalization (Hunter.io, 2026). Open rates can be distorted by privacy protections and automated tracking, so treat them as a noisy directional signal. Personalization should also be business-relevant and non-sensitive; a first-name merge tag is not the same as useful research.

Compare the two approaches:

• Generic: "Hi there, I hope this email finds you well. My name is [name] and I work with [company] helping businesses like yours grow."
• Specific: "Saw [company] just opened a second location in [city]. Curious how you're handling scheduling across both sites."

The second one signals research in the first five words. The first one signals a template, and most recipients can tell within a sentence which one they're reading.

Shorter can help, but the exact number depends on the dataset and audience. Instantly's 2026 benchmark reports that its best-performing campaigns use emails under 80 words, while defining overall reply rate across initial and follow-up messages (Instantly, 2026). Hunter's analysis of 31 million emails found strong reply-rate bands at 61 to 80 words and again around 181 to 200 words, so the relationship is not a straight line where shorter is always better (Hunter.io, 2026). Gong's analysis of 28 million cold emails reports that emails of 100 words or fewer and 3 to 4 sentences perform best in its dataset (Gong, 2025/2026).

An older Boomerang analysis of 40 million general emails found response rates above 50% across the 50 to 125 word range, peaking around 51% at 75 to 100 words, then dropping at much longer lengths (Boomerang, 2016). It was not limited to cold email and is now dated, so use it only as historical context rather than evidence that a particular cold-email length will work.

The practical takeaway: write to around 100 words as a default, and don't pad an email to hit a target length. If the message genuinely needs 150 words to make sense, that's fine. If it's 250 words because you tried to cover three points, cut two of them.

The close does two jobs: it asks for something specific, and it signs off in a way that doesn't undercut the rest of the email.

The ask. End with one clear, low-friction request, ideally something the recipient can answer with a yes, a no, or a quick reply, rather than an open-ended "let me know your thoughts" or a request for a long call with no agenda. Gong's research on 28 million cold emails found that language which reads as a pitch or a hard sell reduces reply rates by up to 57%, while the reps who get 4.2 times more replies than average tend to use simpler, lower-friction asks (Gong, 2025). A specific yes/no question, such as "Worth a 10-minute call next week?", gives the recipient an easy way to respond without committing to much.

The sign-off line. An older Boomerang analysis of 350,000 general email threads found that gratitude-based closings correlated with higher response rates: "Thanks in advance" saw 65.7%, "Thanks" 63%, and "Thank you" 57.9%, compared with 51.2% for "Best" and a 47.5% baseline (Boomerang, 2017). The study was not limited to cold email and does not prove that a sign-off causes a reply. Use it as a directional historical observation, not a hard rule.

A short list of what the current data supports, beyond structure and length:

• Personalize the subject line and the opener, not just the greeting. A first-name merge tag isn't personalization anymore; a specific, checkable detail is.
• Send one clear ask, not two. An email that asks for a call and offers a resource and requests a reply gives the recipient three ways to do nothing.
• Plan for follow-ups from the start, but test the cadence. Instantly reports that 42% of replies came from follow-ups and describes 4 to 7 touchpoints as a test range. Woodpecker repeats and cross-references those figures alongside other sources rather than providing independent confirmation (Instantly, 2026; Woodpecker, 2026). Woodpecker also reports a 65.8% lift from adding one follow-up in its dataset; treat that as source-specific, not guaranteed.
• Mind send timing, loosely. Instantly's 2026 report found Wednesday as the peak-engagement send day and Friday as the weakest, with a surge in auto-replies rather than real engagement (Instantly, 2026). Treat this as a starting point to test against your own list rather than a fixed rule.
• Fix deliverability before you fix wording. A well-written email from an unauthenticated domain is more likely to face delivery problems, but authentication and warm-up are not guarantees. Monitor provider guidance, complaint rates, and list quality. For the mechanics of subject lines specifically and what triggers spam filters, see the data-backed breakdown of cold email subject lines.
• Don't let AI drafting replace the research step. AI tools can speed up drafting, but the research and personalization step is what separates a reply-worthy email from a templated one; see why a lot of AI-generated cold outreach is converting below 1% in 2026 for what tends to go wrong when that step gets skipped.

The hardest part of cold email isn't writing one good email, it's writing dozens of specific, researched emails without the process collapsing into a template with variables swapped in. That usually comes down to two things: having a genuinely researched, verified contact list, and sending from infrastructure that is authenticated and monitored. Authentication and cautious ramp-up can support delivery, but neither guarantees inbox placement or makes unwanted email appropriate.

For agencies and small teams sending cold email into local business niches such as plumbers, dentists, clinics, and salons, Anomalead builds prospect lists from Google Maps searches by category and city, finds the actual owner's or decision-maker's email instead of a generic info@ address, and sends from domains and mailboxes that are already authenticated and warmed. That combination is what makes the research-driven opener and single low-friction ask described above realistic to do at scale, rather than only on the first ten emails before the process reverts to a template. For a broader look at building the list and pipeline this feeds into, see what B2B lead generation actually involves and the best lead generation tools for small businesses.

A cold email is the first message you send to someone who has no prior relationship with you and hasn't asked to hear from you. It is a relationship and outreach category, not a legal status. A well-targeted email may be more relevant than a bulk message, but U.S. commercial-email rules and EU GDPR, ePrivacy, and national rules can still apply.

Start with one business-relevant, checkable detail about the recipient or their company, without using sensitive personal information. Use an honest, specific subject line, open with the reason for writing, state the value in one or two sentences, and close with a single low-friction ask. Start around 100 words, then test shorter and longer versions for your audience. Authenticate the sending domain and follow the applicable sender and opt-out rules; authentication alone does not guarantee inbox placement.

There is no universal word count. Hunter's analysis found strong reply-rate bands at 61 to 80 words and again around 181 to 200 words, while Gong reports that emails of 100 words or fewer and 3 to 4 sentences perform best in its dataset. Start around 100 words, keep every sentence useful, and test length against your own audience rather than treating 50 to 125 words as a rule.

Open with a business-relevant detail that explains why you are writing, such as a product launch or a change visible on the company's website. Avoid sensitive personal details and do not assume that publicly available information can be used without limits. Hunter's dataset reports higher open and reply rates for messages with two custom attributes, but open rates are noisy and the result is a benchmark, not a guarantee. A generic opener like "Hope you're doing well" often wastes the sentence that should be doing the most work.

End with a single, low-friction, specific ask, such as a yes/no question about a short call, rather than multiple asks or vague next steps. Gong reports that language which reads as a pitch or a hard sell can reduce reply rates by up to 57%. For the sign-off itself, an older Boomerang analysis of 350,000 general email threads found that gratitude-based closings correlated with higher response rates, but it was not limited to cold email and should be treated as historical context rather than a current benchmark.

A cold email is the first message to someone with no prior relationship with you. A six-part structure is a useful drafting framework: a specific subject line, a researched opener, a brief value statement, optional proof, a single low-friction ask, and a short sign-off. Current benchmark reports disagree on the best length and cadence, so start around 100 words, test against your own audience, and use follow-ups only when each one adds relevant value and respects opt-outs.

None of that works if the email never reaches the inbox, so treat deliverability, list quality, sender authentication, and complaint monitoring as part of the process. Authentication can support delivery, but it cannot make unwanted or non-compliant email appropriate.