A cold email gets replies when four things are right: it goes to a person who plausibly needs what you do, it arrives in the inbox instead of spam, it is short and specific about why you wrote, and it asks for something small. Most of the improvement comes from the list and the deliverability, not from clever wording. This guide covers each part in order, shows what published data says and does not say, explains how to test without fooling yourself, and points out the legal and reputational limits. Where a figure is quoted, the source and its limits are stated, because almost all public cold email benchmarks come from vendors or from outreach that is not quite like yours.
Start with the list, not the wording
A well-written email to the wrong person fails, and an average email to the right person often works. Before writing anything, define who you are writing to and why they might care.
- A clear target. Role, company type and size, and a visible reason they might need you now, such as a recent hire, a new product or a problem you can see from outside.
- A real, verified address. Bounces damage your sending reputation. Remove addresses that are guesses.
- A reason you can state in one line. If you cannot say why you picked this person, the recipient will not guess it.
Google's Gmail guidelines say to send only to people who want to get messages from you and not to buy email addresses from other companies. Cold outreach sits close to that line, so it makes the case for small, relevant, researched lists, not large purchased ones.
The anatomy of a short cold email
A workable structure has five parts, and each is one or two sentences.
| Part | Job | Example wording |
|---|---|---|
| Subject | Look like a normal message from a person | "Quick question about [their topic]" |
| Opener | Say why you chose them | "I saw that [company] just opened a second office." |
| Value line | Say what you do and for whom, in plain words | "We help firms like yours cut the hours spent on client onboarding." |
| Ask | One small, easy question | "Is that something you are dealing with at the moment?" |
| Sign-off and opt-out | Identify yourself and make leaving easy | Name, company, and "If this is not relevant, tell me and I will not write again." |
A full example, with placeholders:
Subject: Question about onboarding at [Company]
Hi [First name],
I saw that [Company] recently opened a second office, which usually means more client set-up work for the same team.
We help service firms cut the time spent on client onboarding by automating the repetitive steps, such as forms, folders and welcome emails.
Is onboarding something that is eating time on your side right now?
If not, no problem and I will not follow up again.
[Name], [Company]
Notice what is missing: no long company history, no list of services, no attachment and no request for a 30-minute call.
What published data says, and what it does not
Three studies are widely quoted. All have limits.
- Length and style. Boomerang's 2016 analysis of over 40 million tracked emails found the best response rates for messages of 50 to 125 words, subject lines of three to four words, writing at about a third-grade reading level, and one to three questions. These were everyday emails that users asked the tool to track, not sales outreach, and the findings are correlations from a vendor's blog post. Use them as a nudge toward brevity and plain language.
- The ask. Gong's analysis of 304,174 emails, measuring a meeting booked within 10 days, reported that an interest-based call to action performed best in cold email, ahead of asking for a meeting directly. The page does not show the cold email rates in text, and it comes from a sales software vendor, so treat it as a reason to test a softer ask.
- Follow-ups. Backlinko's study of 12 million outreach emails found an overall response rate of 8.5 percent, that one follow-up increased replies by 65.8 percent, and that sequences of three or more messages did best. Those emails were mostly outreach to website owners, so your reply rate will differ.
The honest summary is that shorter, plainer, softer and followed-up tends to beat long, formal, hard-selling and single-shot. The exact numbers for your market you will only learn by testing.
Choose the ask
The ask decides how much effort a reply takes.
| Type of ask | Example | Trade-off |
|---|---|---|
| Interest | "Is this relevant to you?" | Easy to answer and tends to draw more positive responses, but a reply is not yet a meeting |
| Open meeting request | "Would you be open to a short call?" | Higher effort, fewer replies, closer to the goal |
| Specific time | "Are you free Thursday at 2 pm?" | Highest effort, can feel pushy to a stranger |
A common approach is to start with the interest ask, then move to a call once someone replies. Which one wins depends on your audience, so test them against each other.
Follow up with something new
A single email is easy to miss, and the Backlinko data above suggests follow-ups matter, so plan them before you send the first email.
- Message 1. The short email above.
- Message 2, a few days later. A new angle, such as a relevant example or a one-line result a similar business saw. Do not write "just checking in".
- Message 3, about a week after that. A final short note that closes the loop politely, and gives them an easy way to say no.
These gaps are a starting point from common practice, not measured optima. Stop the sequence when someone replies, whatever they say, and honour every opt-out.
Protect deliverability
An email nobody sees gets no replies. Gmail's sender guidelines set a baseline: all senders to personal Gmail accounts need SPF or DKIM authentication, valid reverse DNS and TLS, and a spam rate in Postmaster Tools below 0.3 percent, with Google recommending staying below 0.10 percent. Senders of more than 5,000 messages a day need SPF, DKIM and DMARC and one-click unsubscribe for marketing mail. Small cold email programmes sit below that 5,000 threshold, but the logic of the rules, which is to authenticate, avoid spam complaints and make leaving easy, applies to everyone.
Practical habits that follow from it:
- Set up authentication (SPF, DKIM and DMARC) on the domain you send from.
- Keep volumes modest per mailbox, and grow them slowly.
- Verify addresses before sending, and remove bounces.
- Watch complaints and bounces, and pause if they rise.
- Consider a separate sending domain for cold outreach, so that a problem does not affect the domain you use for client email. This is common practice with a cost: it needs setting up and warming up properly.
Know the rules
The law depends on the country and on whether the recipient is a company or an individual. In the UK, the ICO says the PECR email marketing rule does not apply to corporate subscribers, but it does restrict marketing emails to individual subscribers such as sole traders, and the ICO says you should honour opt-outs from corporate subscribers anyway. Data protection law applies separately, and other countries have their own rules. Check the rules where you and your recipients are based, identify yourself honestly and give a clear way to opt out. This is general information, not legal advice.
Test without fooling yourself
Small samples are noisy. If one version of an email gets 3 replies from 100 sends and another gets 5 from 100, that looks like a win, but the plausible range for 3 in 100 is roughly 1 to 8.5 percent and for 5 in 100 roughly 2 to 11 percent, and they overlap heavily. Even at 500 sends per version, 15 replies (3 percent) and 25 replies (5 percent) have ranges of about 1.8 to 4.9 and 3.4 to 7.3 percent, which still overlap. At 1,000 sends per version, 30 and 50 replies give about 2.1 to 4.3 and 3.8 to 6.5 percent, which only just separate. These ranges are 95 percent Wilson intervals.
What this means in practice:
- Change one thing at a time, such as the ask or the opener.
- Do not declare a winner after a few dozen sends.
- Track what matters downstream: positive replies and meetings, not only replies.
- Record bounces and complaints for each batch.
Common mistakes
- Writing to everyone. A small, relevant list beats a big, generic one.
- Starting with yourself. Open with why you picked them.
- Asking for too much. A 30-minute call is a big ask from a stranger.
- Skipping follow-ups, or sending empty ones. Add something new each time.
- Ignoring authentication. Missing SPF or DKIM hurts inbox placement.
- Hiding the opt-out. It is the right thing to do and it reduces spam complaints.
- Judging on tiny samples. Wait for enough sends before changing course.
Next step
Write your short version, send it to a small, well-researched list, and track replies, positive replies and bounces separately. If you would like an outside view of your outreach and how it fits with your other channels, you can book a free audit. To compare outreach with the other lead channels, read eight lead generation channels compared. For an example of outreach in a specific industry, see how to automate candidate sourcing without sounding like a bot, or browse the rest of the blog.