A cold email gets a reply when the person reading it can say yes to something small without taking any risk. That is the whole mechanism. Every step below exists to build a message that offers exactly that, written in the order I write one.
The part most people get backwards is what the first message is for. Its whole job is to earn the next sentence. Explaining your company, proving you are good and getting anyone to buy anything all come later, if they come at all.
A vendor benchmark is useful here as a floor rather than a goal. Instantly's 2026 cold email report gives the average reply rate across the market as about 3.4%, and since that comes from a sending platform's own campaign data, read it as directional. Take it at face value for a moment anyway. Roughly ninety-six people out of a hundred ignore the average cold email. The goal is to be the one message in that inbox that asks for something a busy person can reasonably give.
1. Decide what the email is selling before you write a line
Your first message sells the meeting. It does not sell your service.
That single decision fixes more cold email than any rewrite. The service is complicated, it costs money, it needs a decision from more than one person, and none of that can happen inside an email. A short conversation is cheap, reversible, and one person can agree to it alone.
So write the email that sells fifteen minutes. Not the case studies, not the feature list, not the three packages. When founders send me a sequence to look at, I read the first message and usually stop there, because the first message has already decided the outcome of the other four.
A useful test: read your draft and ask what the reader is being asked to believe. If the answer is "that this company is excellent", the email is selling the service. If the answer is "that a short call would not be a waste of time", it is selling the meeting.
2. Earn the first line with a reason you are there
Pitching in the opening message is the fastest way to be ignored. The reader has seen that exact move fifty times this quarter, and the pattern is recognisable inside one line.
What replaces it is a reason. Something true about that company, that role or that situation which explains why this email landed in front of them and not someone else. A compliment about their "impressive growth" does not count.
This is also where most personalization budgets get wasted. The relevance a reader wants is about their situation, not about their LinkedIn banner. Mentioning that you both went to the same university counts as personalization, and it still tells the reader nothing about why you wrote.
3. Say one thing, and say it in plain words
One claim. One audience. One problem you can name the way your reader would name it.
Short, human emails beat polished templates, and they beat them for an unglamorous reason: a template has to work for everyone on the list, so it says nothing specific to anyone on it. The more people a sentence could apply to, the less weight it carries with each of them.
Write the way you would type it to someone you already know. Contractions, ordinary vocabulary, no throat-clearing. If a phrase in your draft would never come out of your mouth on a call, take it out. The problem gets worse when a language model writes the draft, which is why so much outbound now reads like a machine.
And if you do not have results to point to yet, do not borrow any. There are other kinds of proof, and what to put in the email when you have no case studies is a separate problem with its own answer.
4. Make the ask answerable in one line
The last sentence is the only one with a job the reader can act on, so write it first and build the rest of the email up to it.
Two rules cover almost every case. Ask for one thing, not two. And make the reply short enough to send from a phone while walking between meetings.
"Would it make sense to talk for fifteen minutes next week?" is answerable. "Let me know a few times that work for you and I will send over a deck plus some case studies for your team to review" is homework. Whether the first message should ask for the call directly at all, or ask a question instead, is its own decision, and I use a simple rule to pick between them.
5. Write the subject line last
The subject line is the promise the body has to keep, so it cannot be written before the body exists.
Written last, it is a summary. Written first, it becomes a hook you then have to justify, which is how people end up with clever subject lines attached to emails that disappoint the person who opened them. That trade is worse than it looks, because the disappointment is what teaches the reader to skip your next one. Subject lines are a separate topic, including which of them stopped working.
6. Cut before you send
Every draft I write is too long, and so is yours. The first pass after writing is always a deletion pass.
I will not repeat the method here because it already has its own post: the cuts I make to cold email copy, line by line, with what goes and why. Run it on your draft before anything else.
One thing that post does not cover: send the cut version to your own inbox and read it on your phone before you read it anywhere else. A draft that looked tight in a document turns into four grey blocks on a small screen, and that is the version your prospect gets.
7. Plan the next four messages before you send the first
One message is a lottery ticket. A first email that gets no reply tells you nothing at all, because a busy inbox goes quiet by default, and that quiet carries no information about your offer.
Plan five or six touches across channels before the first one goes out, with real time between them. In our campaigns the order matters too: a LinkedIn touch first, so your name has been seen somewhere before it arrives in an inbox. The structure of a follow-up sequence that works covers the spacing and how many messages is too many.
What none of this fixes
Neither a bad list nor a weak sender responds to better writing, though both disguise themselves as one.
If the people receiving your excellent email do not have the problem you solve, the writing is irrelevant, which is why the list decides more than the email does. And the same message sends at very different rates depending on who signs it: I have seen the same email return 2% at one company and 7.5% at another, with nothing changed but the sender.
So before you rewrite anything for the fourth time, check those two. They are slower to fix and they decide more.
Open the last cold email you sent. Find the sentence that asks for something. If that sentence asks for thirty minutes and gives a reason that would apply to every other company on your list, you already know which part to rewrite first.