There’s a lot of cold email copy advice online and most of it is templates. This is the shorter list: the cold email copywriting frameworks people who send for a living keep coming back to, and the two I’ve tested on my own campaigns.
The list does most of the work
Jordan Crawford’s line, “the list is the message,” gets quoted more than any other by people who do this well. I’ve heard it from three different practitioners who don’t know each other.
What it means for the writing is that a good list lets the email be plain. If every person on the list runs a furniture showroom that buys wholesale, an email that says “we work with showrooms that buy wholesale” is already relevant. The work moved from the sentence to building the right list.
Give something before you ask for something
The other idea that keeps coming back is Crawford’s permissionless value proposition. Offer something the reader can use without replying, and let some of them ask for more.
One agency owner took it literally and built a free web tool for his prospects. The tool was the outreach. I haven’t run that pattern myself, so I can’t give you numbers for it, but it’s the clearest example I’ve seen of the idea.
Short, with one question
Two people who send at volume gave me their lengths without being asked. One writes 50 to 75 words most of the time, with the link in the signature. The other rarely sends anything over 70 words anymore, because longer emails from unknown senders had become harder to land. He thinks of length as a deliverability decision as much as a writing one.
Mine sit in that range. I haven’t tested longer against shorter on the same list, so I can’t tell you where the line is for your audience.
Don’t give the whole offer in the first email
An agency owner put it this way: never put the full offer in email one. If you do, the reader consumes it and moves on. The first email exists to find out whether the person cares about the problem. The offer comes in the reply.
That matches what I do. It’s his experience and mine, not a law.
Personalisation: test it before you pay for it
I assumed personalisation worked. Then I tested it.
One campaign ran two versions of the same email. One opened “Hi there.” The other opened with the person’s first name. “Hi there” produced four leads. The first name version produced none.
One campaign proves nothing general. What it did was stop me assuming. In a teardown I sat in on, an AI written personalised line got 2.6% replies. The client’s own version, written off something the prospect had posted on LinkedIn that week, got 4.4%. Same list, same offer.
Two data points, pointing the same way. A name in the greeting did nothing for me. A line that showed the sender had actually read something did better for them. Run the same test on your own list before you pay for enrichment.
Social proof, and the wrong kind
Cold email needs proof more than a LinkedIn message does. On LinkedIn your profile is the proof. In an inbox the sending domain is a stranger.
The wrong proof hurts, though. In that same teardown, a campaign to boutique retailers had named two large national brands as case studies. The boutique owners didn’t see themselves in it and the campaign died. The fix was to remove the names and keep the outcome.
My own proof is numbers rather than logos. 122 replies across four campaigns to one audience. 13 interested replies from 60 emails on another. Whether numbers travel better than logos for your audience is worth a test.
Words I stopped using
Two words I banned for myself are “premium” and “exclusive,” after they kept showing up on the spam trigger lists the sending platforms publish. Those lists are worth reading once. They change, and a plain email from a real person avoids most of the words on them without trying.
How to test copy without guessing
The method that came up most often: write five to seven different angles, not variations of one, send each to 800 to 1,000 contacts, keep the one that gets answered, drop the rest. One angle on 200 contacts tells you nothing, because the numbers are too small to separate a good email from a lucky week.
I do something close to that with fewer angles, because my clients’ lists are usually smaller.
Questions we get
How long should a cold email be? The figures I’ve heard from people sending at volume are 50 to 75 words and under 70. Mine are in that range.
Does personalisation work? Some of it. Test “Hi there” against the first name before you pay for enrichment, then test one real, specific line against no line at all, on your own list.
What reply rate is good? Depends on who you’re writing to. There’s a breakdown by audience here.
Should I use AI to write it? To draft, fine. To send unread, no. The 2.6% email above was AI written. The 4.4% one was written by a person who had read the prospect’s post.
