Why Three Emails, Not Twelve
Ask ten agencies how long a cold sequence should be and you will hear everything from five to fourteen touches. After running campaigns for dozens of B2B clients, we have settled on three emails, and the decision was driven by data rather than taste.
The pattern in our numbers is consistent: the first email does most of the work, the second email produces a meaningful bump, the third catches stragglers, and everything after that generates more spam complaints than meetings. Long sequences do not just underperform; they actively damage sender reputation and burn leads you could have re-approached cleanly a quarter later with a new angle.
Three emails, roughly a week apart, then out. Here is what goes in them.
Email 1: The Only Email That Gets Written From Scratch
The first email carries the campaign, and it has a strict anatomy:
The personalized opener. Two lines: one observation about the prospect drawn from their actual background, and one sentence connecting who you are to their world. I covered how we generate these at scale in the AI personalization article; the short version is that they are built from real LinkedIn and website data, never invented.
The problem, in their language. One or two sentences naming the specific pain your ICP feels. Not your product category. The pain. "Most ops teams we talk to are still stitching this together in spreadsheets" lands; "we are an AI-powered workflow platform" does not.
The proof. One sentence, one number, one named or described customer. "We helped a 40-person logistics company cut quoting time from two days to twenty minutes." Specificity is what separates proof from noise.
The soft ask. End with an interest-based question, not a calendar demand. "Worth a look?" and "Open to seeing how that works?" consistently outperform "Do you have 15 minutes Tuesday?" A cold prospect is ready to express interest long before they are ready to commit time.
The whole email should be under 100 words. Every extra sentence measurably hurts reply rates. If you cannot say it in 100 words, you do not understand your offer well enough yet.
Emails 2 and 3: Replies, Not New Emails
The follow-ups send as replies in the same thread, so your first email gets pulled back to the top of the inbox with full context attached. This one mechanical detail roughly doubles the effectiveness of a follow-up, because the prospect never has to remember who you are.
Email 2, sent around day 4 to 6, adds one new piece of value. A different proof point, a relevant insight, a one-line case study. It never opens with "just bumping this" because a bump with no new information is an admission that the first email deserved to be ignored. Two or three sentences, then the same soft ask.
Email 3, sent around day 10 to 12, is the breakup. Short, warm, zero pressure. Acknowledge the timing probably is not right, leave the door open, and make it easy to say no. Breakup emails routinely pull replies from people who read both earlier emails and simply never got around to answering. Some of the best deals in our client campaigns started with a reply to email three.
The Rules That Protect Your Results
A few hard rules we hold every sequence to, learned the expensive way:
- No links and no images in email one. Both hurt deliverability, and a cold prospect will not click anyway. Send the deck after they reply.
- Plain text only. Designed HTML templates scream marketing blast and go straight to promotions or spam.
- Vary your copy. Sending the identical string thousands of times is a spam-filter fingerprint. We rotate synonyms and phrasings across sends so no two emails are byte-identical.
- Subject lines stay short and boring. Two to four lowercase-ish words that read like a colleague wrote them. Clever subject lines get opens from curiosity and replies from no one.
- One CTA per email. Two asks equals no ask.
Test One Variable at a Time
The three-email structure stays fixed; the content inside it is what you test. Run two versions of the problem statement against each other. Try a different proof point. Swap the soft ask. Because the structure is constant, you always know what caused a change in results.
And measure replies, not opens. Open tracking is unreliable now and opens never paid anyone's invoices. Positive reply rate is the number that predicts booked meetings, and booked meetings are the number that predicts revenue.
If you want a second opinion on a sequence you are running, send it over. I read a lot of these and I am happy to tell you what I would change.