The Cold Email Framework: First Email, Sequence, Follow-Ups

Most cold email fails before it's read: opening with 'I' or 'We', pitching three features in 150 words, asking for a 30-minute call up front, then stopping after one email when the reply never comes. The fix isn't a better template. It's a first email built on real research, a sequence that presents the value from different angles, and follow-ups that earn the second look instead of just bumping the thread.

The First Email: 50-90 Words, 4 Parts

50-90 words is a forcing function, not a magic number. At 200 words you hedge and bury the ask. At 70, every sentence earns its place. Personalization isn't a first name, it's one real signal: a funding round, a job posting, their tech stack, a leadership hire. We pull this at scale with AI agents.

1. 01Hook: the specific signal you found, one sentence.
2. 02Problem: one pain point that signal implies. Don't solve it yet.
3. 03Proof: one concrete result for a similar company.
4. 04Ask: one low-friction question, answerable in five words, not 'are you free for a call?'

Subject line: short and specific beats clever. Lowercase reads as a real person, not a campaign. Skip 'free' and 'guarantee', both are spam-filter bait and they signal template before the prospect opens it.

The Sequence: 4 Steps, 14 Days

Length shrinks step to step, 50-90 words down to 15-30 by the breakup, and that shrinkage is itself a signal: a short email at Step 4 reads as someone with better things to do, which is the point. From Step 2 on, reply in the same thread rather than sending new emails. The recipient sees prior context without re-reading your pitch, and thread replies get better inbox placement than new cold threads because the thread already carries engagement signals. Set Step 2 onward to 'reply to previous step' in your sending platform, not 'send as new email.'

Follow-Ups: The Only Rule That Matters

Every follow-up must add new value, a new angle, proof point, framing, or piece of context, never the same message with a different opener. Most replies come from Step 2 or 3, not Step 1. Stopping after the first email is walking away from most of the pipeline.

• Step 2 (Day 4): reframe the value prop, add one named data point, or address a common objection directly. Never 'just following up.'
• Step 3 (Day 7): switch the pain point entirely, speed to cost, cost to risk, and reference something timely, a job posting or recent news.
• Step 4 (Day 14): no new information. It's an exit, not another pitch. Give permission to say no, or ask 'should I close your file?' Length itself is the signal here.

How Do You A/B Test It?

One variable at a time, minimum 100 sends per variant before drawing conclusions. Don't test Steps 3-4 until Steps 1-2 are optimized, the sample sizes are too small to mean anything. Opens aren't a reliable read at any step: Apple Mail Privacy Protection and security scanners inflate them regardless of what actually happened.

What Should You Do First?

1. 01Pick one ICP segment: one title, one company type, one pain point.
2. 02Research a handful of accounts manually before using any automation, enough to know what a real signal looks like for this ICP.
3. 03Write the 4-step sequence, Step 1 under 90 words, shrinking through the breakup.
4. 04Don't scale until one variant has a meaningful positive reply rate.