Hey 👋
I was reading a sequence this week where the first note was sharp.
It named a hiring post. One stake. Three sentences. The kind of opener that would fail a company-swap test in a good way.
What struck me wasn't the craft. The craft is fine.
It was touch three.
"Just checking in." No hiring post. No new angle. No reason that could stand alone in the inbox. The sequence had spent the signal on day one, then spent the next two touches reminding the buyer they had already decided not to reply.
Last week was personalization theater: research that never made the note. This is the next leak: the note had a reason, then the follow-ups erased it.
Here's what that actually looks like.
TL;DR
Follow-ups (steps 2 to 6) account for 58.6% of email replies; step 1 alone is only 41.4% (Belkins, 7.5M emails, 2025)
Step 3 alone books 35.6% of email-sourced meetings; over 53% of meetings land at step 3 or later (Belkins)
"Bubble-up" follow-ups under ~30 words are 15× less likely to book a meeting than 30 to 150 word value follow-ups; "Thoughts?" cuts meetings ~20% (Gong Labs, 304k follow-ups)
Pitching in cold email can drop replies by up to 57%; top length band is under ~100 words / 3 to 4 sentences (Gong, 28M+ emails)
Verified ICP + subject-matter fit: 51.9% reply / 6.5% meeting vs 13.0% / 1.4% for broad role targeting (Prospectio, controlled test)
This week's move: every follow-up must stand alone with a fresh reason. Kill "just checking in" in the approve queue. No new reason means no send.
Weekly finds
Step 3 is where the meetings hide. Belkins tracked 7.5M emails. Most teams quit right before the touch that actually books.
"Did you see my last email?" is not a strategy. Gong has a name for these bumps, and a 15x worse booking rate to match.
No new fact, no next touch. Explorium's sequence rule is brutally simple. It will probably delete half your cadence.
"Anyone with this job title" is not an ICP. A controlled test where tighter targeting booked about 4.6x more meetings. Awkward news for copywriters.
Sonnet got faster. Your bump still has no reason. Claude Sonnet 5.5 drafts quicker and cheaper. It still can't invent a second reason for you.
THE ISSUE
The Follow-Up Erased the Reason
You'll like this one if your first touch is sharp and your cadence still feels like spam by Friday.
Signal-led outbound fixed the open: you finally had a why-now.
Then the sequence did what sequences do. It continued. On autopilot. With nothing left to say.
That's how a good reason dies: not in the first sentence, in the third touch.
Replies skew early. Meetings skew late.
Belkins' split is the chart most dashboards hide.
Step 1 collects the biggest slice of replies (curiosity, referrals, auto-responders, polite nos). Step 3 books the biggest slice of meetings. Over half of email-sourced meetings show up at step 3 or later.
If you kill the campaign because day-four reply rate looked soft, you often kill it right before the touch that actually converts.
Practical: Judge sequences on meetings per 1,000 accounts after the full cycle, not on step-1 reply vanity. Protect step 3 as a real message, not a nudge.
Bubble-ups hand the work back to the buyer
Gong's name for the disease is perfect: bubble-up emails.
"Just floating this up." "Did you see my previous note?" "Thoughts?"
They assume the buyer remembers your value, still has the thread open, and wants to do archaeology for a stranger. Shorter than ~30 words is about 15× worse at booking. "Thoughts?" can cut meetings ~20%. "Following up" is empty calories with a 5% meeting tax.
You didn't follow up. You reminded them they already declined.
Practical: In the approve queue, auto-fail any draft whose entire job is to point at a prior email. If it cannot stand alone in a new thread, rewrite it or kill it.
The sequence ran out of information, not patience
Explorium's framing is the honest one: most sequences die at email 2 because the sender spent their one fact on email 1.
Touch 2 restates it. Touch 3 becomes a reminder about touch 1. Persistence without a new verifiable detail is just volume with guilt attached.
You do not need a new CRM schema for "hours since signal." You need a boring gate: before you approve a follow-up, name the new reason out loud. Hiring update. Different stake. Fresh angle. Dated fact. If you cannot name it, the touch does not fire.
Practical: Cap the approve queue. For every follow-up, glance at the original signal and ask what is new since. No new reason: archive until one appears. Do not invent a bump because the calendar said "day 4."
Targeting still beats a prettier bump
Prospectio's controlled test is awkward for people who want the fix to be copy alone.
Broad role targeting: ~13% reply, ~1.4% meeting. Verified ICP plus subject-matter fit: ~52% reply, ~6.5% meeting. That is targeting explaining more variance than another rewrite of "just checking in."
A sharp follow-up on the wrong person is still the wrong person. A mediocre bump on a verified fit still burns trust.
Practical: Before you write touch 3, re-check fit. Title drift and "they engaged once" are not the same as ICP. If the row would not earn a first touch today, it does not earn a follow-up either.
What we change
No new stack required.
One reason per touch. Fresh, dated, specific. Or no send.
Step 3 is a first email. Different hook, different proof, own ask. Not a polite poke.
Kill bubble-ups on sight. "Just checking in" is not a draft. It is a delete.
Approve the reason, not the cadence. The calendar does not get a vote if the account went quiet for a reason.
Measure meetings by step. If your board only shows blended reply rate, you are flying blind into the only touch that books.
This week's experiment
Pull the last twenty follow-ups you approved (or almost approved).
Tag each one: new reason or bubble-up.
Send only the new-reason pile. Rewrite or kill the bubble-ups so each stands alone without the prior thread. Same-day where the original signal is still hot.
Compare reply quality and meetings booked for one week. The spreadsheet usually shuts up..
