Outbound sequence design that survives first contact
Most sequences fail for structural reasons, not copy reasons. Here is the shape of one that works, and the four decisions that determine whether it does.
Almost every underperforming outbound programme I have looked at is diagnosed the same way by the team running it: the copy is not good enough. So they rewrite the copy. Three weeks later the numbers are unchanged, and the conclusion is that outbound does not work in their market.
The copy is rarely the binding constraint. Sequence structure — how many touches, spaced how far apart, across which channels, with what exit rules — sets a ceiling that no amount of rewriting can lift.
The four decisions that matter
Every sequence encodes four decisions. Most teams make three of them by accident.
1. Step count
The honest answer is between six and nine touches over three to four weeks. Fewer than six and you are relying on a prospect being available on the day you happened to write. More than nine and marginal reply rate goes negative — you are no longer following up, you are being remembered for the wrong reason.
The distribution matters more than the count. Front-load: touches one through four inside the first ten days do most of the work.
| Touch | Day | Channel | Purpose |
|---|---|---|---|
| 1 | 0 | The reason you are writing, specific to them | |
| 2 | 2 | LinkedIn view + connect | Recognition, no ask |
| 3 | 4 | Email reply in thread | One new piece of evidence |
| 4 | 7 | Call | The only touch that can become a conversation today |
| 5 | 11 | Different angle, not a nudge | |
| 6 | 16 | LinkedIn message | Short, referencing the original reason |
| 7 | 23 | Explicit close-the-loop |
2. Spacing
Even spacing is a tell that nobody thought about it. Real follow-up is dense at the start and thins out, because the probability that a message is relevant decays and the probability that it is annoying does not.
Two constraints override the schedule: send inside the prospect's working hours in their timezone, and never send two touches on the same day across different channels. The second one is the difference between persistent and unpleasant.
3. Channel mix
A single-channel sequence is a bet that your one channel is the one they read. Email is the backbone because it scales and it is threaded. Everything else earns its place by doing something email cannot:
- LinkedIn provides recognition. A profile view before touch three measurably lifts reply rate on touch three, because the name is no longer cold.
- Calls are the only channel where a conversation can start immediately. One well-placed call around day seven is worth three extra emails.
- SMS is high-risk and occasionally decisive. Use it only where a prospect has given you the number.
4. Exit rules
This is the decision teams skip, and it is the one that damages the programme.
A sequence that keeps sending after a reply is not a sequence with a bug. It is a sequence that is actively destroying the relationship it just created.
At minimum, an enrolment must stop on: a human reply, a hard bounce, an unsubscribe, a suppression match, and a booked meeting. "Human" is doing work in that sentence — an out-of-office is not a reply, and treating it as one silently removes your best prospects from the sequence on the week they come back.
What to measure
Reply rate is the headline and it is not enough on its own, because it hides where a sequence dies.
- Delivery rate per mailbox. Below 97% and nothing downstream means anything.
- Reply rate per step. A sequence where 80% of replies land on step one is not a sequence — it is one good email with six wasted follow-ups behind it.
- Positive reply rate. Total replies including "no" is a vanity metric.
- Meetings per 100 enrolled. The only number a board asks about.
A healthy B2B sequence into a well-defined ICP produces 4–8% positive reply rate. Below 2%, the targeting is wrong and no rewrite will fix it. Above 12%, you have found something worth scaling carefully — reputation is the constraint at that point, not creativity.
The uncomfortable part
Sequence quality is bounded by list quality, and list quality is bounded by how honestly you defined your ICP. A perfectly structured sequence into a badly built list produces polite silence, which is the hardest failure mode to diagnose because nothing visibly breaks.
Start with the list. Then structure. Then copy — in that order, because that is the order in which the constraints bind.