Most replies to cold email do not come from the first message, but every extra step brings fewer. Below: how many steps to plan, what each step should do differently, when a sequence stops, and how a sequence differs from a cadence.
What is an email sequence?
A sequence is a set of email steps with delays between them, attached to a list of contacts. Each contact moves through the steps on their own clock. A reply takes the contact out; silence moves them to the next step.
A good cold sequence has three layers. The message is written once, by a person. Spintax varies the wording, so a hundred recipients do not get a hundred identical copies. A short generated clause adds something specific about each contact. Leadspicker's guidance is to generate fragments with AI, never whole messages, because native speakers spot generated text immediately.
How many steps, and how far apart?
Plan for about four steps. Replies keep arriving after the first message but fall with every step, and the benchmarks agree on that shape.
| Step | Smartlead, H1 2026 | Woodpecker, Q4 2025 (median) |
|---|---|---|
| 1 | 1.18% | 1.0% |
| 2 | 0.73% | 0.9% |
| 3 | 0.56% | 0.6% |
| 4 | 0.49% | 0.5% |
| 5 | 0.37% | Not reported |
Reply rate by step. Smartlead: more than 850 million emails sent in H1 2026. Woodpecker: median across its users, Q4 2025. Both are vendor data from each tool's own customers.
Instantly's 2026 benchmark found that 58% of replies come from the first step and 42% from follow-ups. An older study of 12 million outreach emails by Backlinko and Pitchbox found that one follow-up lifted replies by 65.8%, though that was link-building outreach, not sales. Stopping after one email gives away close to half the replies; a fifth step adds little.
Spacing matters less than the content of each step. Leave a few working days between steps and longer gaps later, as in the example below.
Step-by-step benchmarks come from sending tools. Smartlead's H1 2026 data (850 million plus emails) shows reply rate falling from 1.18% on step one to 0.37% on step five. Woodpecker's Q4 2025 benchmarks show medians of 1.0%, 0.9%, 0.6% and 0.5% for the first four steps. Instantly's 2026 report attributes 42% of replies to follow-ups. Backlinko and Pitchbox found a 65.8% lift from one follow-up across 12 million link-building emails in 2019. All four measure their own users, so treat them as shape, not target.
What should each step do differently?
Each step needs a new reason to reply, not the same message again.
| Step | Day | Job | Example line |
|---|---|---|---|
| 1 | 1 | Reason for writing and one question | {Saw|Noticed} that [AI clause: the company opened a Brno office in March]. Who runs {new business|outbound} for the new team? |
| 2 | 4 | Short follow-up in the same thread | {Quick|Short} follow-up on my note below. Is this on your list this quarter? |
| 3 | 9 | A new angle: one concrete example | A team in a similar setup started with [one specific example]. Worth a short call? |
| 4 | 16 | Polite close | {I will leave it here|I will not follow up again}. If timing changes, a one-line reply is enough. |
Illustrative sequence. Curly braces mark spintax alternatives, square brackets mark the AI-generated clause. Day spacing is an example, not a benchmark.
Everything outside the brackets is written by hand. Inside Leadspicker, spintax cuts repeats to roughly one identical message in 150. Keep every step plain text: no images, no tracking pixel, one link at most.
When does a sequence stop?
- On a reply: any reply, positive or negative, stops all remaining steps for that contact.
- On a bounce: a hard bounce removes the address; writing to it again damages the domain.
- On an opt-out: the contact moves to a blacklist that applies across campaigns.
- After the last step: no reply ends the sequence. Do not restart the same sequence on the same contact.
- On a reply elsewhere: in a multichannel sequence, a reply on another channel stops email too.
How does a sequence differ from a cadence?
A cadence is the plan: the rhythm, channels and timing a team agrees on, typically seven or eight touches across email, phone and a social network. A sequence is its implementation in a tool: the actual steps, templates and delays, typically about four email steps inside that cadence. Many tools, Leadspicker included, use 'sequence' for multichannel flows too, so check which one a vendor means.
Leadspicker has no published reply figure by step. Its guidance is to write the sequence once by hand, vary it with spintax, add one AI-generated clause per contact, and stop on reply.
Frequently asked questions
How many emails should a cold sequence have?
About four. Benchmarks show reply rate falling with every step, from just over 1% on the first message to about 0.5% by the fourth, while follow-ups still bring in around 42% of all replies.
Should AI write the whole sequence?
No. Write the message once by hand and use AI for one specific clause per contact. Fully generated messages are easy for a native speaker to recognise.
When should a sequence stop?
When the contact replies, bounces or opts out, or after the last step. In a multichannel sequence, a reply on any channel should stop every remaining step.
- Smartlead, How many follow-up emails is too many? (H1 2026)
- Woodpecker, Cold email benchmarks (Q4 2025)
- Instantly, Cold Email Benchmark Report 2026
- Backlinko and Pitchbox, 12 million outreach emails study (2019)
- GTM playbook for 2026 by Leadspicker, chapters: Spintax; AI personalisation; Writing cold outreach that works
