The cadence that works (the numbers)
From campaigns in our system with 500+ emails and statistically meaningful reply data:
The sequence structure that performs
Day 1: Initial email — personalized, references a specific signal or context. Day 3: Follow-up email — different angle, shorter. Don't just bump the thread. Day 5: LinkedIn connection request — with a brief personalized note. Day 8: Follow-up email — new value angle or social proof. Day 11: LinkedIn message (if connected) or comment on their post. Day 15: Final email — direct ask, acknowledge the sequence, give an easy out. Day 18 (optional): Phone call — for high-value prospects only.
This structure works because it's multichannel (email + LinkedIn + phone), varies the angle on each touch, and gives enough breathing room that it doesn't feel like harassment.
The biggest cadence mistake: stopping too early
Most teams send 2 emails and move on. The data says this leaves 44% of potential replies on the table.
Why: B2B buyers are busy. Your first email arrived when they were in a meeting. Your second arrived on a Friday afternoon. Neither got read — not because the message was bad, but because the timing didn't align with an open moment.
Touchpoints 3–5 exist to catch the prospect in a different context. A LinkedIn message catches them scrolling during a commute. A Thursday morning email catches them during a quiet inbox window. Persistence (with varied angles) is not spam — it's how outbound works.
The line between persistence and spam: each touchpoint adds new value or a new angle. If you're sending the same 'just checking in' message five times, that's spam. If each touch references a different reason the prospect should care, that's a sequence.
Why multichannel beats email-only (by 2.5×)
Email-only sequences are fighting in one arena. The prospect's inbox is crowded — 20+ cold emails/day for anyone in B2B SaaS. Adding LinkedIn and phone creates additional surface area.
The multiplier works because channels reinforce each other: - A LinkedIn connection makes the email sender name recognizable → higher open rate - An email makes the LinkedIn message feel contextual, not random - A phone call after email + LinkedIn engagement is warm, not cold
The key: coordinate the channels as one sequence, not three separate campaigns. The prospect should feel like one person is reaching out through multiple channels — not like three different people are blasting them independently.
Send timing: matters less than you think
Every blog post about cold email cadence obsesses over send time. Tuesday at 9:14 AM. Thursday at 7:42 AM. The data shows this optimization is real but minor — a 5–10% lift at best.
What matters 10× more than time-of-day: signal-based timing. Sending your first email the week a prospect changes jobs, their company announces funding, or they post about a relevant problem on LinkedIn — this produces a 2–3× reply rate lift. That dwarfs any send-time optimization.
If you're optimizing send time before you've implemented signal-based triggering, you're polishing the dashboard while the engine is off.
Frequently asked questions
How many follow-up emails should I send?
4–6 follow-ups after the initial email. 44% of positive replies come on the 3rd touchpoint or later. Most teams give up too early — the biggest cadence mistake is stopping after 2 emails.
How long should I wait between cold emails?
3 days between email 1 and 2. Then 4–5 days between subsequent touches. Shorter gaps feel aggressive; longer gaps lose momentum. The sweet spot is 14–21 days for a complete sequence.
Should I mix channels in my cadence?
Yes. Email + LinkedIn + phone sequences get 2.5× the reply rate of email-only sequences at the same volume. The channels reinforce each other — a LinkedIn connection makes the email feel less cold.
When should I send cold emails?
Tuesday through Thursday, 8–10 AM in the recipient's time zone. But timing matters less than you think — signal-based triggering (sending when something relevant happens at the prospect's company) outperforms time-of-day optimization by 3×.