Your cold emails get ignored
Run two 3-email sequences against matched lists of 100 prospects each. Compare reply rate and meetings booked. Kill the loser, double the winner.
When to use
You have a sharp ICP (ran problem interviews already), a clean list of 200+ prospects, and a sending setup that won't land in spam.
Hypothesis template
If we run sequence A (pain-led) vs sequence B (insight-led) against 100 matched prospects each, one will outperform on reply-to-meeting rate by at least 2x.
Method
The problem: Cold email "doesn't work" — except it does for everyone who tests properly. You're sending one version, getting nothing, and concluding the channel is dead.
What works: Treat each sequence as one experiment. Two variants only. Same list quality, same send window, same sender. Different angle.
How to do it:
- Build two matched lists of 100 ICP-fit prospects (same role/company size/segment). Split randomly.
- Write two 3-email sequences:
- A — Pain-led: "I noticed [specific signal that suggests they have the problem]. Most [persona] we work with deal with [pain]. Worth 15 min?"
- B — Insight-led: "We just analyzed [N] [persona] and found [surprising stat]. Thought you'd want to see it. Open to a 15-min walkthrough?"
- Send same days, same hours, same sender. 3 emails over 8 days.
- Track: open, reply, positive reply, meeting booked, meeting attended.
- After 2 weeks: declare the winner on meetings booked (not opens — opens are a vanity metric in 2026).
Output: Winning sequence + the reason it won (you'll learn whether your ICP responds to pain or to insight). Use this to write the next test.
Success metrics
- •Reply rate per variant
- •Positive reply rate per variant
- •Meetings booked per 100 prospects
- •Meeting-to-qualified rate
Prerequisites
- Sharp ICP
- List of 200+ matched prospects
- Email sending tool (Instantly, Smartlead, Apollo, etc.)
- Domain warmed up
Common pitfalls
- •Comparing sequences with different list quality
- •Optimizing for opens instead of meetings
- •Sending fewer than ~100 per variant — too noisy to call
- •Three variants instead of two — too thin per arm
Source: Standard outbound playbook. Predictable Revenue (Aaron Ross) + every modern outbound shop.
Suggested ICE scores
Ready to run this experiment?
Sign up free to use this playbook with step-by-step guidance and track your results.