Sequence Architecture

Pipeline contract

Stage 16 of 20. Reads 14-emails.csv, 15-linkedin.csv. Writes 16-sequence.md (steps, day offsets, channel per step) and 16-followups.csv (segment, step, day_offset, channel, subject, body).

• Each follow-up adds something new (proof, angle, resource). A bump that only says "following up" does not ship.
• Follow-up copy rules live in /cold-email section 7; this skill owns order, timing, and escalation.
• Drafts only. Sequence settings are a recommendation for the human's sender.

Files live in runs/<run>/. Column definitions: /prospecting-system. Standalone use: ignore this block.

When to Use

• Building a new outbound campaign from scratch
• Restructuring sequences with poor step-2+ engagement
• Designing multi-channel cadences (email + LinkedIn + phone)
• Setting up A/B testing frameworks for sequence optimization

Framework

Sequence Design Principles

1. Every step earns the next — Step 1's job is to get Step 2 opened. Step 2's job is to get a reply. Never front-load everything into Step 1.
2. New value every touch — Each step must introduce a new angle, insight, or proof point. "Following up" is not a strategy.
3. Decay is natural — Open and reply rates drop with each step. Design for this: your best material goes in Steps 1-2, your boldest ask goes in the final step.
4. Variants multiply learning — Run 2-3 subject/body variants per step. At 200+ sends per variant, you have statistical significance.

The 4-Step Core Sequence

This is the baseline architecture. Adapt timing and angles to your ICP.

Total campaign duration: 14 days Total touches: 4 emails

Extended Sequence (Multi-Channel)

For high-value accounts or enterprise targets, layer in LinkedIn and phone:

Total campaign duration: 14 days Total touches: 8 (4 email + 3 LinkedIn + 1 phone)

Variant Strategy

Every email step should have 2-3 variants to enable testing:

Testing rules:

• Only test ONE variable per variant pair (subject OR body, not both)
• Minimum 200 sends per variant before drawing conclusions
• Winner = highest positive reply rate (not total reply rate — ignore "not interested" replies)
• Rotate winning variants into the "control" and test new challengers

Timing Optimization

Sequence Settings Checklist

Configure these before launching any campaign:

Escalation Logic

What happens when the standard sequence doesn't convert:

Standard Sequence (14 days)
    ↓ No reply
Wait 30 days
    ↓
Re-engagement Sequence (different angle, 3 steps)
    ↓ No reply
Wait 60 days
    ↓
Trigger-based Re-entry
    (Only re-enter if new signal detected: funding, hiring, tech change)

Never:

• Re-enroll someone in the same sequence
• Contact someone who explicitly said "not interested"
• Send more than 2 sequences to the same person in 6 months without a new signal

Campaign Naming Convention

Use a consistent naming structure so performance data is analyzable:

{Client} - {Segment} - {Angle} - {Version}

Examples:

• Acme - Enterprise VP Sales - Hiring Signal - v1
• Acme - SMB Founders - Referral Ceiling - v2
• Acme - Mid-Market Growth - Tech Stack Change - v1

This makes it trivial to filter analytics by client, segment, or angle.

Architecture Decision Tree

Is the target account high-value (>$50K ACV)?
├── YES → Extended Multi-Channel Sequence (8 touches, 14 days)
│         Layer email + LinkedIn + phone
│         Personalize every touch individually
│
└── NO → Standard 4-Step Email Sequence (4 touches, 14 days)
         Use signal-based personalization at scale
         Rely on variants for optimization

Is this a new persona/segment you haven't tested?
├── YES → Run 3 variants of Step 1 with 200+ sends each
│         Wait for data before building Steps 2-4
│
└── NO → Clone your best-performing sequence
         Swap personalization for new segment
         Keep winning structure intact

Templates

Campaign Launch Checklist

Pre-Launch:
- [ ] ICP and persona defined
- [ ] Lead list built and verified (bounce rate < 3%)
- [ ] Email accounts warmed (2+ weeks, 30+ emails/day)
- [ ] Sequence written (2-3 variants per step)
- [ ] Settings configured (tracking off, text-only, stop on reply)
- [ ] Naming convention applied
- [ ] Test send to yourself and team for QA

Post-Launch (Day 3):
- [ ] Check open rates (target: 60%+)
- [ ] Check bounce rate (must be < 3%)
- [ ] Check spam complaints (must be 0)
- [ ] Review any replies for pattern recognition

Post-Launch (Day 14):
- [ ] Calculate positive reply rate per variant
- [ ] Identify winning variants
- [ ] Document learnings
- [ ] Plan next iteration

Tips

• Build your sequence backwards — write the breakup email first, then work back to Step 1. This forces you to think about the full arc.
• The best Step 2 is a reply to Step 1 (literally in the same thread), not a new email. This uses the "thread effect" for higher opens.
• Don't over-engineer your first sequence. Launch with 4 steps, 2 variants each, and iterate from data. Perfection before launch = wasted time.
• Keep a "swipe file" of every reply you get — positive and negative. The language prospects use to say "yes" and "no" is gold for future copy.
• If your open rate drops below 40% mid-campaign, stop sending. You have a deliverability problem, not a copy problem.

Progressive disclosure: load channel-specific templates and industry playbooks only when building sequences for a specific campaign.