Outbound Engine

What it does

outbound-engine turns cold outbound from a "write-and-blast" task into an auditable engineering process across three phases:

Phase 1 — Discovery & Audit. Clarifies ICP and anti-ICP via references/icp-template.md; collects business context (what you sell, the primary offer, proof points, real URLs). In optimize mode it calls the Instantly v2 API to pull active campaigns, sending accounts, warmup scores, and domain inventory, flagging any account with score <80 or <14 days warmup as "NOT ready" — the single biggest deliverability lever.

Phase 2 — Recursive Expert-Panel Scoring. A 10-expert panel (subject-line, pattern-interrupt, CTA, deliverability, personalization lenses) scores the copy 0-100. The target is 90/100, non-negotiable: every round outputs a score table + aggregate + top-3 weaknesses + changes made + revised copy, iterating until it lands. The iteration trail ships as part of the value.

Phase 3 — Deliverables. Brutal-truth pre-analysis, ICP summary, infra status, all scoring rounds, Instantly-ready final copy, implementation plan, capacity math, weekly metric targets, and STOP/START lists.

When to use it

• Standing up a cold outbound sequence (4-7 touches) from scratch.
• Repairing a low-reply or spam-foldered existing campaign.
• Planning warmup and capacity for new domains/inboxes.
• Narrowing ICP and rewriting messaging per vertical.

Method / frameworks

• Recursive expert-panel scoring — 90/100 bar, no padding.
• Capacity math: ready accounts × daily sends × 22 working days → monthly volume → replies → pipeline.
• Deliverability discipline: SPF/DKIM/DMARC, staged warmup (5-10/day ramping to 50/inbox), bounce <2%, spam complaints <0.1%, one-click unsubscribe — aligned to 2025 bulk-sender rules.
• Sequence design: 4-7 touch sweet spot, each touch adding new value.
• Human review gate: the doc is for review; no API writes without explicit approval.