Signal-triggered outreach and lead-scoring pipeline
The system I built to run my own outbound: it watches for buying signals, scores every account, drafts the first line, and queues the send for review. The same system I build for teams tired of spraying static lists.
// 01_problem
Cold outbound breaks the same way everywhere: one static list from a data vendor, every send identical, reply rates under 1%. Meanwhile the accounts most likely to buy are raising rounds and posting job ads in plain sight, and nobody is watching.
So I built the opposite. Reach out only when there's a real reason to, and say the reason out loud. I pointed it at my own pipeline first.
// 02_system
Three stages: detect, score, send. Signals come in from job boards, funding feeds, and tech-install data via Clay. An n8n pipeline dedupes and enriches each account, then a scoring model ranks it against the ICP. Anything above threshold gets a Claude-drafted first line that names the specific signal, queued in Smartlead with human review for the first 100 sends.
// 03_details-that-mattered
01 Signal decay. A funding announcement is worth a lot on day 2 and nothing on day 40. Scores decay on a half-life, so the queue stays fresh without manual pruning.
02 Review thresholds. Claude drafts, a human approves, until the approval rate crosses 95%. Then only low-confidence drafts hit the review queue.
03 Deliverability first. Warmup, rotation, and per-domain send caps were built in before the first email went out. No burned domains.
// 04_outcome
Running against my own pipeline now, with every reply rate and booked call tracked in the open, no rounded-up vanity numbers. Every workflow is documented and every prompt versioned, so when I build it for a team, they own it after handoff.
Want this pipeline pointed at your ICP?