Signal Stack Engagement Playbook

Purpose

This recipe executes the full Signal Stack consulting lifecycle: from identifying raw intent signals in a target industry through building an automated intelligence pipeline that converts those signals into qualified sales dossiers. The end state is a self-reinforcing flywheel where each delivered dossier generates feedback data that improves signal classification accuracy. [src1, src2]

Prerequisites

• Target vertical defined with at least 1,000 addressable companies
• Sales baseline — current conversion rates, average deal size, sales cycle length
• CRM access with historical conversion data for at least 6 months
• Budget confirmed for data sources ($500-$2K/month) and LLM API costs ($200-$500/month)
• Sales team identified who will consume and act on delivered dossiers

Constraints

• All data sources must be publicly accessible or properly licensed — no scraping behind authentication without legal agreement. [src2]
• Human-in-the-loop validation required for first 100 dossiers per vertical.
• No platform extraction until 3 paying customers in vertical #1. Hard gate. [src3]
• GDPR, PECR, and CAN-SPAM compliance mandatory before any outreach delivery.
• Minimum viable pilot: 10 dossiers/week for 4 consecutive weeks.

Tool Selection Decision

Which path?
├── Client has existing sales data + CRM
│   └── PATH A: Data-enriched pipeline
├── Client has target vertical but no CRM data
│   └── PATH B: Signal-first pipeline
├── Client needs multiple verticals simultaneously
│   └── PATH C: Platform-first (only if 3+ paying customers exist)
└── Client unsure about vertical selection
    └── PATH D: Discovery engagement

Execution Flow

Phase 1: Signal Audit (Weeks 1-2)

Duration: 5-10 days · Tool: Web research + structured scoring framework

Execute full signal source audit for the target vertical. Inventory every available data source: regulatory databases, behavioral signals, visual signals, and unstructured media. Score each on accessibility, cost, refresh rate, and signal-to-noise ratio. [src1]

Verify: Signal audit report with minimum 15 scored sources across 3+ categories. · If failed: Vertical may lack signal density — run viability scoring.

Phase 2: Taxonomy Workshop (Weeks 2-3)

Duration: 2 days (workshop) + 3 days (validation) · Tool: Workshop facilitation + live data testing

Day 1: domain expert interviews, data source deep-dives, trigger event brainstorming. Day 2: signal strength scoring, false positive threshold calibration, compound signal design, taxonomy validation against live data. [src5]

Verify: Taxonomy validated against 50+ real-world examples. False positive rate < 30%. · If failed: Add qualifying signals, tighten thresholds, revalidate.

Phase 3: MVP Pipeline Build (Weeks 3-6)

Duration: 2-4 weeks · Tool: Python + LLM API + enrichment APIs + cron

Build minimal pipeline: cron job pulling public data, LLM classification using taxonomy, enrichment via Clearbit/Apollo, dossier generation as PDF, email delivery with tracking. No UI required. [src2]

Verify: Pipeline producing 10+ dossiers/week with < 5% error rate. · If failed: If accuracy < 70%, return to taxonomy.

Phase 4: Pilot Execution (Weeks 6-10)

Duration: 4 weeks · Tool: Pipeline + tracking + weekly iteration

Select 2-3 pilot customers with measurable baselines. Deliver 10-20 dossiers/week. Track conversion rates. Iterate taxonomy weekly. Human validates first 100 dossiers. [src1, src3]

Verify: 2x conversion improvement vs cold outreach after 4 weeks. · If failed: Diagnose signal quality vs dossier format.

Phase 5: Platform Extraction (Weeks 10-14)

Duration: 3-4 weeks · Tool: Software architecture + refactoring

Hard gate: 3 paying customers required. Refactor into reusable engine + vertical-specific config layer. Target: subsequent verticals < 50% effort. [src4]

Verify: Second vertical launched in < 50% of vertical #1 timeline. · If failed: Identify tight-coupled components, refactor before vertical #3.

Phase 6: Vertical Scaling (Weeks 14-20)

Duration: 4-6 weeks per vertical · Tool: Platform + config + domain advisors

Launch additional verticals using the platform. Per vertical: recruit domain advisor, verify data sources, build classifier rules, design templates, identify pilot customers, ensure compliance. [src3, src5]

Verify: Each new vertical achieves > 1.5x conversion improvement within 4-week pilot. · If failed: Evaluate signal density — kill non-viable verticals quickly.

Output Schema

{
  "output_type": "signal_stack_engagement",
  "format": "multi-deliverable",
  "sections": [
    {"name": "signal_audit", "type": "object", "description": "Scored inventory of data sources"},
    {"name": "signal_taxonomy", "type": "object", "description": "Classification schema with weighted scoring"},
    {"name": "mvp_pipeline", "type": "object", "description": "Deployed pipeline specification"},
    {"name": "pilot_results", "type": "object", "description": "Conversion metrics and ROI"},
    {"name": "platform_architecture", "type": "object", "description": "Reusable engine design"},
    {"name": "vertical_playbooks", "type": "array", "description": "Per-vertical launch configs"}
  ]
}

Quality Benchmarks

If below minimum: Stop and diagnose. Signal density may be insufficient or taxonomy needs recalibration.

Error Handling

Cost Breakdown

Anti-Patterns

Wrong: Building the platform before proving vertical #1

Investing in reusable architecture before confirming product-market fit. Result: over-engineered system nobody uses. [src2, src3]

Correct: Prove value with ugly code first

Build the simplest pipeline for vertical #1. Hardcode everything. Extract a platform only after 3 paying customers.

Wrong: Skipping human-in-the-loop validation

Automating delivery from day one without review. Result: false positives destroy credibility with pilot customers. [src1]

Correct: Human validates first 100 dossiers per vertical

Every dossier gets human review for first 100 deliveries. This calibrates taxonomy and catches edge cases.

Wrong: Launching multiple verticals simultaneously

Trying to prove the model in 3 verticals at once. Result: none get enough attention, all produce mediocre results. [src3]

Correct: Sequential vertical execution

Fully prove vertical #1 (3 paying customers), extract platform, then launch vertical #2.

When This Matters

Use when an agent needs to plan or execute a full Signal Stack consulting engagement — from identifying intent signals through building an automated intelligence pipeline. This is the master recipe orchestrating all sub-recipes into a cohesive engagement lifecycle.