Clay toolkit skill

Clay Toolkit

Location: tools/clay-toolkit · Skill name: clay-toolkit

Overview

Clay is the operating system for data-driven GTM — spreadsheet-native enrichment, scoring, and CRM push. The mistake: using Clay as a lookup table with one provider and no verify step. This skill covers table architecture, LeadMagic-first waterfalls, Claygent, credit optimization, and output routing.

Clay tool stack:

• clay-toolkit (this skill) — one-time tables and waterfalls
• clay-loops-toolkit — recurring signal loops
• clay-automation (automation/) — process and rollout, not column config

Authoritative Foundations

• Clay. Table-native enrichment: source → waterfall columns → formula → output. Separate company and person tables; join on domain.
• LeadMagic on Clay. Default first step for email: Find (1 cr) → Verify (1 cr) → Enrich Person (2 cr). Highest accuracy-per-credit for outbound tables. Load leadmagic-toolkit for column wiring.
• Pat Spielmann — Clay + enrichment outbound. Signal table → LeadMagic waterfall → validate → AI personalization column → sequencer push. Copy structure: Hook-Line-Sinker. Canonical playbook → references/pat-spielmann-outbound-copy.md.
• Clay University / GTME Pulse. Production table patterns at scale.
• ColdIQ. Signal tables feed clay-loops-toolkit — not mixed into static tables.

When to Use

Trigger phrases: "Clay setup", "Clay waterfall", "Clay enrichment", "Claygent", "Clay table architecture", "Clay automation", "Clay credit optimization", "Clay Sculptor", "Clay formula", "Clay CRM push", "Clay signal table"

For recurring signal monitors → clay-loops-toolkit (tools/clay-loops-toolkit). For AI column prompts → ai-prompts-toolkit. For rollout process → clay-automation.

Core Concepts

Table Architecture

• Source Sheet: Google Sheets, CSV, or API — where your data lives
• Enrichment Columns: Each column is an enrichment step
• Waterfall Columns: Chain multiple providers per data point
• Formula Columns: Transform data between enrichments
• Output: CRM push, webhook, or export

Waterfall Strategy

Goal: Find email for a list of contacts

Column 1: LeadMagic Email Finder
  → Found? Done.
  → Not found? Fall through to...

Column 2: Apollo Email Finder
  → Found? Done.
  → Not found? Fall through to...

Column 3: ProspectingTool Email Finder
  → Found? Done.
  → Still not found? Flag for manual review.

GTM Table Blueprints

Load full column specs in references/gtm-table-blueprints.md.

Blueprint	Rows	Primary Use
Outbound list	Contacts	Waterfall email → verify → ICP score → CRM
Account ABM	Companies	Firmographics → technographics → tier → owner
Signal outbound	Contacts	Monitor column → enrich if signal → draft → route
CRM hygiene	Contacts	Stale check → re-verify email → update CRM
Event follow-up	Contacts	Attendee list → enrich → sequencer within 48h

Company vs person rule: Always separate tables. Join on domain.

Clay Loops (Overview)

Loops = recurring GTM monitors. Full designs in clay-loops-toolkit.

SOURCE → MONITOR (cheap) → ENRICH (if signal) → ACTION (route)

Common loops: funding, job change, hiring, stale opp, champion move.

Sculptor + Claygent (GTM)

• Sculptor: Natural-language table builder — start from blueprint in references/gtm-table-blueprints.md, then refine in UI.
• Claygent: Use prompts from ai-prompts-toolkit P01–P10 — never raw "find email" without source-url rules.

Sequencer + CRM Output Patterns

Output	When	Skill
HubSpot/SF create/update	Score ≥ threshold	crm-toolkit
Smartlead/Instantly webhook	Verified + approved	sequencing-toolkit
Slack notify	Tier-1 signal	proactive-alerts
n8n webhook	Complex branching, inbound SLA, reply routing	n8n-toolkit

Push fields: email, first_name, company, icp_score, why_now, personalization_source, clay_status.

Credit Optimization

• LeadMagic: 1 credit = find + verify. Best accuracy.
• Apollo: credits per export. Good breadth.
• Claygent: AI-powered research agent. Use sparingly (expensive).
• Rule: Run cheapest/highest-accuracy provider first. Fall back to more expensive.

Artifacts

Standard Clay Table Template

COLUMNS (left to right):

1. INPUT COLUMNS:
   - First Name (text)
   - Last Name (text)
   - Company (text)
   - Domain (text or formula: clearbit domain from company)
   - LinkedIn URL (optional)

2. ENRICHMENT WATERFALL — EMAIL:
   - LeadMagic Find Email (name + company → email)
     Credit cost: 1. Confidence: High.
   - Apollo Find Email (name + company → email)
     Credit cost: 1. Confidence: Medium.
   - ProspectingTool (name + domain → email)
     Credit cost: 1. Confidence: Medium-Low.

3. VERIFICATION:
   - LeadMagic Verify Email (email → status + confidence)
     Credit cost: 1. Status: valid/invalid/risky/unknown.

4. ENRICHMENT — PERSON:
   - LeadMagic Enrich Person (email → title, company, phone, social)
     Credit cost: 2.

5. ENRICHMENT — COMPANY:
   - Clearbit Company (domain → employee_count, industry, revenue)

6. SCORING:
   - Formula: ICP Score =
     (Title matches ICP: 30) +
     (Company size 50-500: 20) +
     (Email valid: 30) +
     (Industry matches: 20)

7. OUTPUT:
   - CRM Push (HubSpot / Salesforce)
   - Google Sheets sync
   - Webhook → n8n → downstream workflows

Common Clay Formulas

// Extract domain from email
=RIGHT(Email, LEN(Email) - FIND("@", Email))

// ICP Score calculation
=SUM(
  IF(CONTAINS(Title, "VP"), 30, IF(CONTAINS(Title, "Director"), 20, 0)),
  IF(AND(Employees > 50, Employees < 500), 20, 0),
  IF(Email_Status = "valid", 30, 0),
  IF(Industry = "Technology", 20, 0)
)

// Clean company name for domain lookup
=LOWER(SUBSTITUTE(SUBSTITUTE(Company, " ", ""), ",", ""))

// Time since last enrichment (days)
=DATEDIF(Last_Enriched, TODAY(), "D")

Implementation Checklist

• Source data cleaned (no duplicates, consistent formatting)
• Waterfall order optimized (highest accuracy first, cheapest per hit)
• Credit budget set per table (prevent runaway costs)
• Webhook configured for async enrichment completion
• CRM push tested with 5 records before full sync
• ICP scoring formula validated against 50 known-fit contacts

Common Pitfalls

1. Flat enrichment (no waterfall). One provider. Missed contacts stay missed. Fix: Waterfall. LeadMagic → Apollo → fallback → manual review.
2. Credit burn on bad data. Enriching 10,000 contacts with bad names/domains. Fix: Clean data first. Deduplicate. Validate domains before enrichment.
3. No ICP filter before CRM push. All contacts pushed to CRM. SDR team overwhelmed. Fix: ICP score column. Only push score > X to CRM.

Output Format

The agent delivers a Clay table blueprint and supporting implementation assets:

• Table Architecture: ordered column list — input columns (name, company, domain, LinkedIn URL), waterfall enrichment columns (LeadMagic → Apollo → fallback), verification column, person/company enrichment columns, ICP scoring formula column, and CRM push output config
• Waterfall Configuration: provider order with credit cost per hit, fallback trigger condition (no result), and manual-review flag for exhausted chains
• Formula Reference: ready-to-paste Clay formulas for domain extraction, ICP score calculation, company name normalization, and enrichment-staleness detection (days since last run)
• Credit Budget Estimate: projected credit consumption per 1,000 contacts at the configured waterfall depth with cost-optimization recommendations
• CRM Push Rules: ICP score threshold for inclusion, HubSpot/Salesforce field mapping, deduplication check before sync, and webhook trigger for async completion

Quality Check

Before delivering, verify:

• All required sections complete
• Output matches the user's stated need
• No vague or unsupported claims
• Frameworks cited where applicable

Execution Artifacts

• references/framework-notes.md — Clay architecture principles
• templates/output-template.md — table design deliverable
• scripts/check-output.py — validates table blueprint output
• references/pat-spielmann-outbound-copy.md — Clay waterfall → copy pipeline, Claygent vs LeadMagic roles (Pat Spielmann)
• references/gtm-table-blueprints.md — Outbound, ABM, signal, hygiene tables
• templates/waterfall-config.md — provider order per field

Related Skills

• clay-loops-toolkit — Recurring signal loops (funding, job change, hiring)
• ai-prompts-toolkit — Claygent and LLM column prompts + prompt loops
• clay-automation — End-to-end Clay workflow process
• leadmagic-toolkit — LeadMagic provider in waterfalls
• sequencing-toolkit — Push enriched contacts to sequencers
• waterfall-enrichment — Multi-provider waterfall design
• gtm-role-descriptions — Hire GTM Engineer to own Clay stack (gtm-engineer-hiring.md)