SuperAPI AI Directory

crawford-method

Data & analytics

View repository

Updated 10-16 05:03

About this skill

The Crawford Method

"The list is the message. If you have a good list, you have a good message."

Targeting IS messaging. Build the list on data heuristics that reflect the buyer's real situation, then describe that situation back to them. This skill is methodology; it sources its tension heuristics from ColdIQ endpoints (see callouts) but writes no API calls of its own.

Two message types

  • PQS — Pain-Qualified Segment: 2–5 public heuristics identify prospects in the same painful situation as an existing customer; tell the story of how your customers navigated it. Formula: name the obvious choice → why your customers made the non-obvious choice → the benefit.
  • PVP — Permissionless Value Prop (always superior): synthesize 3–5 public data sources into a non-obvious insight that is independently valuable. Acid test: would the recipient pay to receive this even if they never buy? If yes → PVP. Use PVP whenever you can.

The Tension Triangle

Tension = a specific, data-provable, time-bound situation that makes them need you NOW. Not a single signal — multiple data points pulling against each other.

  • Heuristic 1 — a unique, data-provable characteristic
  • Heuristic 2 — a second characteristic that PULLS AGAINST the first
  • Growth amplifier — funding / headcount / revenue growth that increases the pull
  • Existential data point — the metric that separates good-fit from bad-fit (sweet-spot range)

Source these heuristics from public data via ColdIQ: → GET /v1/signalbase/funding-signals · ? cr · signalbase.funding_signals (unverified) — funding amplifier → POST /v1/limadata/jobs · 2 cr · limadata.jobs — hiring / no-SDR heuristics → POST /v1/limadata/research/ai-search · 0.3 cr · limadata.research.ai_search — filings, regulation, public events → POST /v1/limadata/enrich/company · 1 cr · limadata.enrich.company — growth, tech, firmographic heuristics

7-component email architecture

  1. Visceral emotional opening — a moment they've physically felt, not a fact.
  2. Connection to existing behavior — frame the offer as an extension of what they already do.
  3. The insight (contrarian) — "Most people think X, but it turns out Y." Must pass the recognition + articulation tests.
  4. Social proof / authority (light) — one line, founder credentials not company stats.
  5. Information gap — enough to create curiosity, never enough to answer everything.
  6. Qualification as disqualification — a threshold that excludes non-fits.
  7. Soft CTA (ask for truth, not time) — "Am I close, or is it different at {{companyName}}?" NEVER "Do you have 15 minutes?" / "Book a time."

11 rules (condensed)

Never mention your product in step 1 · write about their situation not your solution · target on events not roles · use Cialdini systematically · create information gaps · use their language · sell against incumbent behavior not competitors · acknowledge multiple solutions exist · one perfect email > ten mediocre · data-driven > surface personalization · quantify the meeting in dollars.

Gold-standard example (PVP — Texada)

Earl, your CAT 365 excavator seems to be not on the move right now.
They just filed a large construction project for 123 Main Street — half a mile from you.
AI says it'll likely need your tool for ~six weeks.
Public rental rates put that at about a $170,000 job.
Here's the foreman, their email, their phone. Hope this was useful.

Independently valuable, no product mention, three public sources synthesized, quantified — the message IS the demo.

The process (zero → finished email)

  1. Understand best customers (call transcripts: what was their situation when they bought?).
  2. Extract tension (H1, H2, growth amplifier, existential data point).
  3. Find the data sources — can you get the heuristics programmatically? (filings, jobs, tech, ad libraries, funding — source via the ColdIQ endpoints above).
  4. Build the list on tension, NOT firmographics.
  5. Write backwards: define the perfect message, then find the data that makes it true.
  6. Choose PQS or PVP. 7. Apply the 7-component architecture. 8. Run the quality checklist.

Anti-patterns

"I noticed" openings · "companies like yours" · "do you have 15 minutes" · features dump · surface personalization (location, mutual connections, LinkedIn posts) · horizontal targeting · over-sequencing (max 2–3 touches, each adding new information).