Permissionless Value Props Playbook

Pick the finding that would help the account even if they never buy from you.

• The framing comes from Jordan Crawford at Cannonball GTM: combine two to five data sources into an insight the prospect doesn't already have.
• Write the source map before any copy:
  • The market question
  • The public source of truth
  • How a record joins to a company
  • Who resolves the buyer
  • The first-party filter
  • The evidence fields a reader needs to trust it
  • What gets dropped before anything sends

Claude Code fans out across public conversations to find where the data lives and what buyers call it.

• Claude Code runs a research scout across Reddit, X, YouTube transcripts, GitHub, Hacker News and the web, ranked by engagement and recency.
• It returns the datasets people name, the scrapers already in use, and the words real buyers use.
• Write the prompt like a data operator: ask for records, source systems, fields and evidence.
• The scout finds the source trail. The pipeline comes after it.

• Claude Code

Claude Code turns the trail into a build sheet before a single row is pulled.

• Claude Code names the exact artifact, its canonical URL, the join keys and the route for each source.
• A probe gate comes first: can we pull 25 rows with stable IDs and a current status?
• Every source gets a cost check before the full run.

• Claude Code

Permits, inspections, licenses, filings and registries become rows, and every row is matched to a company.

• Claude Code pulls the rows and keeps the record ID, the date and the source URL on every one.
• Each record is matched to a company, facility or owner, with the match confidence written down.
• A record with no confident match is dropped, so the workflow never scores the wrong company.
• Sources that work this way:
  • OSHA inspections, for safety and compliance vendors
  • FMCSA carrier records, for fleet, insurance and compliance
  • City business licenses, for restaurant owners
  • MSHA accident and inspection data, for mining safety

• Claude Code

Accounts are ranked on what the record shows, for a person to review.

• Jev and OpenRouter score each account on the evidence in its record.
• The score orders the review queue. It never becomes a claim in the email.

• Jev
• OpenRouter

Only once the account has a reason does anyone look for a person.

• The enrichment waterfall uses IcyPeas, Prospeo, LeadMagic and GetLeads to find the person who owns the problem the record shows.
• Each vendor only sees the contacts the ones before it missed.

• IcyPeas
• Prospeo
• LeadMagic
• GetLeads

Every email is checked twice before anything sends.

• MillionVerifier checks every address first, then BounceBan checks every address that passes.
• An enrichment vendor's own check never counts as one of the two.

• MillionVerifier
• BounceBan

Customers, open deals and closed-lost accounts come off the list.

• Every account is checked in HubSpot before it moves on.
• The reason for each suppression stays on the row.

• HubSpot

Claude Code checks each line against the fields in the row before anyone reads it.

• Claude Code lists what the record supports, what it doesn't, and any word that overstates certainty.
• Nothing implies risk, urgency, savings, legal exposure or intent unless the row supports it.
• An overstated claim is rewritten or dropped.

• Claude Code

A person approves each send, and the opener leads with what you found.

• A Slack approval puts a person on every send.
• Emails in EmailBison lead with the finding, cite only fields in the row, and stay under 45 words.
• One version asks for a correction instead of a meeting.
• The source URL and record ID go into the CRM note.

• Slack
• EmailBison