Credibility Evidence Selector

Scenario: B2B security startup with a credentials wall Trigger: Founder says "our homepage lists 14 analyst quotes and 40 customer logos but nobody books a demo. How do I make the security claim more credible?" Process: (1) Lock claim: "Our system blocks every known credential-stuffing attack pattern." (2) Inventory — 14 entries under External Authority, 2 under Vivid Details (an incident-response write-up mentioning a specific attacker TTP), 1 under Sinatra (they run security for a top-3 global bank's consumer login). (3) Rank: the bank reference is Sinatra-grade; testable credentials missing; authority dominates inventory (credentials wall). (4) Sinatra Test on the bank reference — Hardest case? Yes (top-3 global bank, consumer login = highest-volume attack surface). Legibly hard? Yes (audience instantly recognizes). Verifiable? Partial — the bank will not be named publicly. VERDICT: PASS with anonymization ("a top-3 global bank, consumer login, 400M accounts, zero successful credential-stuffing incidents in 18 months"). (5) Diagnostic flags credentials-wall (14 analyst quotes). Fix: cut to 3. (6) Recommendation: lead with the anonymized bank Sinatra hero; back with two incident-response details; cut 11 of 14 analyst quotes. Output: credibility-plan.md with the bank Sinatra as lead, the quote wall shrunk, and a note that the founder should push for a testable credential (open benchmark or public red-team) to upgrade the credibility stack further.

Scenario: Nonprofit fundraising email, abstract statistics Trigger: Fundraiser has a draft saying "3.2 million children are affected by food insecurity each year" and asks "why isn't this landing?" Process: (1) Lock claim: "Child food insecurity is a scale emergency in our region this year." (2) Inventory — 1 raw statistic (the 3.2M number), 1 External Authority (a public-health study), 0 Testable, 0 Sinatra, 0 Antiauthority, 0 Vivid Details. (3) Rank: only two entries, both weak; the statistic is a Raw Statistic failure mode and needs illustration. (4) Sinatra Test: no hero candidate exists. Do NOT fabricate. Instead: convert the statistic. "3.2M children = every elementary-school seat in the five largest school districts in the country, empty at breakfast." (5) Diagnostic: Raw Statistic HIT. Fix: the BB-translation above. Also: add one Antiauthority — a single named family from the region, their story, their specific morning (vivid-details pairing). (6) Recommendation: lead with the illustrated statistic AND the one-family Antiauthority story side-by-side (Mother Teresa effect — identifiable victim paired with illustrated scale). Cut the authority-study citation to a footnote. Output: credibility-plan.md recommending a two-piece lead (illustrated stat + identifiable family) with a note that the fundraiser should recruit an antiauthority spokesperson BEFORE next campaign.

Scenario: Academic publishing a counterintuitive research finding Trigger: Researcher has a finding that contradicts 20 years of prior consensus. Asks "how do I make this credible to reviewers who will dismiss it on sight?" Process: (1) Lock claim: "Finding X contradicts prior consensus Y because of mechanism Z." (2) Inventory — External Authority is dominant (peer-reviewed replication); Testable Credentials possible (open data + reproducible pipeline); one Antiauthority candidate (a respected prior defender of Y who reversed position); no Sinatra hero. (3) Rank: override default — audience is credentialed peers, so External Authority upgrades to #1. But lead with the Testable Credentials (open data) because a reviewer who can reproduce the result in an hour is more convinced than one reading a methods section. Antiauthority (the reversed defender) comes next — a Barry-Marshall-pattern Sinatra hero by analogy: someone who publicly reversed position is the "hardest case" this audience responds to. (4) Sinatra Test on the reversed defender: Hardest case? Yes (they were a public critic). Legibly hard? Yes (the reversal is documented). Verifiable? Yes. VERDICT: PASS. (5) Diagnostic: no credentials wall, no raw-stat failure. (6) Recommendation: lead with the open-data link (testable), follow with the reversed-defender statement (Sinatra-pattern antiauthority), support with the peer-reviewed replication. Cut nothing. Output: credibility-plan.md with a testable-credential lead, a named Sinatra-pattern antiauthority, and a note that the open dataset should be ready BEFORE submission — the credibility plan depends on it.