TrustInsider Methodology v3.1

Every TrustInsider score is computed from public, machine-readable sources. We do not scrape sites whose Terms of Service forbid it (LinkedIn, Levels.fyi). We do not use paywalled datasets as primary sources. When data is unavailable, we declare ND  rather than estimate.

Each component of every composite score carries a Confidence label (ALTA / MEDIA). Components with intrinsically low confidence do not enter numerical scores — they are presented as qualitative observations only.

• SEC EDGAR API — financial filings, segment reporting, risk factors
• USPTO PatFT/AppFT API + Google Patents BigQuery — patent data
• BLS Employment Multipliers I/O tables — indirect employment estimates (USA)
• Eurostat I/O tables, BEA Industry Economic Accounts — economic multipliers
• Yahoo Finance API, Alpha Vantage — historical stock data
• GitHub REST API + GitHub Archive (BigQuery) — open source contributions
• OpenSecrets, EU Transparency Register, UK House of Commons Register — lobbying spend
• EPA ECHO, EU Industrial Emissions Portal — environmental violations
• OSHA Establishment Search, NLRB Case Search — labor disputes
• FTC Enforcement Actions, EU Commission DG Competition — antitrust history
• NHTSA, FDA, FCC, EU RAPEX — recall databases
• CDP Climate, GRI Database — sustainability disclosure
• UN Global Compact participants registry — UNGC verification
• Brand Finance Top 500, Interbrand Top 100 — verified brand value
• Trustpilot, Glassdoor — public reviews
• SimilarWeb, Ahrefs, Semrush — public-facing digital data
• Crunchbase free tier — basic company data
• Wikidata SPARQL — identity facts (CC0)
• HackerOne directory, AICPA SOC report listing, B Corp directory — certifications
• ProPublica Nonprofit Explorer — 990 forms (philanthropy)

The headline Trust Score is the weighted average of four equally-weighted pillars (25% each), minus a capped penalty layer. Each pillar is a composite of deterministic, source-grounded leaf indices; missing leaves redistribute their weight across the present ones — never zero-filled.

• Affidabilità (25%) — Mantiene le promesse?
• Reputazione (25%) — Le persone si fidano?
• Trasparenza (25%) — Si comporta correttamente?
• Solidità (25%) — Sarà affidabile anche domani?

Penalty layer (cap 30): severe negative-signal indices (reputation event shock, product recall, sanctions exposure, regulatory compliance track, enforcement track) subtract from the weighted pillar average instead of being diluted — so a verified scandal bites the headline. The penalty is shown only where a real penalty_applied driver exists (hidden otherwise).

Verdict bands:

• LEADER (90–100) — Punto di riferimento del settore
• TRUSTED (75–89) — Azienda altamente affidabile
• GOOD (60–74) — Complessivamente positiva ma con limiti
• WEAK (40–59) — Fiducia fragile / problematiche rilevanti
• RISK (0–39) — Rischio elevato / fiducia compromessa

5-component formula. Verified sources only.

Crisis Track Record and Talent Stability were removedfrom the index in v2 because their proxies are too subjective or require ToS-blocked sources (LinkedIn, Levels.fyi). Stress scenarios are presented as qualitative observations, not score components.

Sum of weighted positive minus weighted negative components. All components from verified registries.

SDG mapping is limited to verifiable UN Global Compact participation; we do not invent ad-hoc SDG scores.

Pricing Power Index = (gross_margin − sector_median_margin) / sector_stdev, normalized 0-100. Source: SEC 10-K + S&P sector aggregates.

Hamilton Helmer's 7 Powers framework is presented as a qualitative narrative summary  — not a numerical score, because the classification is intrinsically qualitative.

Industry HHI is computed only for the ~30 sectors with public top-10 market share data (smartphone, cloud, browser, EV, etc.).

• No scraping of LinkedIn, Levels.fyi, or other sites whose ToS forbid it.
• No estimated brand value for companies absent from Brand Finance / Interbrand lists.
• No invented financial figures when the company hasn't disclosed them.
• No subjective qualitative scoring smuggled into composite numerical indices.
• No retention of user data beyond what's required for service delivery.

Every numerical score on a TrustInsider report displays the confidence level of each component (ALTA / MEDIA). Components with intrinsically low confidence do not enter the numerical score — they appear as qualitative observations or as ND.

This page is versioned. Changes to scoring formulas trigger a methodology version bump, displayed at the top of this page. Major methodology revisions are also referenced in each report's sources section so historical reports remain auditable.

Each composite index has a dedicated methodology page with formula, sources, and changelog. Raggruppati per famiglia.