Our Scoring Methodology

How we evaluate AI tools for regulated industries

Why Governance Scoring Matters

Regulated industries face unique challenges when adopting AI tools. Compliance requirements, data privacy concerns, and the need for explainability demand a rigorous evaluation process. Traditional software reviews don't capture these critical dimensions.

GovernAtlas provides a standardized framework for evaluating AI governance. Our methodology is built on industry standards, regulatory requirements, and best practices from healthcare, financial services, legal, and government sectors.

Every tool in our directory is scored across five dimensions, providing buyers with comparable, objective data to inform their procurement decisions.

Governance Score Overview

Each tool receives a Governance Score from 0 to 100, calculated as a weighted average of five dimensions. Here's how to interpret the scores:

90-100
Excellent
Exceeds industry standards
75-89
Good
Meets industry standards
60-74
Moderate
Basic governance practices
0-59
Needs Work
Significant gaps

Five Scoring Dimensions

Security (25%)

Evaluates the security controls and certifications in place to protect data and systems.

What We Evaluate

  • Industry certifications (SOC 2, ISO 27001)
  • Encryption standards (at rest and in transit)
  • Access controls and authentication
  • Penetration testing frequency
  • Incident response procedures
  • Security audit history

How It's Scored

Points are awarded for each verified certification and documented security practice. SOC 2 Type II earns more points than Type I. Regular third-party audits increase the score.

Data Sources

Vendor trust pages, SOC 2 reports, public documentation, security certifications

Transparency (25%)

Measures how clearly the vendor communicates about their AI systems and practices.

What We Evaluate

  • Model documentation (model cards)
  • Decision explainability features
  • Training data disclosure
  • Limitation acknowledgment
  • Update and changelog practices
  • Public technical documentation

How It's Scored

Points for published model cards, explainability features, clear documentation about capabilities and limitations, and transparent communication about data usage.

Data Sources

Technical documentation, published papers, product features, public statements

Fairness (20%)

Assesses efforts to ensure AI systems treat all users equitably.

What We Evaluate

  • Bias testing procedures
  • Demographic parity measures
  • Third-party fairness audits
  • Remediation procedures
  • Diverse training data practices
  • Ongoing monitoring for bias

How It's Scored

Points for documented bias testing, third-party audits, published fairness metrics, and active bias mitigation programs.

Data Sources

Published audits, vendor disclosures, academic partnerships, fairness reports

Privacy (20%)

Evaluates data privacy practices and regulatory compliance.

What We Evaluate

  • Privacy framework compliance (GDPR, CCPA, HIPAA)
  • Data minimization practices
  • Consent mechanisms
  • Data retention policies
  • User data rights support
  • Data processing agreements

How It's Scored

Points for verified compliance with privacy regulations, clear privacy policies, documented data handling procedures, and available data processing agreements.

Data Sources

Privacy policies, compliance certifications, data processing agreements, regulatory filings

Accountability (10%)

Measures governance structures and oversight mechanisms.

What We Evaluate

  • Audit trails and logging
  • Human oversight mechanisms
  • Error correction procedures
  • Regulatory compliance history
  • Governance committee/board
  • Incident reporting processes

How It's Scored

Points for documented oversight mechanisms, clean compliance history, available audit logs, and clear accountability structures.

Data Sources

Product features, regulatory filings, news/press, governance documentation

Verification Process

1

Data Collection

We gather information from vendor trust pages, certification bodies, public documentation, and direct vendor submissions.

2

Verification

Analysts verify certifications with issuing bodies, review documentation, and cross-reference multiple sources.

3

Ongoing Monitoring

Scores are reviewed quarterly. We monitor for certification changes, security incidents, and significant updates.

Verified vs. Self-Reported

Verified: Confirmed through official sources, carries full weight in scoring
Self-Reported: Vendor-submitted, pending verification, carries reduced weight

Frequently Asked Questions

How can vendors improve their score?

Vendors can improve their score by obtaining relevant certifications (SOC 2, ISO 27001, HIPAA), publishing clear documentation about their AI systems, conducting and publishing bias audits, implementing robust privacy practices, and establishing clear accountability mechanisms. We recommend starting with the dimension where you score lowest.

How often are scores updated?

Scores are reviewed quarterly. We also update scores when vendors notify us of new certifications, when we discover significant changes through our monitoring, or when users report inaccuracies. Major changes (like new certifications or security incidents) trigger immediate reviews.

Can vendors dispute their score?

Yes. Vendors can submit a dispute through their vendor dashboard or by contacting us directly. Disputes should include documentation supporting the requested change. We review all disputes within 10 business days and will update scores if the evidence supports a change.

Is the scoring automated or manual?

Scoring is a combination of both. We use automated tools to gather publicly available information and verify certifications. Human analysts review the data, assess qualitative factors, and make final scoring decisions. All scores are reviewed by at least two analysts before publication.

What's the difference between verified and self-reported data?

Verified data has been confirmed through official certification bodies, public records, or direct vendor documentation we've reviewed. Self-reported data comes from vendor submissions that we haven't independently verified. Verified data carries more weight in scoring calculations.

Do you accept payment for higher scores?

No. Our scoring is completely independent of any commercial relationship. Paid listings may receive enhanced visibility or features, but they have no impact on governance scores. Our methodology is applied consistently regardless of vendor relationship status.

Questions about our methodology?

We're happy to discuss our scoring approach with buyers and vendors.

Contact Us