Termsparency

Methodology

How Termsparency scores privacy.

Termsparency evaluates published privacy disclosures using a consistent, versioned methodology. Each completed analysis produces one deterministic privacy score based on category-level findings and the methodology version used to evaluate them.

What we evaluate

13 categories of privacy practices.

Each analysis evaluates the company's published disclosures across the same privacy categories. Categories are weighted according to their importance in the methodology.

Data Collected

8%

Sensitive Data

12%

Location

7%

Tracking

7%

Advertising

10%

Third-Party Access

12%

Data Sale and Sharing

12%

AI Training

7%

Profiling

6%

Data Retention

6%

Account Deletion

3%

Data Deletion

6%

Children’s Data

4%

Category weights total 100%. Only the Children's Data category may be classified as not applicable. When it is, its weight is excluded from the score denominator.

How findings are classified

Disclosure determines how each category is scored.

Each category receives one of three classifications. Missing, vague, or incomplete disclosure does not earn favorable privacy credit.

Applicable

The policy substantively addresses the category. It receives a rating from 0 to 4 based on the protections explicitly established by the policy and contributes to the privacy score.

Not Disclosed

The policy does not substantively address the category. It receives a rating of 0 and its full category weight remains in the score denominator.

Not Applicable

This classification is permitted only for Children's Data when the policy affirmatively establishes the methodology's non-applicability conditions. Its weight is excluded from the score denominator.

Important: Policy silence is never treated as not applicable. If a category is not substantively disclosed, Termsparency assigns it a rating of 0 rather than assuming a privacy-protective practice.

How the score is calculated

One deterministic score.

Every completed analysis produces one privacy score from 0 to 100. There are no provisional scores or possible score ranges.

Category Rating

Applicable categories receive a rating from 0 to 4. A rating of 4 earns the category's full weight; lower ratings earn a proportional share.

Not Disclosed

A category that is not substantively disclosed receives zero points while remaining in the score denominator.

Not Applicable

When Children's Data is validly not applicable, that category's weight is excluded from both the earned points and score denominator.

Score formula

Each applicable category earns its weight multiplied by its rating divided by 4. Termsparency divides the total points earned by the total applicable weight and multiplies the result by 100.

Disclosure coverage

Termsparency also reports disclosure coverage: the percentage of applicable methodology weight that the policy substantively addresses. Disclosure coverage provides context about the breadth of the policy's disclosures, but it is not a separate score.

Evidence and transparency

A score should be traceable.

Termsparency is designed so published results can be traced back to the policy and methodology used to produce them.

Source Evidence

Findings can include excerpts from the analyzed policy so users can inspect relevant language supporting an evaluation. Evidence helps explain and trace a finding but is not itself a publication requirement.

Policy Versions

Findings are tied to the specific policy version analyzed, including its publisher-stated effective or updated date, source, and when the policy content was captured for analysis.

Methodology Versions

Published scores identify the methodology version used so changes to the scoring framework can be distinguished over time.

Score History

Published analyses can be preserved as historical records, allowing changes in company disclosures and future scores to remain visible.

AI interprets the policy. The published rubric judges. Code calculates the score.

Termsparency separates policy interpretation from deterministic score calculation so the published methodology—not the model itself—determines how category findings translate into a privacy score.