12 KPIs × 5 Categories
Criteria & Weights
Each KPI carries a different weight depending on the evaluation category. Weights reflect the relative importance of each metric to the category's primary use case. All weights sum to 100% within each category.
Weight Distribution Matrix
| KPI | Text | Visual | Voice | Auth | Academic |
|---|---|---|---|---|---|
| Detection Accuracy | 20% | 20% | 20% | 10% | 20% |
| False Positive Rate | 15% | 15% | 15% | 5% | 20% |
| Model Coverage | 12% | 10% | 10% | 5% | 10% |
| Speed/Throughput | 5% | 5% | 8% | 5% | 3% |
| Transparency | 10% | 8% | 8% | 15% | 8% |
| API Availability | 5% | 5% | 5% | 8% | 3% |
| Privacy Compliance | 8% | 8% | 10% | 15% | 12% |
| Audit Status | 5% | 5% | 5% | 12% | 5% |
| Update Frequency | 8% | 8% | 8% | 5% | 5% |
| Cross-Language | 5% | 5% | 5% | 5% | 8% |
| Accessibility | 3% | 3% | 3% | 5% | 5% |
| Research | 4% | 8% | 3% | 10% | 1% |
Scoring Formula
Raw scores for each KPI are normalized to a 0-100 scale using min-max normalization across the evaluation universe. The overall integrity score is calculated as the weighted sum of normalized KPI scores, using the category-specific weight distribution shown above. Platforms are ranked by descending overall score within each category and across the full index.
Overall Score = Σ (KPIi × Weighti) for i = 1 to 12
where KPIi = (rawi − mini) / (maxi − mini) × 100
where KPIi = (rawi − mini) / (maxi − mini) × 100