Onbrd Docs — Scoring

Scoring Methodology

0-100 Deterministic Score

Each audit produces a deterministic score from 0 to 100, where 100 represents perfect onboarding experience and 0 indicates critical issues that prevent user success. Our scoring algorithm is fully deterministic and reproducible - the same input will always produce the same score.

Weights v1

Our deterministic scoring algorithm uses weighted categories to calculate the final score:

• Signup friction: 25%
• Activation path: 30%
• Copy & clarity: 20%
• Momentum: 15%
• Technical: 10%

Formula (simplified): Score = Σ(weight_i × category_i_subscore). Each category subscore averages its checks, with heavier weight on critical blockers.

Transparency: We publish weights and update them in the changelog. Reports include the build ID and scoring version. All scores are deterministic and reproducible across multiple runs.

Benchmarks v0.2 Metrics

Our expanded benchmark corpus (60 fixtures) validates scoring accuracy:

• Macro-F1 Score: ≥ 0.75 for boolean checks
• Calibration R²: ≥ 0.70 (predicted vs ground-truth)
• Out-of-band: ≤ 10% per category
• Fixture Coverage: 5 categories × ~12 fixtures each

See benchmarks documentation for detailed validation results and integrity checks documentation for information about robustness validation and falsification testing.