Methodology
How percentiles are calculated
Most of our tools use a normal distribution (bell curve) model with published mean and standard deviation values. Your input is converted to a z-score, then mapped to a percentile using the cumulative distribution function (CDF).
z = (your_value - population_mean) / standard_deviation
percentile = normCDF(z) × 100
We use the Abramowitz & Stegun approximation (1964, equation 7.1.26) for the normal CDF, which is accurate to 1.5 × 10⁻⁷.
Reverse-scored items
In quiz-style tools, some items are phrased negatively (e.g., "I feel useless at times"). These are reverse-scored: a response of 5 is counted as 1 and vice versa. This is standard practice in validated psychological instruments. Reverse items are marked with data-r="1" in the source code.
Multi-dimensional scoring
Some tools (Big Five, Dark Triad, Enneagram) measure multiple dimensions. Items are tagged with dimension codes (e.g., data-d="O" for Openness) and scores are calculated per dimension separately.
Clinical screening tools
Several tools are based on validated clinical instruments:
| Tool | Scale | Reference |
|---|---|---|
| Depression | PHQ-9 | Kroenke et al. 2001 |
| Anxiety | GAD-7 | Spitzer et al. 2006 |
| Social Anxiety | LSAS (adapted) | Liebowitz 1987 |
| OCD | Y-BOCS (adapted) | Goodman et al. 1989 |
| Bipolar | MDQ | Hirschfeld et al. 2000 |
| PTSD | PCL-5 (adapted) | Weathers et al. 2013 |
| Self-Esteem | Rosenberg RSES | Rosenberg 1965 |
| ADHD | ASRS | Kessler et al. 2005 |
| Autism | AQ-10 | Baron-Cohen et al. 2001 |
| Insomnia | ISI | Morin et al. 2011 |
| Big Five | BFI/IPIP | Goldberg 1993 |
Important: Our implementations are adapted versions of these scales — shorter and simplified for self-screening. They are not diagnostic tools. A high score means "consider talking to a professional," not "you have this condition."
Lifestyle and comparison tools
Non-clinical tools (salary, rent, sleep hours, etc.) use published population data:
- Government surveys: BLS Consumer Expenditure Survey, OECD datasets, WHO Global Health Observatory, CDC NHANES, Eurostat
- Central banks: Federal Reserve Survey of Consumer Finances (SCF)
- Research orgs: Gallup World Poll, Pew Research Center, National Sleep Foundation
Where direct percentile data is not available, we model the distribution using reported mean and standard deviation values from the most recent available data.
Limitations
- Normal distribution assumption: Not all traits follow a bell curve. Income, for example, is right-skewed. We use the best-fit model available.
- Single-population norms: Most norms are US or Western-centric. Percentiles may not apply equally across all cultures.
- Self-report bias: All inputs are self-reported. People tend to overestimate socially desirable traits and underreport stigmatized ones.
- Simplified scales: Clinical tools are abbreviated. Full-length versions with professional interpretation are more accurate.
- Data currency: Population norms change over time. We use the most recent published data available at time of development.
Open source verification
Every calculation is implemented in client-side JavaScript and is fully inspectable. View the source code on GitHub to verify the mean, standard deviation, and scoring logic for any tool.