Methodology
How we build our data
Transparency is core to getCourage. This page explains our data sources, processing pipeline, confidence tiers, and AI risk methodology.
Salary data sources
| Country | Source | Frequency | Tier |
|---|---|---|---|
| United States | H1B LCA filings + BLS OEWS | Quarterly / Annual | DIRECT |
| United Kingdom | ONS ASHE (SOC 4-digit) | Annual | DIRECT |
| Germany | Destatis + Entgeltatlas (KldB 5-digit) | Annual | DIRECT |
| Switzerland | BFS Salarium (ISCO 3-digit) | Biennial | SUPPORTED |
| Netherlands | CBS StatLine (ISCO 2-3 digit) | Annual | SUPPORTED |
| France | Recruitment guides (Robert Half, Hays) | Annual | INFERRED |
Confidence tiers
Every data point in getCourage is assigned one of four confidence tiers. This lets you judge how much weight to give each number.
Data comes from a primary official source (e.g., H1B filings, BLS surveys). Highest reliability.
Data is cross-validated from multiple independent sources that agree within acceptable variance.
Calculated from proxy indicators (e.g., country multipliers applied to a known baseline). Transparent methodology.
Model-based projection where direct data is unavailable. Clearly marked and used sparingly.
Skill demand analysis
Skill demand frequencies come from analysis of 5,099 real job postings collected via the Adzuna API, Arbeitnow, and Remotive. We map each posting to our standardized skill taxonomy (based on ESCO and O*NET) and count mention frequency per role.
Skills are ranked by their salary impact — the correlation between skill presence and compensation level within each role. This helps identify which skills to prioritize for maximum career return.
AI automation risk score
The composite AI exposure score combines three peer-reviewed research indices:
- Eloundou et al. (2023) — 50% weight. GPT-4 exposure scores for occupations based on task-level analysis.
- Microsoft Research (2023) — 30% weight. Occupation-level AI impact assessment considering complementarity.
- Felten et al. (2023) — 20% weight. AI occupational impact based on AI capability benchmarks.
At the skill level, each skill is classified into one of four resilience categories:
- Amplifier — AI makes this skill more valuable (e.g., Python, statistics)
- Resistant — Difficult to automate, retains value (e.g., system design, leadership)
- Vulnerable — At risk of automation (e.g., basic data entry, routine testing)
- Neutral — Minimal AI impact in either direction
Career transition paths
We model 240 directional career transition paths (16 roles × 15 targets). Each path is derived from O*NET skill adjacency data, expert knowledge, and similarity analysis across role skill profiles.
Timeline estimates are based on skill gap size and learning complexity. As users report actual transitions, these estimates will be refined with real-world data.