Methodology
How we built this, what we trust, and where the limits are.
Approach
This product synthesizes data from 26 primary sources to answer: where is the European job market heading, and what do you need to get there? We prioritize European-sourced data over US proxies.
Every number on the signals page traces to a Tier 1 or Tier 2 source. We don't interpolate, forecast, or model. We report what the data says and flag where it's thin.
Source Tier Ratings
| Tier | Definition | Examples |
|---|---|---|
| Tier 1 | Official statistics or large-sample primary research (n > 1,000). Reproducible methodology. | Eurostat, OECD, Stack Overflow (49K), Ravio (1,500+ companies), LinkedIn (1.3B members) |
| Tier 2 | Industry reports with documented methodology and meaningful sample size. | Atomico SoET, Figma Design Hiring (2,500), Bridge Group SDR (600), ISC2 Cybersecurity (16K), McKinsey Marketing (500) |
| Tier 3 | Expert analysis, qualitative sources, or small-sample surveys. Used for context, not for numbers. | Claude research reports, Indeed Hiring Lab screenshots, Bitkom (152 startups) |
Data Architecture
Role Families
We analyze 9 role families, each mapped to ISCO-08 occupation codes where possible. These are the same families used in the AI Exposure Map (opens in new tab) for cross-referencing.
Role-level data (hiring rates, attrition, salary YoY, seniority) comes primarily from Ravio Compensation Trends 2026, which covers 1,500+ European tech companies. Ravio data is by function, not by ISCO code — the mapping is approximate.
Country Signals
Country-level data comes from 6 sources, creating varying "data depth" per country:
- Eurostat AI adoption (isoc_eb_ai) — 33 countries, 2021-2025
- Eurostat ICT specialists % (isoc_sks_itspt) — 36 countries, 2015-2024
- OECD Employment Protection Legislation — 26 European countries, Version 4 (2019)
- Stack Overflow Developer Survey 2025 — 29 European countries with 50+ respondents
- Eurostat employment by ISCO 2-digit — 35 countries, 2023-2024 YoY
- LinkedIn hiring rate data — 4 countries (DE, FR, UK, NL)
Countries with 6/6 sources: Germany, France, Netherlands. Countries with 5/6: 19 countries including AT, CH, PL, Nordics. The "data depth" badge in the country table indicates how many sources contribute to each country's row.
Salary Data
Ravio salary data is in GBP and represents European tech company compensation. Stack Overflow salary data is in USD (converted by SO using their methodology). These are not directly comparable — Ravio covers tech companies specifically, SO covers all developers regardless of employer type.
Limitations
- Ravio bias: 1,500+ companies skews toward funded tech companies, not the full economy. Hiring rates may not represent traditional employers.
- LinkedIn bias: Over-represents white-collar knowledge workers. Country-level data available for only 4 European markets (DE, FR, UK, NL) plus global aggregates.
- Stack Overflow bias: Self-selected developer community. Salary data is self-reported and unverified. Sample sizes vary significantly by country (DE: 3,025 vs LV: 62).
- Temporal mismatch: Ravio data is Q3 2025, Eurostat is 2024-2025, OECD EPL is 2019, LinkedIn is Oct 2025, SO is June 2025. The page combines these as if they represent "early 2026" but they're snapshots from different moments.
- No salary purchasing power adjustment: SO salaries are in USD but not PPP-adjusted. A $60K salary in Poland has very different purchasing power than $60K in Switzerland.
- AI impact attribution: We report LinkedIn's finding that the hiring slowdown is macro-driven, not AI-driven. But the boundary between macro effects and AI effects is blurring and will become harder to separate.
- Regulatory demand estimates: Scope numbers (NIS2: 160K entities, DORA: 22K) come from regulatory documents but translate to hiring demand imprecisely.
Connection to the AI Exposure Map
The AI Exposure Map (opens in new tab) scores ~130 ISCO occupation groups for technical and regulated AI exposure across 36 countries. This companion product takes the next step: for the roles most people actually hold (PM, engineer, designer, etc.), what does the hiring market look like?
ISCO codes bridge the two products. Each role family here maps to 1-3 ISCO codes. When you click through to the exposure map, you see the AI exposure score for that specific occupation group.
Updates
This is a point-in-time snapshot (early 2026). We plan to update when major new data drops (Ravio H2 2026, Eurostat annual releases, LinkedIn quarterly). The data architecture supports incremental updates without rebuilding the page.