Methodology
From raw job ads to useful salary context. Transparent rules, reproducible calculations, and a clear publishing threshold.
Our goal is not to publish as many salary pages as possible. We want to publish numbers whose origin and meaning are easy to understand.
Current status: pilot. The current pilot publishes only pages with at least 30 observations from approved sources. City, skill, and comparison pages are published only when their own data exists. The method below describes the full production pipeline.
1. Data collection
Planned daily imports read public APIs, approved job feeds, and official statistics. Only job title, city, salary range, date, source, and relevant skills are stored. No applicant profiles or personal data are collected from job ads.
2. Cleaning and normalization
Duplicate ads are merged. Job titles are mapped to a controlled role list. Hourly, monthly, and annual pay is normalized to annual gross euros. Currency conversions and working-hour assumptions are versioned.
- Outliers and contradictory values are quarantined for review.
- Bonuses, equity, and benefits in kind are excluded.
- Minimum salaries are labeled and not treated as actual pay.
3. Calculation
The median is the middle value of the sorted sample. The 25th percentile marks the lower bound and the 75th percentile the upper bound of the middle 50%. Experience groups are inferred from job-ad requirements. Trends compare like-for-like monthly samples.
4. Publishing rules
- 30 data points minimum: Smaller groups show “Not enough data yet”.
- Monthly recomputation: Imports run daily; published benchmarks are recomputed monthly.
- Unique context: Every page contains its own values, sample size, sources, and data-based commentary.
- Reviewable changes: Every publication is a Git commit and can be rolled back.
Limitations
Advertised pay is not the same as actual pay. Employers that disclose pay may differ from those that do not. Self-reports can be biased. Salary data is orientation, not a guarantee or individual compensation advice.