HumanAverage

humanaverage.com answers one question per page: where does a country sit, and where do you sit inside it? Each page takes a single published statistic — the age at which half of young people have left home, household wealth — and turns it into a map of Europe, a sortable table of every country, and a comparison with whatever the reader enters about themselves.

Every figure comes from Eurostat, the OECD or the ECB. Nothing on the site is estimated on top of what those sources publish, and no number is typed into a page by hand: the text is calculated from the data at build time, so it cannot go stale when a source updates.

How it is built

Astro with static output, no client-side framework, and exactly one dependency. The data pipeline is plain Node.

npm install
npm run dev                  # http://localhost:4321
npm run build                # static build into dist/, then the sitemap
npm run preview              # serve the build locally

Each data source has three steps — fetch into data/raw/ (git-ignored), build a compact JSON file that is committed, validate it:

npm run leaving              # Eurostat: leaving the parental home
npm run wealth               # ECB: household wealth
npm run geo:build            # Natural Earth: the map geometry

The gate

Every dataset has a scripts/validate-*.mjs that must pass before anything is published. It does not trust the built file: it recomputes each published figure from the raw API response, checks that aggregates like EU27 were not counted as countries, that each country carries its own reference year, and — for Eurostat — that no comparison spans a break in the series.

That last one matters more than it sounds. Eurostat reorganised the Labour Force Survey in 2021 and flags the break; the break year itself is transitional. Measured from 2005 the spread between countries appears to grow, and measured across comparable years it shrinks. The code refuses to calculate the first version rather than footnoting it.

Where each dataset comes from, what it covers and what its licence allows is recorded per source in data/sources.md.

What is where

Path
src/pages/	The English site
src/components/, src/lib/	Map, table, hero and the calculations behind them
scripts/	Fetch, build and validate, one set per source
data/	Committed JSON; data/raw/ is ignored
public/nl/, public/tools/, public/articles/	A frozen Dutch set of browser-based reaction, typing and memory tests. Still served, no longer developed
research/	How the metrics were chosen and how the design was arrived at

Publishing

Push to main; Vercel builds and deploys. Before that: the data gate passes, npm run build succeeds, and the page has been looked at in a browser.

No advertising code is on the site. When it is added, the consent banner and the privacy policy go first.