A schema for crowdsourced café prices

Free-text submissions are easy to collect and a pain to analyse. A small, strict schema up front saves days of clean-up later.

{
  "country": "HU",          // ISO 3166-1 alpha-2
  "city": "Budapest",
  "cafe_type": "regular",   // regular | specialty | chain | hotel
  "price": 1290,
  "currency": "HUF",        // ISO 4217
  "size_ml": 180,
  "observed_on": "2026-10-14"
}

Validation rules

• Reject prices more than five median absolute deviations from the city median and queue them for review.
• Normalise city names against a reference list, so "Bp" and "Budapest" count as one place.
• Keep the raw submission next to the cleaned row, because you will want to re-run the cleaning later.

Wages live in a separate table keyed by country and year, and are joined in at analysis time.