StackData Browse the catalog

Open government & public data

Published
everywhere.
Findable nowhere.

Statistical offices, ministries, agencies and city governments release public data constantly, each portal with its own vocabulary, licences, formats and silences. None of them lists the others. StackData reads them, gives every dataset one category, one coverage range, one cadence and one licence, and puts the result in a single index — so a labour-force survey from one country sits next to the same survey from another.

Browse the catalog Read the method

Classification

Every dataset gets one shelf.

Publishers name things however they like, so the catalog needs a single axis everything hangs from. Where governments already agree on a classification StackData adopts theirs rather than inventing one — the categories below are a statistical scheme its publishers already share, which is why no crosswalk had to be invented for them. Where sources disagree, the mapping is published rather than hidden. The eight largest categories hold 14,773 of the 18,144 datasets indexed so far.

Slug Field Datasets Statistics Share of 18,144

  • Slug GOV Field Administration and Public Finance datasets 4,560 48 statistics
  • Slug JUS Field Justice, Security and Environment datasets 2,428 55 statistics
  • Slug AGR Field Agriculture, Forestry and Fisheries datasets 1,967 74 statistics
  • Slug EDU Field Education, Culture, Sports and Life datasets 1,648 29 statistics
  • Slug ECO Field Business, Household, Economy datasets 1,211 90 statistics
  • Slug SOC Field Social Security and Sanitation datasets 1,054 114 statistics
  • Slug LND Field Land and Climate datasets 1,004 2 statistics
  • Slug POP Field Population and Households datasets 901 21 statistics
All 17 fields Browse the catalog

One record

What an entry looks like.

Every dataset in the catalog carries the same fields, whatever portal it came from and whoever published it — and says so plainly when the source leaves one empty. This is the shape of a record, with the kind of value each field holds.

LAB · 0117 example

Labour force survey, quarterly

National statistics office

Category
Labour and Wages — LAB
Coverage
First and last period the data covers
Cadence
One value, from whatever the source wrote
Shape
Rows and columns, where published
Formats
Every file the record offers
Licence
One licence, however the source spelled it
Last change
When the source last moved
Status
Live, stale, or stopped

What the source did not publish The fields a portal leaves blank are the ones a catalog is tempted to invent. This one records them instead: coverage years unstated, cadence given as free text rather than a code, licence written two different ways across the same record’s files. Every original string stays on the record beside the value StackData stores, so any normalization can be checked against what the publisher actually said.

Method

How a dataset gets in.

Four passes, applied the same way to every source.

  1. 01

    Harvest

    Read each portal on its own terms — an open API, a keyed one, or just a published list — on whatever schedule it keeps.

  2. 02

    Normalize

    Licences, cadences and formats arrive as free text. Collapse each to one value, and keep the published string beside it so the mapping stays checkable.

  3. 03

    Classify

    Give every dataset one category. Where publishers already share a classification, adopt theirs rather than inventing one; where they do not, publish the mapping.

  4. 04

    Verify

    Re-fetch on schedule, mark a series stopped once it goes quiet for longer than its own cadence, and resolve the same dataset appearing in two places.

Read the method in full Normalization

Where the index stands

Early, and honest about it.

The catalog is real but young. National portals are indexed end to end today, with city and agency portals in progress, and every figure on this site is read from a live source rather than estimated. StackData is not a global catalog yet and does not pretend to be one.

  • 01 Datasets indexed 18,144 across 23 publishers
  • 02 Categories in use 17 one shelf per dataset
  • 03 Machine-readable 5.3% carry a CSV; half are PDF
  • 04 Filed with no category 357 kept and marked, not dropped

One index for public data.

Free to read, free to query, licensed for reuse, and nothing behind an account. Every count on this site was read from a source portal in August 2026, and says which one it came from.

Browse the catalog Read the method