For newsrooms
Data journalism partnerships.
Some stories only appear once the same question has been put to hundreds of government bodies at once. If your desk has the question, we can do the collection.
UnGovr is a nonprofit. We keep a registry of 330,000+ government entities across 200+ countries, crawl what they publish, file records requests where publishing stops, and return structured data with a source on every row. The topic does not have to be about government. It has to be answerable by government: permits, registers, contracts, inspections, budgets, licences. One agency is a phone call. Four hundred agencies in nineteen countries is the part we own.
Not in the list means we have not started there yet. Write to us anyway: that is how a country gets started.
What we hold
What we could do there
What we bring
Labelled, so you can tell what exists.
Live is running today. Pilot is real but limited. In development is not something we can offer you yet.
-
LIVE
The registry to hang your rows on
Stable IDs for 330,000+ entities, cross-referenced to GeoNames, Wikidata, OCD and FIPS, so a value stays attached to a body that renames or merges. Free through the API and bulk downloads.
-
LIVE
Collection at scale, within the rules
Our crawler identifies itself and respects robots.txt. It never bypasses a CAPTCHA; where a site refuses us we file a records request instead. Slower, and it survives being written about.
-
LIVE
Documents turned into data
OCR, transcription, translation. It runs on hardware we operate, so nothing you share reaches a third-party model service.
-
LIVE
The disclosure law, already looked up
398 records laws across 140+ countries: deadlines, fees, residency rules, exemptions, appeal routes. Free at law.ungovr.org/records.
-
PILOT
A public view built around the story
A dataset for your desk, and where it helps a public view that outlives the article. Six built for ourselves, none yet with a newsroom, so treat it as a pilot you would be early to.
-
IN DEVELOPMENT
Batch filing and response tracking
One template against a target set of agencies, with deadline monitoring across the set. We run this internally; the partner-facing version is being built.
How it runs
Four steps, and the second one stops most of them.
- You bring the question. We say which bodies hold the answer and roughly how many there are.
- You get the feasibility in writing before anyone commits: what is published, what needs a request, who will refuse, and the realistic floor on coverage.
- We collect, and you watch progress against the target list.
- You get the dataset, the gaps, and a source on every value. You publish first.
Worked examples
Six of these, already running.
UnGovr Monitors are this shape when we are the ones asking. Each began as one question put to a large number of agencies. The topics are chosen for how the data behaves, never for a position on the subject.
-
watchwatch.org
Which surveillance technology has each agency deployed? 14,104 on record.
-
computecompute.org
Where is government compute infrastructure? 1,523 facilities.
-
agegating.org
What does each age-verification law require? 250 jurisdictions.
-
sheriffoversight.org
Does civilian oversight of the sheriff exist? 3,062 counties.
-
caoversight.org
What oversight structures exist across California government? 22,478 reports.
-
civilgrandjury.org
What have California's civil grand juries found? 13,082 reports.
Where we stand
Nonpartisan by method, not by promise.
What we do
- Choose topics on whether the data can be collected the same way everywhere, not on where the answer lands politically.
- Collect the whole population of bodies in scope, and publish the coverage gap beside the data.
- Publish the method: what was asked, of whom, under which law.
- Let the finding be that nothing is wrong, when it is.
What we do not do
- Stories about a party, a campaign, an election contest, or an individual politician.
- Shape a collection to reach a conclusion someone already wrote.
- Collect in a way that breaks the law where it is collected.
- Ask for copy approval, an advance look, or a say in what you conclude.
Embargo and licensing
You break it. Then everyone gets it.
- Embargo
- We will consider a partnership under news embargo. While it holds the dataset stays with your desk: not published by us, not offered to another newsroom. The lift is a date, agreed in writing before collection starts.
- After the lift
- Our preference is open publication under Creative Commons Attribution 4.0, crediting your newsroom and UnGovr. We ask for a credit and a link and nothing else. There is no fee.
- Where we say no
- Exclusivity that keeps a public-interest dataset closed indefinitely, or a collection method that would not survive being described in the article. If open release will not work for your story, tell us why at the start.
Tell us the question.
One sentence about what you are trying to count and where you think it lives. A first reply is usually a plain yes, a plain no, or a note on what exists.