UN Data Commons platform team
10 questions still open. The team that runs the platform. Questions addressed to them are things only they can answer — roadmap, sanctioned paths, and the structural facts we have found in the graph.
What a city can say in the UN's data, what it cannot, and what we are asking the platform team.
New here? Start here How this site works, what the questions in the margins are for, and how to take part — whether you have a URL or a git clone.New York City published the world's first Voluntary Local Review of the SDGs in 2018. It is a PDF, produced every few years, by a small team. Meanwhile NYC Open Data and the Mayor's Management Report publish the same underlying indicators continuously, and the UN System Data Commons now holds the authoritative global series for the same concepts. Nobody has connected the two for the person who actually needs it — the analyst at DOT, DOHMH, HPD or OMB writing a budget justification, a council testimony, or a program review.
This project is a live Voluntary Local Review workbench in three layers: an open indicator crosswalk carrying a written comparability grade on every mapping; benchmark views that put the caveat on the face of the chart rather than in a footnote; and an agent that answers a staffer's question with a citable card — number, source, vintage, method, caveat.
The finding that makes the case: against the United States, NYC's homicide trend reads as a success story. Against every reporting country, it sits in the bottom half. Same number, different comparator, opposite conclusion — which is why a grade and a caveat are not decoration.
The scope, measured rather than recalled. The SDG goal framework holds 689 base indicators across 6,025 variables. Walked from the graph's own root instead, the governed corpus holds 1,661 — so the goal framework is a minority of what is there, which is a fact about the platform rather than about this project.
It is no longer only a proposal. NYC's own series now run in a custom Data Commons instance at commons.databook.nyc — 16 variables and 572 observations, each read back out of the running instance and matched against the CSV that produced it. It settled the question the pitch depended on: sub-annual data survives, so the frequency NYC has and a national graph does not is not flattened by the platform.
It also produced the project's argument without the UN graph in it. Homicide is the one concept where the instance holds both our series and an independent publisher's for the same city, and across 16 overlapping years NYPD and the FBI agree to +0.4% — until 2023, when the sign flips and the gap triples to −10.4%. Same concept, same city, same year, two official sources, and nothing on a chart tells you which you are looking at. Asked to draw them together, the platform helpfully labelled the FBI's counts with our rate's unit.
One Hetzner box, €42 a month; a team's tool rather than public infrastructure.
⚠️ Every automated check in this project has, at least once, produced a confident answer that was wrong, and each time a person reading the output for five minutes caught it. That is the project's thesis and its own failure mode at the same time, and it is why the questions in these documents are numbered and published rather than settled quietly. The figures on this page are computed from the probe caches, not typed — for the same reason.
10 questions · 10 open · 0 resolved · 66 documents
Every figure on this page is computed by the build. None is typed, so none can rot.
UN Data Commons platform team
10 questions still open. The team that runs the platform. Questions addressed to them are things only they can answer — roadmap, sanctioned paths, and the structural facts we have found in the graph.
The documents being worked on, and what each is still waiting for.
8 sections · 0 questions
2 sections · 7 questions
6 sections · 0 questions
Finding: data quality across the graph
13 sections · 3 questions
Finding: the electoral data gap
9 sections · 0 questions
1 sections · 0 questions
Addendum: errors in the UN data
5 sections · 0 questions
The tokens this site is built from, read from the stylesheet.
2026-09-21 · Devin Balkind
2026-09-18 · Devin Balkind
2026-09-18 · Devin Balkind
Answers come from what is published here. It cannot change anything. This conversation is stored and the team can read it — kept for 90 days.