1001 Evidence Brief
🔍 Who actually evaluates smart-city projects?
Almost nobody who can compel an answer: of the 1140 projects documented here, 62 cite an audit office or a court — 5.4%. Academic research reaches 22.8% of them, the press carries most of the rest, and 145 entries (12.7%) cite no voice from outside the organisations that built, run or paid for the system.
✔ Every published entry is checked against the public sources it cites before publication.
How far outside the project the record reaches
Examined from outsideReported on, not examinedInside voices only
| Category | Entries | Outside scrutiny | Examined | Reported only | Inside only |
|---|---|---|---|---|---|
| All published entries | 1140 | 302 examined from outside, 693 reported on, not examined, 145 inside voices only | 26.5% | 60.8% | 12.7% |
| Mobility | 435 | 89 examined from outside, 306 reported on, not examined, 40 inside voices only | 20% | 70% | 9% |
| Data & IoT | 257 | 64 examined from outside, 147 reported on, not examined, 46 inside voices only | 25% | 57% | 18% |
| Governance | 173 | 55 examined from outside, 97 reported on, not examined, 21 inside voices only | 32% | 56% | 12% |
| Environment | 157 | 61 examined from outside, 76 reported on, not examined, 20 inside voices only | 39% | 48% | 13% |
| Energy | 100 | 25 examined from outside, 59 reported on, not examined, 16 inside voices only | 25% | 59% | 16% |
| Society | 18 | 8 examined from outside, 8 reported on, not examined, 2 inside voices only | 44% | 44% | 11% |
Society is eighteen entries, so its shares move on one or two projects; the size sits next to every row for that reason. The gap that is large enough to mean something is Environment against Mobility — 1140 entries in all, and the domain most often examined from outside is the one where an air-quality figure can be checked by anybody with an instrument.
Representative entries
- Buenos Aires's Fugitive Facial Recognition System, stopped by the courts One of only 8 entries in the atlas that cite a court. The city's own security law and its two never-appointed oversight bodies are what stopped the system — not the data-protection regulator, and not the press.
- Réseau express métropolitain: a pension fund's driverless metro Québec's auditor general found that public contributions exceed half the fare revenue and that a return on equity is roughly twenty-five years away. No operator publishes that about itself.
- Restriccion Vehicular: Santiago's Vintage-Tiered Plate Rationing The rarest kind: a peer-reviewed evaluation that found the policy works. Research is the only outside institution in its source list — nobody with authority over the scheme has examined it.
- Integrated Command and Control Centre (Safe and Secure Ahmedabad) The archetypal smart-city artefact — an integrated command centre — and everything this atlas knows about it comes from the city that runs it. That is not a hidden failing; it is the ordinary state of the record.
Method
Every published entry in the atlas carries a list of sources. Each source URL is matched against the source registry, which records who publishes behind that address — an organisation, not a document type and not a measure of quality. An entry is counted once per publisher type present in its list.
Three groups, mutually exclusive, so they sum to the total. Examined means the list reaches an audit or ombudsman institution, a court, or a research venue. Reported means it reaches no such body but does reach the press, an NGO or a multilateral institution. Inside voices only means every publisher in the list is the operator, the supplier, a utility, a government body or a platform that hosts what one of them wrote.
Inside the examined group the brief separates accountability — an audit office, an ombudsman or a court — from research. A university studies a system because somebody chose to; an audit office or a court examines it because the system is answerable to somebody. Folding the two together lets 260 research citations stand in for 62 accountability ones, which is the finding rather than a detail.
The figures cover 1140 published entries and 6020 source URLs, 96% of which carry a publisher in the registry. The classification is regenerated on every build; nothing on this page is typed.
What this cannot tell you
- This measures what an entry cites, not what exists. An audit may have been published that the entry does not list; the number is a floor on scrutiny found, not a ceiling on scrutiny performed.
- The absence of an auditor is frequently the honest state of the world rather than a defect. Most systems here are too small, too new or too uncontroversial for an audit office to have been asked, and nothing in this brief treats an operator's own reporting as untrue — it is the right source for what a system does and the wrong one for whether it worked.
- Publisher type describes who publishes and never how strong the evidence is. A newspaper investigation can be more searching than a perfunctory audit; the grouping cannot see that.
- The registry rests on an editorial decision for part of its rows and derives the rest from a curated row on the same host. Its coverage figure and that split are published on the sources page.
- Categories with few entries move on a handful of projects. The table states each group's size next to its shares for exactly that reason.
The slice behind every number
Every figure above is computed at build time from the published data, and the rows it counts are downloadable as one file: who-evaluates-smart-city-projects.json. Recomputing the numbers is a matter of counting that file. The full datasets are on the open data page, under CC BY 4.0.
Where to read next
The sources page lists every publisher this atlas cites and how far each one reaches. How the record ages asks the same question of the calendar — independent evaluation is written years after a system is switched on, so the newest entries in any catalogue rest on the operator's own account. And the evidence readings group the corpus by the grade each entry carries.
All briefs · Who publishes the evidence · How the record ages · The state of the evidence