Embarrassment Is a Public Good

Twice a year, New York's transit agency tells its own board a number that makes it look bad: its battery-electric buses break down roughly six times as often as the diesels they replace. Four other big American cities bought electric buses in the same years and published no comparable figure at all. So here is a question with an uncomfortable answer: which of those five fleets has taught the rest of the world anything? This essay is about why the embarrassing number is the valuable one — and why the willingness to publish it may be the most underrated piece of smart-city infrastructure there is.

The number nobody else published

Start with the buses, because the contrast is unusually clean. In June 2025, New York's MTA reported to its board that its battery-electric buses averaged 2,111 miles between major failures, against 13,000 miles for the diesel buses they replace, and were available for a full day of service 60% of the time, against 91% for diesel. Those are not numbers a press office volunteers. They are numbers a measurement culture produces: mean distance between failures and fleet availability are standing performance indicators at the MTA, reported on schedule whether they flatter or not.

Now look at what the numbers did. In July 2026 the agency trimmed an order of 380 buses to 295 and swapped 85 battery buses for diesel hybrids. That retreat was possible because the metric existed — and every other transit agency on Earth got the lesson for free: reliability, not range, is the thing to write into the next procurement contract.

When this atlas surveyed the ten biggest US metro areas, it found four other operators — in Philadelphia, Chicago, Atlanta and Dallas — that bought battery buses in the same years and published no ongoing performance data about them: no telematics, no availability figures, no evaluation. Their entries were rejected from this atlas not because the buses aren't real, but because there is nothing to check. Meanwhile Shenzhen's 16,000-bus fleet went fully electric with the process documented in World Bank and World Resources Institute case studies that other cities cite to this day. The dividing line is not the technology and not the continent. It is whether anyone is obliged, or willing, to publish.

What one honest evaluation is worth

The strongest counterpart sits on another continent, two decades back. When Medellín strung the Metrocable up to its hillside barrios in 2004, the city did something that cost it nothing technically and everything politically: it left the data open. Crime statistics, survey waves, neighbourhood boundaries. Because of that, a team of epidemiologists could later run a natural experiment, comparing the neighbourhoods the first line reached with carefully matched ones it didn't. The result became one of the most-cited findings in urban policy: the homicide rate fell 66% more in neighbourhoods served by the first Metrocable line than in comparable control neighbourhoods, and resident-reported violence dropped 75% more.

Note what the transparency bought. Not certainty — the study covers one line, and the cable car came bundled with library parks and school investment, so nobody can cleanly separate the hardware from the social programme around it. What it bought was a free experiment for every other city: a measured, contestable, replicable data point that planners from Rio to La Paz have used ever since. A city that publishes its evaluation donates an experiment to the world. A city that publishes a press release donates a sentence.

When the data says no — and that is the point

The deepest value of published measurement shows up when the answer is unwelcome. Mexico City has run its Hoy No Circula driving ban since 1989, and because the city's air-monitoring data is public, economists could test it — three times. The first study concluded that across pollutants and specifications there is no evidence that the restrictions have improved air quality; a follow-up found the Saturday extension of the ban equally ineffective, and a third study documented households responding to restrictions by acquiring additional cars. That triple null is not a failure of transparency. It is transparency's finest product: it is how the world learned that plate rationing does not clean the air — knowledge that cities from Manila to Delhi have had to weigh ever since.

The pattern repeats at every scale. Boston's 25 mph speed-limit cut got a controlled evaluation against Providence, and the honest result was subtle: mean speeds fell 0.3 percent, which was not statistically significant, while the odds of a vehicle exceeding 35 mph fell 29.3 percent — the effect lives in the dangerous tail, not the average. A Spanish field study of smart waste bins found bins never crossed 70% capacity under the city's already-frequent schedule, so sensors weren't economically justified there — the only published null result in an industry that sells fill-level sensors on faith. And Copenhagen's City Data Exchange, a marketplace for urban data, is valuable to this atlas precisely because its partners closed it in 2018, citing weak financial sustainability, and said so on the record instead of letting the website quietly rot.

Our own dataset carries a telling signature of this. Entries documenting failure or retreat are nearly twice as likely to rest on outside evaluation as the atlas average: 30% of our 97 failed or scaled-back entries carry peer-reviewed or independent evidence, against 16% overall. Independent scrutiny is where the failure knowledge concentrates — operators' own reporting rarely files that report.

The silence on the other side

Now the other column of the ledger. Sixty percent of the 1,021 entries in this atlas rest on nothing but the operator's own account, and 15% state no impact evidence at all. The pattern has geographic reach: city camera networks from Casablanca to Tashkent to Yangon publish budgets and hardware counts but only self-reported results, with no independent audit anywhere in the record. And the most famous smart-city project on Earth — Virtual Singapore, the S$73 million national digital twin — has spent a decade in keynotes while the OECD's 2024 case study of the programme describes capabilities only: no adoption figure, no documented decision, no outcome. Fame and checkability are different currencies, and much of the industry still pays only in the first.

For a city shopping for technology, the practical rule that falls out of this is blunt: treat the absence of published data as information. A vendor or a peer city that cannot show you operating numbers after three years is telling you something, and "no independent evaluation exists" is a finding, not a gap in your research. This atlas grades every entry's evidence for exactly that reason — the grade is often more informative than the claim it qualifies.

The catch: publishing doesn't protect you

Honesty about the honest cities requires one hard story. Toronto built the best-evaluated speed-camera programme in North America: a peer-reviewed study across 204 school-zone sites found the proportion of speeding vehicles fell 45% and the 85th-percentile speed dropped by 10.68 km/h. It did not save the programme. Ontario abolished municipal speed cameras by provincial law in November 2025, evaluation notwithstanding. The epilogue is the bleakest data point in this essay: a follow-up city report found speeds back up at 101 of 104 former camera sites.

So the case for transparency has to be made carefully:

• Published data wins arguments, not votes. Toronto's evaluation lost to a provincial election promise. But because the evaluation exists, the abolition itself became measurable — and every city that faces the same fight now argues from Toronto's numbers, in both directions.
• Measurement without publication helps no one else. An internal dashboard is management; a published one is a public good. The difference between the MTA and the four silent operators is not sensors, it is disclosure.
• Self-reporting has a ceiling. An operator's own figures — even honest ones like the MTA's — are still grade-B evidence. The projects that changed global practice all crossed the line into outside evaluation: someone with no stake in the outcome got to run the numbers.

What a city can actually do

None of this requires heroism. It requires four decisions, all cheap at contract time and expensive to retrofit:

• Write the data layer into the contract. Where e-bus performance data is public outside the US, it is often because a concession or a state operator made availability a contractual, reportable quantity. Procurement is the moment of maximum leverage.
• Publish the baseline before the ribbon-cutting. Medellín's evaluation was only possible because pre-intervention data existed. A project that starts measuring at launch can never show what changed.
• Budget the evaluation with the hardware. An independent study costs a rounding error of any infrastructure project and multiplies its value to everyone else.
• Publish the nulls. The Spanish waste study and Mexico City's three strikes did more for global practice than a hundred success stories. If the numbers say it didn't work, that number is the product.

The verdict, then, is the title. In a field where showcases are built to be seen rather than used and failures vanish quietly, the cities that publish their operating data — including, especially, the embarrassing kind — are doing something structurally generous: converting their own budget into everyone else's knowledge. The open-data movement made this argument for datasets. It applies with at least equal force to results.