Smart parking

Smart parking is the data layer a city wraps around the kerb: knowing which bays are free, pricing them by demand, deciding who may use them, and collecting from drivers who do not pay. It is the one field in this atlas where the expensive instrument — the sensor under the tarmac — mostly did not survive its first decade.

This guide draws on 18 documented projects in 16 cities and 13 countries, 2009 to 2025. Parking is about storing vehicles and who controls the kerb: charging to enter an area belongs in congestion pricing, keeping traffic moving in traffic management, paying to ride in smart ticketing.

What does a smart parking system actually do?

Four jobs hide behind one label, and cities routinely buy the first when they need the fourth:

1. Detecting. Occupancy sensors under each bay — Melbourne, Wellington, Barcelona's Fastprk. For Santander and Zurich, parking is one application on a general-purpose sensor network.
2. Pricing. SFpark moves meter rates with demand. Madrid's SER prices the car instead: the meter reads the plate, queries the national vehicle registry and charges by emissions label — zero-emission cars free.
3. Permitting. Bratislava's PAAS issues resident permits pulled automatically from five state registers; São Paulo swapped forgeable paper cards for a digital permit checked against a plate database.
4. Enforcing. Warsaw's camera cars read every plate twice to cut false positives; Beijing put plate-reading cameras on poles covering 8–12 spaces each.

What happened to the first generation of parking sensors?

Three of the four first-generation municipal sensor networks here are gone or shrunken, and none was killed by a better idea — they were killed by maintenance. The sensor is the expensive part and the first to fail; the payment record is cheaper, already exists, and proved good enough to price with. San Francisco lost its sensors and kept its programme. Barcelona's ApparkB skipped them entirely, forecasting free bays from payments since 2017.

Melbourne is the genuine counter-example: the oldest network here, still running. The reason looks less like better hardware than a better product — the feed is open data, so it has users outside the parking department. An instrument only a procurement officer can see is easier to switch off.

What does the evidence actually support?

Less than the sector claims. No parking entry in this atlas carries a peer-reviewed evaluation of its own outcomes. The field's most-quoted numbers — a 43% fall in time spent searching for parking, a 30% fall in circling emissions — come from SFMTA's own 2014 evaluation and apply to the 2011–2013 pilot areas versus control neighbourhoods, not the citywide programme running today. Quoting them citywide is not what the source says.

The rest is thinner. Melbourne's council reported better compliance and faster turnover after 2011 — a spokesperson's claim, never audited. Cologne's ParkPilot reports search time cut by up to 45% on one street; Bratislava reports 65,000+ permits, over 80% applied for online. All operator figures. Warsaw's count enforcement rather than benefit: 14.7 million plates scanned in 2024, 384,700 fines, over 61 million złoty.

Who is really paying for the kerb?

Follow the money and the technology reads differently. Beijing's 2019 switch was never sold as convenience: cash paid to kerbside attendants was easy to under-declare, and independent reviews by ITDP and Paul Barter both record private operators pocketing the proceeds. The cameras took the concession back for district treasuries. São Paulo did the same sum with permits: revenue rose from R$54.6M in 2016 to R$88.9M in 2017 after the paper cards went.

The kerb is contested space as well as revenue: New York's West Harlem containerisation put in about 1,100 on-street bins at a cost of roughly 500 parking spaces. Montreal marks the limit of publishing kerb status at all: its app warns drivers 60 minutes before a snow-clearing parking ban, but the orange street sign, not the feed, legally authorises a tow. CBC News, analysing city data, counted close to 3,000 vehicles towed outside the hours the app indicated, 2015 to 2018.

Questions to ask before procuring

• Who replaces the sensors, from whose budget? Wellington's lasted eight years — put the refresh cycle in the business case.
• Can payment data answer it instead? Barcelona's ApparkB never needed sensors; San Francisco fell back on payments and kept its programme.
• Pilot number or rollout number? Ask which areas, which years, and who ran the study.
• Is detection or collection the real problem? Beijing and Warsaw fixed collection — a different purchase entirely.
• Does what you publish carry legal force? In Montreal it does not, and a wrong feed costs the resident the tow.
• Who else can use the feed? Melbourne's survives partly because it is open data.

The practical takeaway: buy the cheapest instrument that answers your actual question. In this record that is almost never the sensor under the road — it is the payment record, the plate read or the permit register. None of those corrode.

Frequently asked