Traffic management

Traffic management covers everything a city does to control how vehicles move through a fixed amount of road: signal timing, access rules, pricing, enforcement and the control rooms that watch it all. The technology has changed enormously since the 1970s. What actually reduces congestion has changed much less.

This guide draws on more than 50 documented entries across nearly 40 cities in 25 countries — the atlas's deepest field, and the only one whose oldest system is still running fifty years on. It connects to the adaptive traffic signals and congestion pricing definitions, and to the long view in fifty years of congestion pricing.

Two architectures, fifty years apart

SCATS went live in Sydney in 1975 and is the oldest live entry in the entire atlas; Los Angeles built ATSAC nine years later. Both are central: sensors report to a control centre, which retimes signals across the network. That model still dominates — Istanbul, Bucharest, Belgrade, London and Bangkok all run versions of it.

The alternative is to push the decision down to the intersection. Pittsburgh's Surtrac lets each junction optimise locally and negotiate with its neighbours, and its own documented lesson is the interesting one: decentralised control scales more easily than a central brain, because adding an intersection does not mean re-tuning the whole network.

Where the evidence is strong — and where it is not

Signal optimisation produces impressive headline numbers, and almost all of them are self-reported. Transport for NSW credits SCATS with 28% shorter travel times across its worldwide deployments; the Surtrac pilot figure of ~25% comes from the research team that built it; Hangzhou's ~15% speed gain is Alibaba's own; Bangkok's 10–41% delay reduction is the city's own assessment. None of these is dishonest. None has been independently audited either, and every atlas entry says so.

The peer-reviewed evidence in this field sits somewhere else entirely: on the demand side. A randomisation-exploiting study found Beijing's plate lottery cut the city's car stock by 14%. An independent study of Barcelona's low-emission zone measured NO₂ down 15.8% inside the zone — while finding particulate effects small and not robust. Singapore's ERP raised zone speeds by nearly 30% in the published record. London's low-traffic neighbourhoods carry peer-reviewed evaluation too.

The pattern is uncomfortable for procurement: the interventions with the weakest independent evidence are the ones a city can buy, and the ones with the strongest are the ones it has to win an argument about.

What decay looks like

London's congestion charge is the field's most important cautionary record. Year one delivered roughly 30% less congestion — and TfL's own later monitoring found those gains had largely eroded by the late 2000s, even though traffic volumes stayed lower. Pricing is not a one-off fix; it needs recalibration. Rome's ZTL makes the same point from the access-control side: two decades of automated gates cap and modestly trim entries into the historic centre, but daily volumes remain high.

Two ways these projects die

Dhaka's signal automation was cancelled after the hardware went in without operator training, maintenance budget or integration with existing police practice — the signals were switched off and traffic reverted to manual control. Algiers failed earlier in the chain: of 500 contracted intersections, 22 of the first 200 were wired and none fully switched on, blocked by unfinished fibre links and a national-security refusal to authorise the cameras.

Neither failed because the algorithms were wrong. Both failed on operations and institutions — which is where Wiesbaden's record also points: a ~€30M citywide digitalisation of 227 signal systems delivered data first and visible congestion relief much later, with local coverage staying sceptical in the meantime.

The trade-off that rarely gets priced

Copenhagen's green wave is unusually honest about this. Synchronising signals to a ~20 km/h cycling speed raised average cyclist speed from about 15 to 20 km/h and cut red-light stops from as many as six to zero or one — at a cost of 27 to 50 seconds of added delay for buses on the same corridors. Every signal plan allocates delay to somebody. Most published evaluations only report the group that gained.

Questions to ask before procurement

• Who is the plan optimising for? Cars, buses, cyclists and pedestrians cannot all be prioritised on one corridor.
• Who retimes it in year three? Adaptive systems drift; name the team and the budget, not just the installer.
• Is the baseline measured? Without before-data, any later number is a claim rather than a result.
• Self-reported or independent? Ask the vendor which of their published figures a third party has verified.
• Does the political option exist? If pricing or access restriction is off the table, be honest that the technology is being asked to substitute for it.
• What happens when it breaks? Dhaka's fallback was manual control; that should be a designed state, not a surprise.

The practical takeaway: signal technology mainly buys efficiency within the traffic a city already has. The measured, independently evidenced reductions in the atlas come from changing how much traffic there is. Cities that expect the first to deliver the second are usually disappointed — slowly, and after the ribbon-cutting.