A junction can appear to be operating acceptably while a queue forms beyond the detector’s field of view, buses lose their scheduled priority, or traffic diverts through residential streets. That is why congestion performance metrics need to do more than report a single average speed. They must show where delay occurs, how reliably the network performs and whether a signal, scheme or intervention is improving conditions for all road users.
For highways authorities and traffic engineers, the objective is not simply to move the greatest possible number of vehicles through a point. It is to manage available road space safely, predictably and efficiently. Meaningful measurement provides the evidence for signal optimisation, corridor management, active travel investment, enforcement and future scheme design.
What congestion measurement should reveal
Congestion is not one condition. A recurring peak-period queue at a signal-controlled junction is different from an unpredictable delay caused by incidents, poor weather, roadworks or uneven detector performance. The first may be addressed through signal timing, staging or capacity changes. The second may require operational resilience, better information or a revised diversion strategy.
A useful performance framework therefore combines flow, speed, delay, queue and reliability data. It also distinguishes between a local issue at a stop line and a corridor-wide problem where one constrained junction causes platoons and spillback elsewhere.
The quality of the underlying detection is central. If a detector misses cyclists, cannot reliably classify vehicles or loses visibility in difficult lighting, the resulting metric can lead to the wrong operational decision. Above-ground radar, wireless sensors and AI-powered video detection can provide the continuous, lane-specific data needed without cutting into the carriageway or creating extended road closures for installation and maintenance.
Core congestion performance metrics for road networks
Journey time and journey time reliability
Journey time remains one of the clearest measures of the road user experience. On a defined route, it shows how long a vehicle takes to travel between two points. Average journey time is useful for strategic reporting, but it can conceal substantial variability.
Journey time reliability addresses that limitation. Measures such as the 95th percentile journey time, planning time index or buffer time indicate how much extra time a person must allow to arrive on time with confidence. A corridor with a reasonable average journey time but frequent severe delays is less dependable for commuters, freight operators and bus services than the average alone suggests.
The right reliability measure depends on the reporting audience. A network manager may need percentile-based reporting to identify operational risk. A scheme appraisal may compare peak and inter-peak changes. Bus operators may need variation by service period and direction. The key is to retain the distribution of journey times rather than relying only on the mean.
Delay per vehicle
Delay compares observed travel time with an agreed reference, usually free-flow or uncongested travel time. Expressed as seconds per vehicle, it helps engineers quantify the operational cost of congestion at a junction, route or network area.
The reference condition requires care. A free-flow speed that is unrealistic for a busy urban street can exaggerate delay, while an overly generous baseline can understate a genuine problem. For urban networks, it is often more practical to establish a representative unconstrained time for the route and apply it consistently across reporting periods.
Delay should also be interpreted alongside safety and place objectives. Reducing vehicle delay by removing pedestrian stages or shortening crossing time may improve one metric while worsening the overall operation of the street. The most useful assessment makes those trade-offs visible.
Queue length and queue duration
Queue length is critical where spillback can block upstream junctions, pedestrian crossings, bus stops or accesses. Maximum queue length shows the scale of the worst condition. Average queue length can reveal persistent pressure. Queue duration shows how long that pressure remains.
For signal control, the most revealing question is often whether the queue clears during each cycle. Repeated failure to clear indicates oversaturation. Once this occurs, delay can rise rapidly and conventional optimisation may have limited effect unless green time, staging, demand or downstream constraints are addressed.
Video analytics and radar detection can support queue monitoring without the disruption associated with carriageway-embedded loops. Their value is particularly clear on approaches where resurfacing, utility activity or changing lane layouts would otherwise create repeated maintenance interventions.
Degree of saturation and demand-to-capacity ratio
Degree of saturation compares traffic demand with the capacity available during the relevant period. It is a familiar junction design and modelling measure, but it is equally valuable in live operations when based on representative detector data.
A value approaching 100 per cent indicates that demand is close to the available capacity. Above that threshold, queues are likely to grow from cycle to cycle. However, the figure should not be treated as a pass-or-fail test. A junction may operate near capacity for a short, predictable peak and still provide acceptable outcomes if queues remain contained and recovery is quick. Conversely, a lower calculated value can hide poor performance if downstream blocking, pedestrian activity or signal coordination interrupts discharge.
Throughput and discharged flow
Throughput is the number of vehicles, cycles or movements served over a period. It is often used to demonstrate capacity, but higher throughput is not automatically better network performance. It may reflect greater demand rather than improved operation, or it may come at the expense of side roads, crossings or cycle movements.
Discharged flow is more useful when paired with queue data. If demand is high but discharge falls below the expected saturation flow, engineers can investigate likely causes: downstream blocking, poor lane discipline, turning conflicts, unsuitable signal timings or inaccurate detection. Vehicle classification adds further value, as a queue of HGVs, buses or mixed traffic will behave differently from one made up largely of private cars.
Speed and speed profile
Speed data identifies where traffic slows and how the pattern changes across a route. A single average speed can be misleading, particularly where vehicles stop at signals by design. Speed profiles over time and distance are more informative. They show whether a problem is confined to an approach, begins after a merge, or moves upstream as demand rises.
For safety-led schemes, speed should be assessed separately from congestion. Lower speeds may be an intended outcome on a high street or near a school, while journey reliability and crossing performance remain the more relevant measures of operational success.
Measure the network, not just the motor vehicle
Road congestion affects different users in different ways. A bus delayed by a poorly coordinated signal may carry dozens of passengers. A cyclist unable to secure detection may face unnecessary delay or take greater risks to obtain a green phase. A pedestrian crossing may experience long waits even while general traffic flows freely.
Where detection technology supports it, performance reporting should separate vehicle classes, turning movements, cycles and pedestrian demand. This is especially relevant when introducing bus priority, cycle stages, pedestrian call facilities or modal filters. The benefit is not merely better reporting. It allows signal strategies to respond to actual demand rather than assumptions about who is using the junction.
Turning raw detection data into decisions
The most effective congestion performance metrics are tied to a clear operational question. Before deploying sensors or changing signal plans, define the decision that the data must support. Is the concern recurring queue spillback? Unreliable bus journey times? A suspected safety conflict? The performance measures, detection positions and reporting intervals should follow from that question.
Establish a baseline before intervention, including comparable days, time periods and weather conditions where possible. Then review both immediate and sustained results. A new timing plan may improve the first week of operation but shift congestion to another approach as drivers adapt. Continuous data makes that effect visible.
Data validation also matters. Compare automated counts with site observations, particularly during commissioning and after layout changes. Check for occlusion, unusual vehicle behaviour, temporary works and detector coverage gaps. An apparently precise dashboard cannot compensate for data that does not reflect conditions on site.
C & T Technology supports this evidence-led approach with non-intrusive detection and traffic analytics solutions designed to provide reliable information while reducing the disruption associated with road-embedded infrastructure.
Use metrics to support practical intervention
No single metric should determine a scheme. A signal timing change that cuts average delay may be worthwhile, but only if it does not create unreliable side-road queues, reduce crossing opportunity or transfer problems beyond the study boundary. Equally, a location with high delay may not need more capacity if better coordination, demand management or detection coverage will resolve the underlying issue.
The practical test is whether the chosen measures explain what road users experience and give engineers a defensible basis for action. When detection is accurate, continuous and easy to maintain, congestion data becomes more than a retrospective report. It becomes a working tool for safer roads, reduced congestion and better use of the network.