A recent Cities Today article on Coral Gables’ use of AI across municipal infrastructure reflects a broader shift in urban operations: cities are moving from periodic observation toward continuous sensing, prediction, and targeted intervention. Traffic management is one of the clearest use cases, because congestion is not evenly distributed across a network. A few intersections, approaches, ramps, or short arterial segments can account for a disproportionate share of delay.
For transportation agencies, the central question is not simply, “Where is traffic heavy?” It is, “Where is the network underperforming, when does it happen, what movements cause it, and which intervention is likely to produce the greatest mobility benefit?” That is where Ticon’s road traffic analytics approach becomes relevant.
Ticon’s methodology starts with broad network visibility. The platform provides near-complete road coverage, including more than 97% of roads classified as FRC 6 and above, and 100% temporal coverage. Instead of relying on a short field count or a limited detector footprint, Ticon consolidates permanent and portable detector data, traffic counters, GPS data, connected vehicle data, GIS information, demographic context, traffic organization records, event information, and other relevant sources. Through cross-verification, filtration, and proprietary processing, these inputs are converted into speed, volume, and derivative traffic performance estimates for about 95% of roadways.
That resolution matters for bottleneck identification. Urban congestion is often localized at a scale smaller than a corridor study can reveal. Ticon’s model can analyze short road segments, down to approximately 35 feet in some cases, with an average segment length of about 225 feet. Time resolution can reach 5-minute intervals, and in many cases 15-second intervals. This makes it possible to distinguish a corridor-wide demand issue from a recurring failure at a specific signalized approach, merge point, turning pocket, or access-controlled segment.
The practical value is that bottlenecks can be ranked by performance metrics, not intuition. Ticon’s virtual transportation model supports numerical ranking of road sections by traffic delay, saturation degree, total driver time loss, speed-volume behavior, and the impact of each section on area mobility. In other words, an agency can move from a map of slow speeds to a prioritized list of locations where interventions may produce the greatest network benefit.
This distinction is important because reducing demand does not automatically reduce delay in proportion. In Ticon’s study, Traffic congestion: what works, what doesn’t, the company examined traffic flows at 126 intersections across nine U.S. states, analyzing approximately 200 million datapoints. The COVID-era reduction in travel demand created a natural experiment: traffic volumes fell by up to 30% or more in many areas, yet delay on signalized roads often decreased by much less. In some cases, delay barely changed even when demand was reduced by almost half. The engineering implication is clear: poor traffic organization can preserve congestion even when fewer vehicles are present.
That finding supports a more disciplined workflow for urban intervention planning. Before adding lanes, restricting access, or deploying expensive ITS equipment everywhere, municipalities need to understand where traffic control, signal timing, saturation patterns, and directional flow imbalances are limiting performance. Based on Ticon’s experience, travel delay reductions of up to 50% can be achieved through signal timing optimization alone in suitable conditions. For cities with constrained budgets, this changes the investment sequence: identify the operational bottleneck first, then select the intervention.
Traffic volume accuracy is central to that process. In AADT Estimation by Various Methods: Accuracy and Reliability, Gregory Brodski and Alex Chaihorsky report that Ticon’s AADT estimation was evaluated across Georgia, Nevada, and California using public DOT sources. After quality screening, 637 counting points remained, many bidirectional, resulting in more than 1,200 estimations. The reported median average percentage error was 4.78%, with relative root mean square error of 11.97%. The study concludes that Ticon can keep expected AADT estimation error within 20% boundaries with 90% confidence.
For bottleneck assessment, AADT alone is not enough. A segment with moderate daily volume may fail during a narrow afternoon peak, while a high-volume arterial may operate acceptably because of balanced flows and adequate control. Ticon’s intraday traffic volume estimation addresses this by estimating day-to-day, hourly, and, for some ITS tasks, 15-minute traffic fluctuations. The methodology incorporates road geometry, intersection geometry, road segment connectivity, vehicle speed distributions, expected vehicle composition, and driver behavior under conditions including weather, time of day, and congestion state.
This allows analysts to identify not just where congestion occurs, but what type of congestion it is. A recurring morning peak bottleneck may point to commuter directional imbalance. A midday bottleneck near a commercial district may point to access management or turning movement friction. A weekend pattern may indicate event traffic, retail activity, or recreational travel. Ticon’s ability to compare volume and speed over the full temporal profile helps separate persistent structural problems from episodic demand spikes.
Ticon’s turning movement estimation uses multivariate analysis of GIS, traffic events and management data, demographics, traffic statistics, connected vehicle information, traffic organization records, location-based services, detector data, and GPS or navigation sources. The platform estimates left, through, and right-turn demand for each 15-minute period over a 24-hour day, with possible aggregation by weekday, weekend, month, season, or year. This is essential for diagnosing urban bottlenecks because many failures are movement-specific. A corridor may appear congested, but the actual constraint may be a saturated left-turn movement, a short receiving lane, or an approach where signal green time does not match observed demand.
Ticon also uses saturation flow analysis and automated impact analysis, which are directly relevant to intervention selection. A speed heatmap can show symptoms, but saturation analysis helps identify whether a facility is operating close to or beyond practical capacity. Automated impact analysis can then support before-and-after studies, comparing travel time delay, saturation degree, and other performance measures after a signal timing change, lane assignment modification, access change, or ITS deployment.
This is also where continuous coverage changes the quality of engineering decisions. Traditional manual counts or short-duration studies often observe only a small fraction of time. A 48-hour count represents about 0.5% of a year, while a one-week count represents about 1.9%. Those methods remain useful, especially for calibration and validation, but they can miss seasonal variation, incident sensitivity, event-driven patterns, and day-of-week differences. Ticon’s continuous temporal coverage provides the basis for identifying whether a bottleneck is chronic, peak-specific, seasonal, or tied to special conditions.
The result is a more practical intervention hierarchy. Some bottlenecks may justify capital projects, but many first require operational correction. Signal timing optimization, progression changes, turn phase adjustments, lane-use changes, incident response improvements, curb management, and targeted ITS deployment can often be evaluated before larger construction programs. Ticon’s TrafficZoom and TrafficScope tools were developed for this purpose, covering speed and volume analysis, saturation analysis, and multiple road types, including highways.
There are also important boundaries to recognize. Ticon’s virtual transportation model can compute many metrics used in practice and recommended by the Highway Capacity Manual, including those needed for congestion analysis and before-and-after evaluation. However, some measures, such as the exact number of stops or queue length, may still require field instrumentation or local detector integration. A credible bottleneck program should combine scalable network analytics with targeted local validation where precision requirements demand it.
The broader lesson from cities adopting AI-enabled infrastructure is that technology is most useful when it improves the sequence of decisions. For urban road networks, that sequence begins with empirical bottleneck identification: define the failing segment or movement, quantify its delay and saturation, understand its temporal pattern, estimate its impact on surrounding mobility, and then choose the least-cost intervention capable of improving performance.
Urban congestion will not be solved by observing traffic more frequently alone. It will be improved by converting continuous observation into engineering diagnosis. Ticon’s contribution is the analytical layer that helps cities see where the network is losing time, why it is happening, and where intervention is most likely to matter.
Join the network to access Ticon case studies and to receive Ticon newsletters
@2023 Ticon
