Supporting Vision Zero: Using Road Traffic Data and Analytics

This webinar started from an uncomfortable fact about road safety: crash reports are lagging indicators. Most agencies identify dangerous locations using several years of collision history, so they find risk only after someone has been killed or seriously injured.

Vision Zero asks agencies to prevent those outcomes rather than count them. That requires risk signals that are fresher, broader and closer to the behavior that causes crashes: operating speeds, congestion, incidents, hard braking, rapid acceleration and near misses.

US road fatalities peaked at roughly 43,000 in 2021 and have fallen for fourteen consecutive quarters, but the gap to Vision Zero remains wide. Normalized for exposure, the US still records two to three times more fatalities per 100 million vehicle miles traveled than Western and Central Europe (about 0.5) and far more than the Nordics (about 0.25). Speeding accounts for almost 30% of fatalities; alcohol impairment, distraction and low seat-belt use follow; and deaths among pedestrians and cyclists have grown even as vehicle-occupant deaths have fallen.

Five perspectives shaped the discussion:

• the state of road safety and the strategies with measurable impact (TomTom)
• telematics-based risk detection and enforcement evaluation (Michelin Mobility Intelligence)
• international speed, policy and enforcement evaluation (Agilysis)
• network-wide Safe System risk modeling for curves and intersections (Abley)
• behavior-driven countermeasure prioritization and before-and-after measurement (Replica)

They pointed to the same conclusion:

“Proactive safety does not replace crash data. It adds leading indicators so agencies can act before harm occurs, and can show why they acted.”

The second lesson concerned accountability. Data-driven prioritization is safer, and it also makes decisions easier to defend. When California found that 83% of injury and fatal crashes happened on 8% of streets, and Atlanta found 2,000-plus crashes on two streets, that evidence both directed the work and unlocked the funding to do it.

“Using crashes and lower-frequency events to try to predict the future is very tough to do. Thinking through that forward-looking framework, using predictive modeling to quantify and measure and identify risk before it even happens — just using the data that's out there is a huge tool.”

• ✓Crash data is a lagging indicator. Treat it as one input rather than the only one.
• ✓Behavior data reveals emerging risk. Speeding, hard braking, rapid acceleration and near misses show where crashes are likely before they happen. One intersection may see a single crash and a thousand harsh-braking events.
• ✓Combine sources. Multi-source data (probe, telematics, incident, volume and crash) produces a more reliable picture of high-injury corridors than any single dataset — and volume data is what lets you compare one corridor fairly against another.
• ✓Speed is the central risk factor. Survival rates fall sharply as impact speed rises. Portland cut deaths 43% in a year after lowering limits, and Amsterdam's citywide 30 km/h limit cut average speeds 5% and crashes 11%.
• ✓Enforcement works when it is targeted with data. San Francisco recorded a 72% speed reduction at 2025 camera sites; Chicago saw pedestrian and cyclist crashes fall 46% near cameras; a six-week data-targeted patrol program in Washington State preceded a 58% drop in severe collisions year over year.
• ✓Evaluate what you build. Probe data measured the Wales 20 mph change five days after it took effect. You no longer need to wait years for crash counts to accumulate.
• ✓Network-wide models stretch tight budgets. Abley's corridor analysis showed a quarter of the network length holding roughly three quarters of the network risk.
• ✓The binding constraint is people and inertia, not funding. Presenters named a retiring workforce, thin capacity and 'we've always done it this way' as the real obstacles in North America.
• ✓Go on the offensive. Move from reactive fixes to proactive intervention. The data is already there.

Setting the Stage: Where Road Safety Stands and What Is Working

Saurav Miglani — Segment Marketing Lead, Public Sector, TomTom

The scale of the problem

US road fatalities peaked at about 43,000 in 2021. Fourteen consecutive quarters of decline since then suggest that safety funding and decisions are paying off, but the gap to where agencies want to be is still significant.

Saurav normalized the comparison using fatalities per 100 million vehicle miles traveled, to answer the common objection that Americans simply drive more. Even on that basis the US rate runs two to three times higher than European comparators.

The primary causes he cited:

• speeding — almost 30% of all fatalities
• alcohol impairment — the second largest factor
• distracted driving
• seat belts — about 50% of passenger-vehicle occupants killed were unbelted
• vulnerable road users — occupant deaths have fallen while pedestrian and cyclist fatalities have grown significantly

Four strategies delivering measurable impact

1. Speed management. At 40 mph the survival rate is about 2 in 10; at 30 mph it improves to about 6 in 10. Speed is the most divisive lever politically and the fastest-acting one in practice.

2. Automated enforcement. Use data to find where drivers actually exceed the limit or the 85th percentile, then place cameras and AI systems there.

3. Roadway redesign. The most expensive lever, and the reason European fatality rates sit lower — priority given to vulnerable road users in street design.

4. Data-driven targeting. Combining crash history with connected-vehicle data — speeds, harsh braking, electronic stability control activations, weather and hazard warnings — to find risk hotspots proactively.

What TomTom is building

TomTom supplies global maps, real-time traffic, hazard warnings and a deep archive of historical traffic data to automotive, enterprise and public-sector customers. Saurav previewed a new product dedicated to road safety that brings that archive together in one place: historical speeds, historical incidents and accidents, acceleration and deceleration rates, lane-level behavioral change, slippery-road and weather archives, and mobile-mapping road-sign data that can reveal missing or non-compliant signage.

Real-World Risk Detection to Drive Vision Zero Implementation

Erik Fretland — Senior Data Scientist, Michelin Mobility Intelligence

Telematics as a leading indicator

Michelin Mobility Intelligence's Safer Roads team uses anonymized, opt-in telematics data — largely from drivers' mobile phones — to locate, quantify and measure risk in the real world, then partners with safety organizations to act on it.

The team's focus is not the general pulse of speeds across a network but the events at the edge: harsh braking, harsh acceleration and harsh swerving. These are the best available proxies for near misses.

“For an intersection you may have one crash. Maybe you have a hundred or a thousand harsh-braking events, which may stand out as abnormal and thus is likely to be an indicator of future crashes happening there.”

Every event is geolocated and timestamped, so speed, weather, lighting conditions and direction of travel can all be attached to it — local context for a single event, and broader patterns at network scale.

Three categories of analysis

Case one — Napa Valley Transportation Authority

MMI built a county-wide risk and behavior map for Napa County, California, covering everything from highways down to county roads. Because NVTA wanted to see how risk evolved, the analysis pulled matched time periods from 2023 and 2024 to test whether the programs they had enacted were working.

Two deep dives followed: how tourism in a well-known winery destination contributes to risk in specific areas, and a holistic evaluation of every school zone in the county, public and private, elementary through high school.

MMI's graph risk prediction model scores each road segment from zero to one for the likelihood of a VRU-related severe crash, using behavior and built-infrastructure characteristics. Combined with the raw behavioral events, it localized which streets around each school were high-risk, what was driving that risk, and produced a single risk score per school so NVTA could rank where to focus.

Case two — Washington Traffic Safety Commission and Washington State Patrol

This project was about resource optimization for a police department operating under real constraints. MMI scanned the interstate and state highway system across the whole state to find where speeding and risky behavior were most frequent and most severe, then worked with both agencies to pick the corridors where added enforcement would do the most good.

Four corridors were selected — three in the west of the state and one east through Spokane. After six weeks of ramped-up enforcement, MMI quantified the relationship between the number of officers deployed and the resulting change in speeds and dangerous-driving frequency, and identified where the intervention worked best and where it worked least.

Three Use Cases for Traffic Data in Road Safety

Richard Owen — CEO, Agilysis Limited

Agilysis has delivered road safety projects for UK road authorities, the World Bank, the World Health Organization, the Asian Development Bank and large insurers. Richard presented three applications of TomTom connected-vehicle data, each answering a different question.

Use case one — National safety performance in Thailand

The Asian Development Bank asked Agilysis to assess safety performance on roads across the whole of Thailand. Doing that with automated traffic counters would have cost millions of dollars, so the team used TomTom probe data over the Overture base map instead, for a fraction of the cost and with far greater coverage and timeliness.

The result is an online dashboard, now published by the ADB and made available to the Thai government. It reports road length, vehicle movements in the sample, average speed and 85th percentile speed, and filters by land use (urban or rural), speed limit and road class. Users can drill from the national picture down to a province such as Bangkok, and then to an individual road section.

Use case two — Evaluating legislation in near-real time

Wales changed its default urban speed limit from 30 mph to 20 mph in September 2023. Agilysis ordered TomTom data for the weekend before the change and the weekend after, cloned the same queries, and had results within about an hour of reordering — across roughly 500 miles of Welsh roads.

Just as important as the drop in the mean was the narrowing of the speed distribution. The tail of very high-speed drivers shrank, and drivers travelled at more similar speeds — which is what road safety practitioners want to see.

Use case three — Enforcement that earns compliance

The UK has been a pioneer in speed enforcement, and average speed cameras — point-to-point, time over distance — are the dominant technology. Evaluations of these schemes typically wait years for collision data. Agilysis evaluated one immediately, on the A5 near Silverstone: a long, straight Roman road that encourages high speeds.

Speeds at either end of the enforced section changed little; the reductions concentrated in the middle, exactly where the cameras are intended to work. Funders could see the effect of their investment immediately.

“You can rapidly assess what's happening on your road networks almost the day after you've made those changes.”

The Abley SafeSystem: Taking the Guesswork Out of Road Safety Investment

Steve Abley — Chief Executive, Abley

The problem it solves

Abley built SafeSystem because international clients wanted better road safety insight in rural areas — areas known to perform poorly, yet with a distinct lack of data pinpointing risky locations beyond traffic flow and exposure metrics.

“Unlike crash-led tools, the Safe System identifies risk everywhere, not just where crashes have already happened.”

The suite spans strategic star ratings through Safe Roads and tactical interventions for rural curves and intersections through Safe Curves and Safe Intersections. Each product aligns with the Highway Safety Manual, flags every road, curve or intersection that deserves attention even where no crash has been recorded, and pairs the risk with real-world countermeasures crews can act on immediately, consistently and defensibly.

Safe Curves across every paved road in the US

Steve demonstrated the national analysis — Alaska to Hawaii to Puerto Rico — structured in three parts: identify, prioritize, intervene. For every horizontal curve on every paved road the model derives:

• the specific risk of a rural roadway departure crash
• the 100% MUTCD-compliant signage, delineation and striping for that curve
• pavement friction areas, with aggressive and comfortable deceleration lengths
• stopping sight distance

Use case one — Finding risk where crashes haven't happened yet

Zooming to the tri-state area of Tennessee, Alabama and Georgia — Hamilton, Marion, Jackson and Dade counties — the identify layer classifies every curve as red, amber, green or gray. The prioritize layer aggregates those individual curves into homogeneous driving corridors, which is what turns a smorgasbord of point risk into an investment programme.

“Priority 1 and priority 2 corridors are only 25% of the network length, but hold about 70% of the network risk. A quarter of any network can be shown to hold roughly three quarters of the network risk.”

Use case two — Turning priorities into defensible action

Selecting one of the worst corridors opens an attributes panel — in the demo, the 20th worst corridor in the study area, with its speed limit and functional classification. The countermeasures panel then sets out three tiers of action:

Every action hyperlinks to the FHWA Proven Safety Countermeasures pages for practitioner guidance. The signs-and-markings layer converts the standard into exact locations, exact treatments and exact justifications: clicking a horizontal warning sign explains why it is justified — in the demo, that the sign is optional because the speed differential to the limit is 5 mph. Ball banking is handled by the model.

“Safe System helps agencies act earlier, act consistently and act with confidence before people are seriously hurt. That's how Abley SafeSystem supports Vision Zero in practice, not just in principle.”

Going on the Offensive: Using Driver Behavior Data to Proactively Prioritize Countermeasures

Sean Johnson — Vice President, Sales, Replica

From reactive to proactive

Everyone on the call has spent time with crash records, and the problems are familiar: the lag from event to record, incompleteness, and the work needed to normalize and cleanse the data. That is not the fault of the reporting department; it is the reality of the source. Driver behavior data gives agencies a much closer-to-real-time view of the risk on the network.

The signals

Case one — Corridors of regional significance

Working with a metropolitan planning organization that had identified corridors of regional significance in its LRTP, Replica's Safety Insights application visualized every crash on those corridors, filtered to unsafe speed as the primary contributing factor, and overlaid excessive speeding of 10–20 mph over the limit. Crashes appear as yellow points; the dark purple shows where excessive speeding concentrates, answering directly whether crashes and risky driving correlate.

Sean noted a recurring pattern in how customers use it: municipal and MPO teams focusing on excessive speeding on suburban arterials with a high presence of vulnerable road users. Volume data settles the prioritization question when two corridors both show, say, 500 excessive-speeding instances.

Case two — Before-and-after measurement

A Washington DC bus priority corridor — a dedicated bus lane, added protected bicycle lanes and reconfigured stops — was evaluated in Replica's before-and-after application shortly after the build. It answers whether dangerous driving decreased, whether speeds changed, and whether travel times rose as opponents of bike lanes commonly predict.

Case three — Near-real-time detection

Replica's newer work surfaces risky-driver-behavior anomalies within hours rather than days. In a Sacramento example, a sharp drop in a device's speed flagged a suspected collision; the detail panel adds context to explain the why — road feature characteristics and volumes, land use and population density, weather conditions, and Google Street View of the built environment. In development is a threshold-and-notification layer: if a corridor's typical ten excessive-speeding events per day jumps to twenty, stakeholders get an automatic alert and can deploy enforcement partners to that location.

“As an industry we're basically at the tip of the iceberg.”

The practical guidance from the presentations, consolidated. Print this section.

Before you build — proactive safety data checklist

• ☐Define your outcome metric. Name it: killed and seriously injured (KSI), fatalities, or injury crashes. Set the target and the timeline.
• ☐Inventory your crash data. Record years available, geocoding quality, severity coding and reporting lag.
• ☐Add at least one leading indicator. Options include probe speeds, telematics harsh events, incident feeds and congestion.
• ☐Confirm coverage and sample size. Check penetration rates by road class, especially on rural and low-volume roads. Two rural Minnesota counties with 70,000 residents still produced 165,000 near-miss events in three months.
• ☐Add exposure or volume data. Without it you cannot fairly rank two corridors that show the same raw event count.
• ☐Align to a common network. Conflate every dataset to the same road segments, or they will not join.
• ☐Set the segment and time resolution. A signalized intersection and a rural curve need different analysis units.
• ☐Establish a baseline before any intervention. Order the data the week before the change, not the year after.
• ☐Plan for transparency. Decide what gets published, and in what form, for residents and grant reviewers.
• ☐Check privacy and licensing terms. Confirm aggregation levels and usage rights for third-party data.

Evaluating countermeasures — a working guide

Speed management — the four levers

1. Set — Set limits that fit the context and the road users present, following Safe System principles.
2. Design — Engineer the road so the intended speed is the natural speed.
3. Monitor — Track operating speeds and compliance continuously with probe data.
4. Enforce — Target enforcement with data and measure the compliance that results.

Programs and guidance referenced

• Vision Zero — The goal of eliminating traffic deaths and serious injuries.
• Safe System Approach — The framework adopted in the USDOT National Roadway Safety Strategy.
• Highway Safety Manual — The best-practice reference Abley SafeSystem aligns to.
• MUTCD — The signage, marking and signal standard that Safe Curves converts into exact locations and treatments.
• SS4A — Safe Streets and Roads for All, the federal grant program that funds safety action plans and implementation.
• FHWA Proven Safety Countermeasures — Federally recognized, effective safety treatments.

A crawl–walk–run roadmap

• Crawl — Map your high-injury network from crash history. Add one leading indicator, such as probe speeds, for the same corridors, and compare the two. Most vendors on this panel offer trials or evaluation data to make this step cheap.
• Walk — Rank corridors by combining crash, behavior and volume data. Pilot countermeasures on the top corridors, with before-and-after measurement.
• Run — Screen systemically across the whole network, including rural curves. Monitor continuously, set alert thresholds on behavior data, report publicly and reallocate budget based on measured results.

The opening demographic poll received 44 responses. Percentages below are reproduced from the official poll report; blank responses account for totals below 100% in some questions.

Questions came from the live audience, hosted by Stuti Srivastava.

“We have the tools in the toolbox. We just need to reach for them.”

Closing question — one message for the audience

“Silos are not going to build and achieve what we're going for. It has to be a joint effort where each of us takes that first step.”

Vision Zero

A strategy to eliminate all traffic fatalities and severe injuries, based on the principle that no loss of life on the roads is acceptable.

Safe System Approach

A framework built on the premise that humans make mistakes and are vulnerable. The system is designed so that errors do not result in death or serious injury.

High-Injury Network (HIN)

The small share of streets where a disproportionate share of fatal and serious-injury crashes occur.

KSI

Killed or Seriously Injured. A standard severity-based safety metric.

Lagging indicator

A measure of harm that has already happened, such as crash reports.

Leading indicator

A measure that signals risk before harm occurs, such as speeding or harsh braking.

Telematics

Data transmitted from vehicles or devices, describing location, speed and driving events.

Probe data

Anonymized speed and travel data collected from vehicles and devices moving through the network.

Harsh braking / rapid acceleration

Sudden deceleration or acceleration events. They are used as surrogate indicators of conflict and risky driving.

Near miss

A conflict that nearly resulted in a crash. It is a surrogate safety measure.

Surrogate safety measure

A non-crash indicator that correlates with crash risk.

Systemic safety analysis

Screening the whole network for risk factors, rather than reacting only to crash clusters.

Countermeasure

An engineering, enforcement, education or policy action intended to reduce crash risk.

CMF

Crash Modification Factor. The expected change in crashes after a countermeasure is applied.

85th percentile speed

The speed at or below which 85% of vehicles travel. It is commonly used in speed studies.

Operating speed

The speed at which drivers actually travel, as distinct from the posted limit.

Speed compliance

The share of vehicles traveling at or below the posted limit.

Average speed camera

Point-to-point enforcement that measures the time taken between two camera sites to calculate average speed over a distance.

VRU

Vulnerable road user. Pedestrians, cyclists and other users unprotected by a vehicle body.

VMT

Vehicle miles traveled. Used to normalize fatality and crash rates for exposure.

MUTCD

Manual on Uniform Traffic Control Devices. The US standard for signs, markings and signals.

Roadway departure crash

A crash in which a vehicle leaves the travel lane or the roadway, common on rural curves.

SPF

Safety Performance Function. A statistical model relating crash frequency to road and traffic characteristics.

Before-and-after evaluation

Comparing conditions before and after an intervention to measure its effect.

Regression to the mean

The tendency of unusually high crash counts to fall naturally, which can make a treatment look more effective than it is.

SS4A

Safe Streets and Roads for All, a US federal grant program for safety planning and implementation.

Action Plan

A comprehensive safety plan. It is a prerequisite for SS4A implementation funding.

LRTP

Long Range Transportation Plan. The long-term plan an MPO maintains for its region.

• TomTom — Webinar series sponsor. A location and navigation company providing global maps, real-time traffic, hazard warnings, historical traffic archives, EV services and a navigation SDK to automotive, enterprise and public-sector customers, with a road safety data product in development.
• Michelin Mobility Intelligence — Its Safer Roads team uses anonymized, opt-in telematics data on real-world driving behavior to locate and quantify risk, build predictive risk models, and evaluate the impact of safety decisions.
• Agilysis — An award-winning UK road safety consultancy working with road authorities, the World Bank, the World Health Organization, the Asian Development Bank and insurers on performance measurement, policy evaluation and enforcement strategy.
• Abley — Developer of Abley SafeSystem — Safe Roads, Safe Curves and Safe Intersections — which identifies network risk everywhere and specifies MUTCD-compliant countermeasures for each location.
• Replica — Provider of Safety Insights and before-and-after analysis applications, combining crash records with driver behavior, travel pattern and volume data so agencies can prioritize corridors and measure the effect of interventions.

From this presentation

Organizations & tools referenced

Standards & guidance referenced

Asset Mapping Intelligence collects practical guidance from public-sector practitioners and data specialists working across transportation, road safety and infrastructure delivery.

Explore AssetMapping.Events →

This report was compiled from the presentation recording and transcript. Statements are attributed to the speaker who made them and reflect their own views rather than those of their employers. Where a speaker referenced a specific program, standard or regulation, we have used the current official designation. Product capabilities described reflect what was stated during the presentation and may change; confirm current specifications with the vendor.