Case Study: Identifying hidden risks for e-scooter riders in Helsinki

As shared e-scooter use grows rapidly in cities, safety management is still largely based on incomplete data. Most incidents—particularly near-misses and rider-reported risk—are never captured in official statistics.

Building on earlier pilots, See.Sense partnered with TIER-Dott and Vianova in Helsinki as part of the EU-funded ELABORATOR project to demonstrate how rider-level behavioural data from shared e-scooters can uncover hidden safety risks at city scale.

Using sensor-based AI insight data, the project reveals where riders are forced into evasive manoeuvres—providing a leading indicator of risk that traditional data simply cannot capture.

Photo: Igor Väisänen, FVH

The Challenge

Cities face three critical gaps when managing e-scooter safety:

• Underreporting of incidents
  Most safety events (near misses, harsh braking, swerving) are not captured in official statistics
• Limited understanding of rider experience
  Existing datasets do not reflect how infrastructure feels to users in real conditions
• Policy blind spots
  Interventions such as geofencing and speed limits are often deployed without evidence of real-world impact

This creates a reactive system—where decisions are based on lagging indicators rather than leading signals of risk.

The Approach

A six-month pilot was deployed across Helsinki using See.Sense’s SUMMIT-2 trackers integrated onto shared e-scooters.

Data Collection

• 40 connected e-scooters equipped with sensor-enabled trackers
• Continuous monitoring of rider behaviour and road conditions
• 3–5 months of live data collection
  

Sensor-Based AI Insight Data

Each journey generated detailed insights including:

• Swerving
• Braking intensity
• Road surface quality
• Speed and dwell time
• Movement patterns

These signals act as surrogate safety indicators, capturing near-miss events that would otherwise go unrecorded

AI Analysis

See.Sense processed billions of data points to:

• Detect abnormal rider behaviour
• Identify “Rapid Evasive Manoeuvres” (near-miss events)
• Score severity based on rider-specific baselines

Platform Integration

Insights were integrated into the Vianova intelligence platform, enabling:

• Risk vs exposure analysis
• Hotspot prioritisation
• Visualisation of safety patterns across the city
  

Risk vs. Exposure. The yellow dots show all the rapid evasive manoeuvres identified using See.Sense technology and the red hexagon shows where the risk vs exposure are the highest. (Image: Vianova)

Key Results

Scale of Data

• 2,490 trips analysed
• 4,512 km travelled
• 40 connected devices deployed

Identification of High-Risk Locations

Sensor data revealed clear clusters of high-risk behaviour across the city, including:

• Complex intersections
• Areas with mixed traffic flows
• Locations with poor surface conditions

By controlling for exposure, the analysis highlighted true risk hotspots, not just high-traffic areas.

Infrastructure Insights

Comparative analysis of parallel routes showed:

• Safer routes exhibited fewer braking and swerving events
• Higher-risk routes often correlated with:
  • Poor surface quality
  • Conflicting vehicle movements
  • Lack of clear segregation

This provides direct evidence to support infrastructure prioritisation.

Impact of Weather

• Wet conditions increased risk events by up to 2.09× for swerving

Crucially, these events occurred in different locations than typical dry-weather hotspots—highlighting risks invisible in traditional datasets.

Time-of-Day Risk Patterns

Analysis of abnormal events showed:

• Variation in risk across different times of day
• Lower risk in late evening and early morning periods

This supports more targeted operational and policy interventions.

Unintended Consequences of Policy

One of the most important findings:

Geofenced restrictions can introduce new risks.

Examples identified:

• Slow speed zones
  • Reduced vehicle stability
  • Increased interaction risk with other road users
• Mandatory parking zones
  • Caused sudden swerving and direction changes
• Categorical zoning (e.g. parks, schools)
  • Impacted adjacent bike lanes and through-routes

These findings demonstrate that well-intentioned policies can create unintended safety issues if not informed by real-world data. 

Approximately 10% of all recorded unusual events occurred inside low-speed zones. This could be to do with pedestrian interactions. (Picture: Vianova)

Outcomes for Helsinki

The project delivered:

• A new safety evidence base incorporating near-miss data
• Identification of previously unknown risk locations
• Insights into the real-world impact of infrastructure and policy
• A scalable framework for ongoing micromobility safety monitoring

Why This Matters

This case study demonstrates that:

• Sensor-based AI insight data works beyond cycling — it applies directly to e-scooters and wider micromobility
• Cities can move from reactive to proactive safety management
• Operators and authorities can co-design better interventions using shared data

Most importantly, it shows that the majority of safety risk exists in data that cities do not currently capture—and that See.Sense provides a practical way to access it.

Key Takeaways

• Near-miss data is critical to understanding true safety risk
• Infrastructure quality and design directly influence rider behaviour
• Weather and time-of-day create dynamic risk patterns
• Policy interventions must be validated against real-world behaviour
• A standardised dataset enables comparison across transport modes