Product Documentation · Road Safety Analytics Platform
User Groups
Question Framework
This document defines the structured question framework used across the road safety analytics platform. Each section includes a platform mind map showing how spatial levels connect to their tabs, and supporting dashboard cards showing exactly how each question is answered on screen.
01 · Introduction
Objective
The question framework structures how different user groups interrogate road safety data. Each question is mapped to a user group, a spatial level, and a stage — ensuring that insights flow progressively from macro city-wide awareness down to granular intersection-level root causes.
⚡
Core Flow: Every question set follows the same analytical progression — Overview → Type → Factors → Anomalies — applied at each spatial level and tailored to the needs of each user group.
02 · System Architecture
Platform Mind Map — How Everything Connects
Each spatial level hosts four analytical tabs. Each tab answers specific questions through purpose-built dashboard cards. The map below shows the full structure from level → tab → card type.
City
City Level
📊 OverviewHotspot map · KPI tiles · Trend chart · Zone comparison
🏷 TypeDistribution donut · Severity bar · Type trend · City avg
⚙️ FactorsPCF donut · Weather share · VRU risk · Time heatmap
🔺 AnomaliesSpike timeline · Peer ranking · AI summary · Outlier map
drills into corridor
Corridor
Corridor Level
📊 OverviewSegment ranking · Risk dist · Corridor vs city · Trend
🏷 TypeDominant type card · Type breakdown · Context combos
⚙️ Factors24h time chart · Night split · Dark table · PCF donut · VRU
🔺 AnomaliesSegment deviations · Localised spikes · Behaviour shifts
drills into intersection
Intersection
Intersection
📊 OverviewCrash freq · Improvement trend · Peer table
🏷 TypeCollision pattern · Movement matrix · Rare + lethal bubble
⚙️ FactorsSignal phases · Lighting table · Behaviour donut · AI synthesis
🔺 AnomaliesRolling avg spike · Peer KSI · AI summary · Deviation ranked
Overview Tab
KPI cards · Hotspot map · Trend lines · Zone/segment risk distribution
Type Tab
Dominant type · Distribution donut · Severity per type · Trend over time
Factors Tab
Timing heatmap · Lighting table · PCF donut · VRU risk index · Weather share
Anomalies Tab
Rolling avg spike · Peer ranking · AI anomaly summary · Deviation scored table
03 · Architecture
Spatial Drill-Down Model
The system connects three levels of analysis, each building on the previous.
City
Identifies where problems exist
Macro-level awareness of crash density, hotspots, and systemic trends across the city.
Corridor
Explains how problems behave
Segment-level patterns, road-specific behaviours, and corridor ranking by risk contribution.
Intersection
Reveals why they occur
Granular diagnosis — collision patterns, causal factors, and outlier behaviour.
04 · User Groups
Users
Three distinct user groups interact with the platform, each with a different analytical purpose.
Operations
Police · Corridor Director
Monitoring and immediate action. Fast identification of problem areas and prioritisation of response.
Analysis
Safety Analyst · Engineer
Diagnosis and solution design. Deep root-cause identification and reliable correlation across variables.
Strategy
City Director · Elected Official
Prioritisation and investment decisions. Clear justification for funding aligned with planning.
05 · Framework
Question Stage Legend
Four progressive stages apply consistently across every user group and every spatial level.
Stage 01
Overview
What is happening?
Stage 02
Type
What kind of problem is this?
Stage 03
Factors
Why is this happening?
Stage 04
Anomalies
Where should attention go?
06 · User Group
Operations — Monitoring → Immediate Action
Value Enabled
Fast identification of problem areas · Behaviour-driven risk awareness · Immediate prioritisation of action
C
City Level
Macro awareness · Hotspot identification
Overview
- Which areas have the highest crash density?
- Which zones are emerging vs. persistent hotspots?
- How has risk shifted over time?
- Which areas require continuous monitoring?
Type
- What collision types dominate these hotspots?
- Which types are increasing fastest?
- Are pedestrian crashes concentrated in specific areas?
- Do specific types repeat across locations?
Factors
- What behaviours are driving crashes (speed, alcohol, violations)?
- Are these behaviours time-dependent?
- Do the same factors repeat across hotspots?
- Are factors geographically clustered?
Anomalies
- Which areas show sudden spikes?
- Which locations transitioned from low to high risk?
- Where do patterns behave abnormally?
- Which hotspots persist despite attention?
R
Corridor Level
Segment risk · Behavioural patterns
Overview
- Which corridors contribute most to crashes?
- Which corridors are worsening over time?
- Is risk concentrated or evenly distributed?
Type
- Which crash types dominate each segment?
- Do patterns shift along the corridor?
Factors
- Do behaviours vary across segments?
- Are crashes linked to infrastructure gaps?
Anomalies
- Which segments deviate from corridor behaviour?
- Where are localised spikes occurring?
I
Intersection Level
Point-level monitoring · Immediate response
Overview
- How frequently do crashes occur here?
- Is this location improving or worsening?
Type
- What exact collision patterns exist at this point?
Factors
- What behaviours consistently cause crashes here?
Anomalies
- Is this location an outlier vs. nearby intersections?
- Is this part of a larger cluster?
07 · User Group
Analysis — Diagnosis → Root-Cause Explanation
Value Enabled
Deep root-cause identification · Reliable correlation across variables · Confidence in analytical conclusions
C
City Level
Long-term trends · Statistical distribution
Overview
- What long-term trends exist across the city?
- Which areas consistently show high risk?
- How does the overall safety picture compare year-on-year?
Type
- What collision types dominate spatially?
- How are crash patterns distributed across districts?
- Which types are statistically over-represented?
Factors
- What factors correlate strongly with KSI outcomes?
- How do multiple variables interact to elevate risk?
- Which environmental or temporal factors amplify severity?
Anomalies
- Which areas deviate statistically from expected patterns?
- Are anomalies consistent over time or episodic?
- Do outliers share common underlying causes?
Supporting Cards — Type Tab · City Level
Distribution · Severity · Ranking · Pattern DNA — from your Type Tab screen
Type · Distribution
Which collision types are most common?
1,499
total crashes
Type · Severity · KSI Focus
Which types have the highest KSI rate?
#1 Veh/Ped
62% KSI
#2 Head-On
58% KSI
#3 Sideswipe
28% KSI
#4 Hit Object
24% KSI
#5 Broadside
19% KSI
Severe minority: 58% minor crashes vs 4% fatal
Type · Ranking · Priority Score
Which types should be fixed first?
Veh/Ped CRITICAL
High volume + extreme KSI + worsening trend
Rear-end HIGH
Highest volume, moderate KSI, rising fast
Broadside HIGH
High volume + signal violations increasing
Sideswipe LOW
Low volume, declining trend
Type · Pattern DNA · Archetypes
Common scenario archetypes
Evening Pedestrian Trap
LT + Ped violation · 6–9 PM · Dark — 78% KSI · 58 crashes
Signal Rear-end Cascade
Rear-end + Signal stop · PM peak · High vol — 11% KSI
Night Speed Fatality
Speed + Dark + Fixed object · After midnight — 65% KSI
Broadside Conflict Zone
HGV + Broadside · High-speed road — 22% KSI · 180 crashes
Type · Comparison · Over-rep
Which types are over-represented in fatalities?
R
Corridor Level
Comparative analysis · Infrastructure correlation
Overview
- How does this corridor compare to similar corridors?
- Which segments contribute most to overall risk?
Type
- Are there hidden temporal patterns in crash types?
- Which segment-type combinations are most diagnostic?
Factors
- What infrastructure or operational factors drive risk?
- How do behaviours interact with physical road design?
Anomalies
- Which segments behave unexpectedly relative to corridor norms?
- Are anomalies explainable by a single causal factor?
Supporting Cards — Corridor · Factors Tab
Timing · Lighting · PCF · Condition combos · VRU risk — from your Corridor Factors screen
Corridor · Factors · Timing
When do crashes happen most on this corridor?
Peak Hour
8–9 AM
Highest single-hour count
Night Share
34%
After dark crashes
Corridor · Factors · Night Split
Are crashes higher at night on this corridor?
Day
66%
Night
34%
Weekday
61%
Weekend
39%
⚠ Night crashes are 1.4× more likely to result in KSI than daytime crashes
Corridor · Factors · PCF
What are the primary contributing factors?
28%
Speeding 28%
Inattention 22%
Fail to Yield 16%
Alcohol 12%
Other 22%
Corridor · Factors · Combos
Most common conditions in crashes
1st — Speeding + Dry Road HIGH
n=342
2nd — Inattention + Daylight MED
n=280
3rd — Alcohol + Night V.HIGH
n=154
4th — Wet Road + Speed HIGH
n=142
5th — Dark Unlit + No markings CRIT
n=98
Corridor · Factors · VRU Risk
Are pedestrians & cyclists at higher risk?
🚶 Pedestrian — 89 crashes
Share
6.9%
KSI Rate
58%
Risk Index
1.8×
🚲 Cyclist — 64 crashes
KSI Rate
38%
Risk Index
2.2×
I
Intersection Level
Statistical significance · Causal stability
Overview
- What is the statistical significance of this location's crash rate?
- How does it compare to similar intersections citywide?
Type
- What combination of collision types defines this intersection?
- Does this match known high-risk patterns?
Factors
- What combination of factors causes crashes at this point?
- Are causal factors stable or shifting over time?
Anomalies
- Is this intersection's behaviour an artefact or a genuine signal?
- Does it cluster with nearby high-risk points?
Supporting Cards — Intersection · Factors Tab
INT-041 · Main St & Harbor Blvd — signal phases · lighting · behaviour · AI synthesis
Intersection · Factors · KPIs
INT-041 key factor metrics
Peak Hour
8–9 AM
Night KSI Rate
19.6%
2.0× daytime
Ped Share
9.0%
58% KSI rate
Behaviour
31%
Speed or alcohol
Intersection · Factors · Signal Phase
Which signal phases are linked to crashes?
Intersection · Factors · Lighting
Does Dark-Unlit dominate KSI here?
Dark-Unlit: 6× higher KSI than daylight. No streetlight on Harbor Blvd NB arm.
Intersection · Factors · AI Synthesis
AI factors summary — INT-041
⏱ Timing
8–9am window dominates crash volume via approach-speed rear-ends on NB arm. PM peak adds left-turn conflicts in declining light. Two different countermeasures needed.
🚦 Phase
NB permissive LT responsible for 62 crashes including 31 KSI. Upgrading to protected-only during PM peak directly addresses the dominant movement-KSI pairing.
🚶 VRU
Pedestrian KSI rate 59% — 34 points above city average. LT vs ped conflict causes 14 of 44 ped crashes.
Intersection · Factors · Ranked Conditions
Which conditions are most dangerous here?
#1 Dark-Unlit + Any Type 59% KSI
Score 96 · Streetlight installation
#2 Fog + Night + LT 68% KSI
Score 89 · Variable message signs
#3 Alcohol + Night Drive 71% KSI
Score 82 · Late-night enforcement
#5 Speeding + AM Peak 68% KSI
Score 68 · Speed camera at junction
Supporting Cards — Intersection · Anomalies Tab
INT-041 · Main St & Harbor Blvd — peer comparison · spike detection · AI anomaly summary · deviation table
Intersection · Anomalies · KPIs
INT-041 anomaly headline metrics
vs Similar Ints
4.8×
Area rank #1
Largest Spike
+156%
Nov 2023
Rare High-KSI
2 types
Head-On + Ped
Corridor Delta
+38%
Above segment avg
Intersection · Anomalies · Peer Rank
Worse than similar intersections?
INT-041 ▶ THIS
14% KSI · 4.8×
INT-039 — Oak & 10th
9.7% · 1.8×
INT-044 — Pine & 1st
8.5% · 1.7×
INT-052 — Cedar & 3rd
7.2% · 1.5×
Peer avg (signalised)
3.2% · 1×
Intersection · Anomalies · AI Summary
AI anomaly synthesis — INT-041
🔵 Peer
Ranks #1 worst among 7 comparable signalised junctions. KSI rate 4.4× peer average. Gap is not explained by volume — adjusted for AADT.
📈 Spike
Nov 2023 produced +156% crash surge. Cause traces to contraflow diversion forcing extra HGV traffic with no signal phase adjustment.
🔴 Rare-KSI
Head-On (61% KSI) + Ped (58% KSI) represent only 14% of volume but 59% of all KSI. Frequency 3.2× peer average.
Intersection · Anomalies · Deviation Table
All anomaly dimensions ranked
Intersection · Anomalies · vs Corridor
Does this intersection diverge from corridor?
+38% above corridor mean — targeted junction treatment, not corridor-wide measures
08 · User Group
Strategy — Prioritisation → Investment Decisions
Value Enabled
Clear prioritisation of high-impact areas · Strong justification for investment · Alignment between data and planning
C
City Level
KPI tracking · Systemic risk · Investment allocation
Overview
- Is the city improving on key safety indicators over time?
- Which areas contribute most to overall risk burden?
- Where are the largest gaps between current and target performance?
Type
- Which crash types drive the most fatalities and serious injuries?
- Where should type-specific interventions be focused?
Factors
- What systemic issues are driving risk at scale?
- Which factor clusters are most amenable to intervention?
- Where will behaviour-change programmes have the greatest impact?
Anomalies
- Which areas require urgent intervention vs. planned investment?
- Where will investment yield the highest safety return?
- Which persistent hotspots have resisted previous efforts?
R
Corridor Level
Return on investment · Capital justification
Overview
- Which corridors offer the highest return on safety investment?
- Where can improvements be implemented quickly and cost-effectively?
Type
- Which crash types on this corridor are most preventable?
- What intervention types are best matched to the dominant pattern?
Factors
- What systemic corridor-level issues justify capital investment?
- How do infrastructure gaps compare in cost vs. impact?
Anomalies
- Which corridor segments are underperforming relative to investment already made?
- Where are the highest-leverage quick wins?
I
Intersection Level
Point prioritisation · Engineering vs. enforcement
Overview
- Why is this location a priority relative to others?
- What is the expected impact of intervention here?
Type
- What intervention type is best matched to the collision pattern?
Factors
- What causal factors are addressable through engineering or enforcement?
- How does cost of intervention compare to expected benefit?
Anomalies
- How does this compare to similar investments made elsewhere?
- Is the risk profile stable enough to justify long-term investment?