How iRAP calculates Star Ratings
From road images every 100 metres to risk scores and Star Ratings for each road user group
Rating the road, not the driver
New cars receive stars from crash tests (NCAP), which show how well a car protects its occupants. iRAP Star Ratings apply the same idea to roads: they look at how much the road's features allow crashes to happen, and how likely those crashes are to cause death or serious injury.
The assessment does not wait for crashes to happen. It uses road attributes recorded from survey images, so it works both on roads with no crash records and on designs for new roads.
The risk factors in the model come from road safety research around the world, and iRAP's Global Technical Committee keeps them under continuous review.
- Every 100-metre road segment is assessed
- Results are reported for 4 road user groups
- Based on 13 crash types
- Driver behaviour is not included, except operating speed
From road survey to Star Ratings
- 1
Road survey
A survey vehicle records geo-referenced images along the whole route, no more than 20 metres apart.
- 2
Road attribute coding
Trained coders review the images and record around 60 road attributes for every 100 metres, following the iRAP Coding Manual.
- 3
Risk score calculation
Road attributes are converted into risk factors, and a Star Rating Score (SRS) is calculated for each crash type and road user group.
- 4
Star Rating banding
Each SRS is compared with the Star Rating bands to give 1 to 5 stars, both for each 100-metre segment (raw) and smoothed along the route (smoothed).
- 5
Investment planning
Fatal and serious injuries are estimated, and cost-effective countermeasures are selected (Safer Roads Investment Plan, SRIP).
Four road user groups and their crash types
Each road user group is involved in crashes in different ways, so iRAP calculates each group separately, using the crash types that cause the most deaths and serious injuries for that group. The same road can therefore be 4 stars for vehicle occupants but 1 star for pedestrians. Click a crash type to see which road attributes are used to calculate it.
Rear-end crashes and sideswipes away from intersections are not yet included in the current Star Rating model. On this page, right and left refer to driving on the left, as in Thailand (driver side = right, passenger side = left).
The Star Rating Score formula
The SRS is the sum of the risk scores for all crash types of a road user group. Each crash type score is the product of five factors. The higher the score, the higher the risk and the fewer the stars.
Likelihood
Attributes that make a crash more or less likely to start, such as curvature, lane width, road condition, and line marking and signs.
Severity
Attributes that determine how badly people are hurt once a crash happens, such as trees or poles close to the road, safety barriers, median type and sidewalks.
Operating speed
The higher of the speed limit and the speed that 85% of vehicles travel at or below (85th percentile). Speed raises both likelihood and severity at the same time.
External Flow Influence
A crash with someone else can only happen when someone else is on the road. This factor is based on traffic volume per lane, or the volume on the intersecting road.
Median Traversability
Used only for vehicle occupants and motorcyclists, for driver-side run-off and head-on crashes. A median that vehicles can cross has a value of 1; a median with a safety barrier or trees has a value of 0.
Because the factors are multiplied, if any factor is zero the risk for that crash type disappears completely. For example, a road with a median barrier has no head-on overtaking risk. No score is calculated for a group with no road users at all, or for segments with major roadworks.
From scores to Star Ratings
The SRS of each road user group is compared with the Star Rating bands below. Vehicle occupants and motorcyclists share one set of bands; bicyclists and pedestrians use another, because their scores are calculated differently.
Bands follow iRAP model v3.10 (the lower value is included in the band, the upper value is not). This model version also adds decimal Star Ratings, such as 2.8 stars, to show where a score sits within its band, and it revised the bands for bicyclists and pedestrians. Assessments made with earlier model versions may therefore use different bands for these two groups.
The global target is for roads to be at least 3 stars for all road users.
| Star Rating | Vehicle occupants Motorcyclists | Bicyclists Pedestrians |
|---|---|---|
| ★★★★★ | 0 – 2.5 | 0 – 3 |
| ★★★★ | 2.5 – 5 | 3 – 9 |
| ★★★ | 5 – 12.5 | 9 – 24 |
| ★★ | 12.5 – 22.5 | 24 – 54 |
| ★ | 22.5 and above | 54 and above |
A real example: one 100-metre segment
A segment of a rural national highway from a ThaiRAP assessment. It is a two-lane road with direction separated by line marking only, straight, with a good road surface.
Recorded attributes
- Speed limit / operating speed (85th percentile)90 / 95 km/h
- MedianCentre line (traversable)
- Roadside object, both sidesTrees 1–5 m from the lane edge
- Paved shoulder, driver side / passenger sideNarrow (< 1 m) / medium (1–2.4 m)
- Lanes1 lane in each direction, ≥ 3.25 m wide
- Curvature / road condition / delineationStraight / good / adequate
- Intersection / property accessNone
- Vehicle flow (AADT)About 1,200 vehicles per day
Attributes with a large effect on this segment's score
Vehicle occupant SRS by crash type
The total vehicle occupant SRS is 14.75, within the 12.5–22.5 band, so the segment is 2 stars (the decimal Star Rating of 2.8 means it is near the top of the 2-star band, not far from 3 stars).
Almost 9 out of 10 parts of the score come from run-off crashes, even though the road is straight with a good surface. A vehicle that leaves the road at around 95 km/h meets trees within 1–5 metres on both sides. The head-on risk is low because traffic volume is still low.
A result like this shows that the most targeted countermeasures deal with the roadside, such as widening the clear zone, installing safety barriers or widening the shoulders.
Why speed matters so much
In the iRAP model for vehicle occupants and motorcyclists, crash likelihood rises with speed, and severity rises with the square of speed. Multiplied together, risk rises much faster than speed, so even a small increase in speed can cost a road a star.
For pedestrians the effect is even clearer: around 90% of pedestrians are killed when hit by a vehicle at 80 km/h or more.
Try changing the speed
Examples from published speed research. These are not values from the iRAP model, but they show the same principle: small changes in speed have a large effect on deaths and injuries.
Lower the mean speed: how many fewer deaths?
Nilsson's Power Model, updated by Elvik from many studies, says the number of deaths changes with the ratio of mean speeds raised to a power. The exponent is higher on rural roads than on urban roads.
Impact speed and the chance of survival
The human body can only tolerate limited impact forces. The Safe System approach therefore uses the impact speed at which the risk of death starts to rise sharply (around 10%) as a design threshold for roads.
Exponents: Elvik (2013) and Nilsson (2019), cited in Transport Scotland, National Speed Management Review 2024, Appendix F · Impact speeds: Wramborg (2005), cited in Austroads, The Safe System Approach · Pedestrian risk of death: Tefft (2011), AAA Foundation for Traffic Safety, US data; older people are at higher risk than average.
What is the difference between raw and smoothed?
The model calculates a Star Rating for every 100-metre segment. This is the raw result, which rises and falls with local features, such as an intersection or trees close to the road.
The smoothed result evens out the scores over a longer length of road, so it is continuous and easy to read. It is used to report on a whole route or to track progress when the road is reassessed.
Use raw results to find locations that need treatment, and smoothed results to report the overall picture or set targets.
Road attributes assessed
Choose a road user group and a crash type to see which attributes affect crash likelihood (L) and which affect severity (S). Click an item to read its description.
L Likelihood
S Severity
Speed and traffic
Speed limit / Operating speed Vehicle occupantsMotorcyclistsBicyclistsPedestrians
The higher of the speed limit and the speed that 85% of vehicles travel at or below. One of the most influential factors. Try changing this attribute →
Speed differential Vehicle occupantsMotorcyclists
Car speeds differ from truck or motorcycle speeds by more than 20 km/h, which leads to more overtaking. Try changing this attribute →
Speed management Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Devices or designs that reduce speeds by at least 10 km/h below the speed limit, such as speed humps or lane narrowing. Try changing this attribute →
Vehicle flow (AADT) Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Average daily traffic. The model converts it to vehicles per lane: the more traffic, the higher the chance of a crash with another road user. Try changing this attribute →
Intersecting road volume Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Traffic volume on the side road that joins the road. Used for intersection crashes. Try changing this attribute →
Motorcycle % MotorcyclistsBicyclistsPedestrians
If there are no motorcycles, no score is calculated for this group. A high share also increases the risk for pedestrians and bicyclists crossing the road. Try changing this attribute →
Heavy goods vehicle (HGV) % BicyclistsPedestrians
Many trucks increase severity when pedestrians or bicyclists are hit (added in model v3.10). Try changing this attribute →
Pedestrian / bicycle flows BicyclistsPedestrians
If no pedestrians or bicyclists use the road at all, no score is calculated for that group. Try changing this attribute →
Cross-section
Number of lanes Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Used to calculate vehicles per lane, overtaking opportunities, and the number of lanes pedestrians must cross. Try changing this attribute →
Lane width Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Narrow lanes make it easier for vehicles to drift out of their lane. Try changing this attribute →
Median type Vehicle occupantsMotorcyclistsBicyclistsPedestrians
What separates the two directions of traffic, from line marking to wide medians or safety barriers. If overtaking is not possible, head-on overtaking risk is zero. Try changing this attribute →
Paved shoulder Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Sealed area outside the edge line. It gives errant vehicles space to recover before reaching a hazard, which reduces run-off severity, and it is a place to walk or cycle where there is no sidewalk or bicycle facility. Try changing this attribute →
Shoulder / centreline rumble strips Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Strips on the shoulder or on the centreline / median that make vehicles vibrate to warn drivers when they leave their lane. Try changing this attribute →
Service road Vehicle occupantsMotorcyclists
A parallel road that separates access traffic from the main road. Try changing this attribute →
Vehicle parking BicyclistsPedestrians
Parked vehicles block sight lines and obstruct pedestrians and bicyclists. Try changing this attribute →
Alignment and surface
Curvature Vehicle occupantsMotorcyclistsBicyclistsPedestrians
How curved the alignment is. Very sharp curves increase the chance of run-off crashes. Try changing this attribute →
Quality of curve Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Whether drivers can tell how sharp the curve is, and whether there are enough warning signs and markings. Try changing this attribute →
Grade Vehicle occupantsMotorcyclistsBicyclistsPedestrians
The uphill or downhill slope along the road. Try changing this attribute →
Road condition Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Whether the surface is smooth and even, or has potholes that could cause loss of control. Try changing this attribute →
Skid resistance Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Whether the surface gives good grip or is slippery, for example with loose gravel or bleeding asphalt. Try changing this attribute →
Delineation Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Line marking, signs and reflective road studs that show the road ahead. Try changing this attribute →
Sight distance Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Whether drivers can see the road and other road users far enough ahead. Try changing this attribute →
Street lighting Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Lighting on the road at night. Try changing this attribute →
Roadside
Roadside severity – object Vehicle occupantsMotorcyclistsBicyclists
The most hazardous object closest to the road, such as trees, poles, deep drains, cliffs, or a safety barrier that reduces severity. Try changing this attribute →
Roadside severity – distance Vehicle occupantsMotorcyclistsBicyclists
Distance from the lane edge to the object. The closer it is, the less time there is to regain control before impact. Try changing this attribute →
Intersections and access
Intersection type Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Three-leg or four-leg intersections, signalised or not, with or without turn lanes, roundabouts, U-turn points and railway crossings. Try changing this attribute →
Intersection channelisation Vehicle occupantsMotorcyclistsBicyclists
Islands or surface markings that guide vehicle paths at intersections. Try changing this attribute →
Intersection quality Vehicle occupantsMotorcyclistsBicyclistsPedestrians
The design, advance warning signs and road markings of the intersection. Try changing this attribute →
Property access points Vehicle occupantsMotorcyclistsBicyclistsPedestrians
Driveways to houses and shops, and minor lanes. The more there are, the more often vehicles turn in and out. Try changing this attribute →
Pedestrians, bicyclists and motorcyclists
Crossing facility BicyclistsPedestrians
Crossings provided for pedestrians and bicyclists, from zebra crossings to footbridges or signalised crossings. Try changing this attribute →
Crossing quality BicyclistsPedestrians
Whether drivers can see the crossing clearly, and whether there are warning signs. Try changing this attribute →
Pedestrian channelisation Pedestrians
Fencing or barriers that guide pedestrians to cross at the crossing. Try changing this attribute →
Sidewalk provision Pedestrians
A sidewalk separated from the traffic lanes. Where there is no sidewalk, the model considers the shoulder instead. Try changing this attribute →
Facilities for motorcycles Motorcyclists
Lanes or paths for motorcycles only. If motorcycles are separated from larger vehicles, the risk of being hit while riding along the road disappears. Try changing this attribute →
Facilities for bicycles Bicyclists
On-road or off-road bicycle facilities. Where there is none, the model considers the shoulder instead. Try changing this attribute →
School zone warning BicyclistsPedestrians
School zone signs or markings, including flashing warning lights. Try changing this attribute →
School zone crossing supervisor BicyclistsPedestrians
Staff who help children cross the road in the morning and afternoon. Try changing this attribute →
No attributes match your search in this view
Every crash type uses speed (V) as a factor · F = flow factor (External Flow Influence) · M = median traversability factor · Conditions = no score is calculated if there are no road users of that group, and no crash type is calculated where there are major roadworks.
As well as the items above, coders record other data that have no risk factor of their own, such as land use on both sides of the road, area type (urban or rural, used to choose the right risk factors), roadworks (no score is calculated where there are major roadworks), upgrade cost, and Star Rating policy targets. These data are used to estimate road user flows, to smooth results and for investment planning.
Try changing road attributes
Choose a road attribute and change it from its current condition to a new one, to see which crash types and road user groups are affected, and whether the risk score (SRS) goes up or down.
This section shows only the direction of the effect; it does not calculate actual scores. The size of the effect depends on the other attributes of the segment, and the Star Rating only changes when the score crosses a band threshold. Actual assessments are calculated with iRAP's ViDA software.
Frequently asked questions
Does a 1-star road mean it should not be used?
No. A Star Rating shows the level of risk that road users will be killed or seriously injured because of the road's features, so that agencies can decide which sections to improve first. A 1-star road can still be used, but it should be improved.
How is this different from black spot analysis?
Black spot analysis looks at crashes that have already happened. Star Ratings look at the road's features, so they work even where there are no crash records, or on designs that have not yet been built. The two methods complement each other.
Why does a road with no recorded crashes get few stars?
Serious crashes on a short section of road are rare and vary a lot from year to year, so the absence of crashes does not mean the road is safe. If high-risk features remain, such as trees close to a high-speed road, the chance of a serious crash remains too.
Why does the same road get different Star Ratings for different road user groups?
Each group is assessed with different crash types and road attributes. A road with a median barrier may be safe for vehicle occupants, but if it has no sidewalks or crossings, pedestrians are still at high risk.
Does the score include driver behaviour?
No, except for operating speed. A Star Rating measures how well the road prevents crashes and reduces their severity when road users make mistakes, in line with the Safe System approach.
Can a Star Rating change?
Yes, when the road is improved, for example by installing safety barriers, building crossings or reducing speeds. Reassessments use the same road segmentation so that before and after results can be compared.
Are Star Ratings related to actual crashes?
Studies have found that roads with higher Star Ratings have lower crash costs per vehicle-kilometre travelled, and that raising Star Ratings reduces crashes and their severity.
Sources
iRAP Methodology Reference Guide: Star Rating, FSI Estimation and Safer Road Investment Plan models v3.10, and iRAP Coding Manual v3.10 (Drive on Left), by the International Road Assessment Programme. Read more at irap.org.
ThaiRAP presents this content as an organisation with a memorandum of understanding with iRAP. iRAP's models, methods and marks are the intellectual property of iRAP.