A vehicle's safety rating comes from a defined programme of staged collisions rather than from real-world outcomes. Understanding what is tested explains both what the rating tells a buyer and what it cannot.
The tests are standardised scenarios
Programmes run a set sequence, typically including a frontal impact into a deformable barrier, an offset frontal test, a side impact and a pole strike.
Each is conducted at a specified speed into a specified obstacle, because comparability requires that every vehicle meets identical conditions.
Real collisions occur across a continuous range of speeds and angles, so the tests represent selected points within that range rather than an average of it.
Dummies are measuring instruments
The occupants are engineered devices carrying sensors that record forces, accelerations and displacements at the head, neck, chest, pelvis and legs.
Different dummy types represent different body sizes, and programmes have expanded the range used because a restraint system tuned for one build performs differently for another.
Readings are compared against injury criteria derived from biomechanical research, and those criteria are what convert a physical measurement into a safety assessment.
Structure and restraints are assessed separately
Inspectors examine how the passenger compartment deformed, whether doors remained closed and operable, and how the steering column and pedals moved.
The restraint system is judged on how it managed the occupant's motion, including belt behaviour and airbag timing relative to the impact.
A vehicle can score well on measured forces while showing structural intrusion that would matter in a more severe collision, which is why both are reported.
Active systems now form part of the score
Programmes increasingly evaluate systems intended to prevent or reduce a collision, including automatic braking and lane keeping, using track scenarios with targets.
Protection of people outside the vehicle is assessed as well, through impacts against the bonnet and bumper representing pedestrian contact.
Weighting between these areas changes over time, which is why ratings from different years are not directly comparable even within one programme.
A star rating compresses a great deal
The headline score aggregates many measurements, so two vehicles with the same rating can have quite different profiles across the underlying tests.
Ratings also compare a vehicle against the protocol rather than against other vehicles, and mass differences mean a collision between two vehicles is not symmetric.
Detailed results are published alongside the summary, and they carry information about specific scenarios that the star figure necessarily discards.