A car's crash rating is presented as a single figure, but it is assembled from several separate assessments measuring different things. Understanding the components explains why two cars with the same headline score can differ substantially.
What the tests represent
Frontal tests are conducted both with the full width of the car striking a barrier and with only part of it, because the two produce very different loads.
An offset impact concentrates force through one side of the structure, which is a more demanding test of how the body distributes energy.
Side impacts and pole tests examine the area where there is least space between an occupant and the outside of the car, and rear tests assess whiplash protection at lower speed.
How injury is measured
Instrumented dummies record forces and accelerations at the head, neck, chest, abdomen and legs during each impact.
Those readings are compared against thresholds derived from biomechanical research, and points are deducted where a reading approaches or exceeds a limit.
Multiple dummy sizes are used, since a structure tuned around an average adult male may protect a smaller occupant less effectively.
Why the score covers more than the crash
Ratings now combine occupant protection with assessments of protection for pedestrians and cyclists, and with the performance of driver assistance systems.
Pedestrian assessment examines bonnet and bumper design, since the shape and give of the front of a car determines the injury caused in a collision.
Assistance testing measures whether automatic braking and lane systems actually intervene in defined scenarios, which is protection against the crash occurring at all.
Why comparisons across ratings mislead
Test protocols differ between regions, so a rating awarded in one market does not translate directly to another.
Protocols are also revised regularly, and thresholds are raised, so a car rated highly several years ago would not necessarily achieve the same score under current requirements.
Comparing scores awarded under different versions of a protocol is therefore not a like-for-like comparison at all.
What the rating does not capture
Tests are conducted against fixed barriers of defined mass, which is not the same as a collision between two vehicles of very different weights.
A heavier vehicle carries an advantage in a real collision that its rating does not describe, since it is measured against the same barrier as everything else.
Real crashes also occur at angles and speeds outside the tested set, so a rating describes performance in specified conditions rather than a general guarantee of outcome.