2. Rating architecture

Euro NCAP does not publish a single "safety number." It publishes a star rating that is derived from a structured, multi-pillar scoring architecture. Understanding that architecture is essential for engineering validation, because it determines which test results actually influence the final rating, how much each result is worth, and where a weak area can be compensated — or cannot. This section explains the rating logic: the four assessment pillars, how pillar scores are computed, how they are weighted into an overall score, and how the overall score maps to a star rating.

The four assessment pillars

Since the 2020 protocol, Euro NCAP organizes assessment into four pillars. Each pillar addresses a distinct safety domain and is scored independently before being combined:

• Adult Occupant Protection (AOP) — frontal offset, full-width frontal, side barrier, side pole, and whiplash/far-side tests, plus restraint-system assessment.
• Child Occupant Protection (COP) — child restraint installation, dynamic performance in frontal and side impacts, and vehicle-based child presence/integration features.
• Vulnerable Road User (VRU) Protection — pedestrian and cyclist head, upper-leg, and lower-leg impacts, plus active VRU avoidance functions.
• Safety Assist (SA) — active safety systems such as AEB, lane support, speed assistance, and occupant monitoring.

Each pillar produces a percentage score. The pillar percentages are then combined using fixed weights to produce the overall score, expressed as a percentage. The star rating is assigned from that overall score, subject to minimum pillar requirements.

From test results to pillar scores

Within a pillar, individual tests contribute points. A test result is first converted into a normalized score, then aggregated. For example, in AOP the frontal offset test contributes a defined maximum point value; the measured dummy injury criteria (such as HIC, chest deflection, and femur loads) are compared against capping limits, and points are awarded on a sliding scale between a lower performance limit and an upper capping limit. A result at or better than the capping limit earns full points; a result at or worse than the lower limit earns zero.

A simplified illustration of the point-award logic for one injury criterion is:

criterion: chest deflection (mm)
lower performance limit: 50 mm  -> 0 points
capping limit:           32 mm  -> full points
measured value:          41 mm

score fraction = (50 - 41) / (50 - 32) = 9 / 18 = 0.50
awarded = 0.50 x max_points_for_criterion

This linear interpolation is a teaching simplification; the actual protocol uses criterion-specific curves and capping values. The key engineering point is that performance is graded, not pass/fail. A result slightly better than the lower limit earns partial credit, so incremental improvements in restraint tuning or structure still translate into rating points.

Weighting and the overall score

Once each pillar percentage is known, the overall score is a weighted sum. The weights reflect Euro NCAP's policy priorities and have shifted over time toward VRU and Safety Assist. A representative weighting used for explanation is:

With these weights, an overall score is computed as:

overall = 0.40*AOP + 0.20*COP + 0.20*VRU + 0.20*SA

example:
AOP = 85%, COP = 80%, VRU = 70%, SA = 75%
overall = 0.40*85 + 0.20*80 + 0.20*70 + 0.20*75
        = 34.0 + 16.0 + 14.0 + 15.0
        = 79.0%

The weights above are illustrative and must be confirmed against the protocol edition applicable to the vehicle's assessment year. Weights and capping limits are revised periodically, so validation targets must always be tied to a specific protocol version.

Star rating thresholds and minimum pillar requirements

The overall score maps to stars through threshold bands. A commonly cited structure is: 5 stars at or above roughly 80%, 4 stars around 70%, 3 stars around 60%, and lower bands below that. However, the overall score alone does not determine the rating. Euro NCAP applies minimum pillar requirements: a vehicle cannot achieve a given star level if any single pillar falls below a defined floor, even when the weighted overall score would otherwise qualify.

This capping rule has direct engineering consequences. It means the rating problem is not a single optimization but a constrained multi-objective problem: each pillar must clear its floor, and the weighted total must clear the star band. A vehicle at 82% overall but with VRU at 55% may be capped below 5 stars, whereas a vehicle at 80% with all pillars above their floors achieves 5 stars.

Reading the rating architecture during validation

During engineering validation, the rating architecture should be used as a planning and diagnostic tool, not just a final label. A practical workflow is:

1. Fix the protocol edition and record its weights, capping limits, and minimum pillar thresholds.
2. Build a pillar-level scorecard from the latest test and simulation data.
3. Identify the binding constraint — the pillar closest to its floor, or the criterion contributing the largest point deficit.
4. Estimate the overall score and check both the star band and every pillar floor.
5. Prioritize countermeasures by points-per-cost within the binding pillar before optimizing already-strong pillars.

The SVG below summarizes the flow from test evidence to pillar scores, the weighted overall score, and the star decision with its capping logic.

Why the architecture matters for engineering decisions

Three practical consequences follow from this structure. First, point deficits are not interchangeable across pillars: recovering 5 points in AOP is worth 2 points overall, while recovering 5 points in VRU is worth only 1 point overall — but VRU may be the binding floor constraint, making it the higher priority despite lower weight. Second, partial credit rewards incremental tuning: restraint optimization that moves chest deflection from 45 mm to 38 mm can convert a zero-point criterion into a partial or full score, which is often cheaper than structural redesign. Third, protocol version control is mandatory: weights, capping limits, and floors change between editions, so a validation target copied from an older program can mislead the entire development plan.

In the next stage of validation, this architecture becomes the scorecard against which simulation and physical test results are tracked, pillar by pillar, criterion by criterion, until every floor is cleared and the overall score reaches the target star band.