METHODOLOGY · RELEASE lens-2026-09-22-f53efbe918e1
ADAS data and methodology
This methodology governs how public vehicle, safety, road-outcome and research evidence becomes an ADAS tracker metric—and when it must stop for human review.
1. Evidence classes
Directly reported by an authoritative source.
Calculated directly from authoritative observed data.
Requires a documented mapping or assumption.
Research that informs interpretation but is not a UK fleet measurement.
2. Vehicle capability and fitment
The capability master stores model generation, applicable model years, equipment basis, source, protocol, confidence and review history. Every feature is classified as standard, optional, unavailable or unknown. Optional equipment is never converted to standard fitment, and unknown equipment remains unknown.
3. Model matching
DfT make, generic-model and detailed-model names are mapped to reviewed Euro NCAP model generations and applicable years. Matches can be exact, generation-level, manually mapped or unmatched. Admin work is prioritised by current vehicle population and latest-quarter registrations.
4. Penetration and coverage
Confirmed standard-fit share = vehicles confirmed standard fit ÷ all relevant vehicles. The denominator includes unmatched vehicles. Once the publication threshold is met, mapping coverage, optional or uncertain population, known non-fitment and unknown population are reported alongside the lower-bound share.
5. Geography and time
UK and Great Britain are separate data domains. GB data never overwrites UK data. Provisional quarterly figures remain visibly distinct from final annual statistics. Every chart shows the underlying data period separately from the date the Site last checked the source.
6. Euro NCAP protocol break
Pre-2026 Safety Assist and 2026+ Safe Driving / Crash Avoidance belong to different regimes. They are stored with protocol versions and never joined into a directly comparable series. Registration-weighted indices include only vehicles with a valid reviewed score and show represented registration coverage.
7. Road outcomes and exposure
STATS19 provides population-level GB context. Where compatible traffic exposure exists, rates use an appropriate DfT vehicle-mile denominator. These trends are never described as causal ADAS effects.
8. International incidents and research
NHTSA Level 2 reports are not exposure-adjusted. They are used as an early-warning signal and never as a manufacturer safety ranking. External studies retain country, population, period, outcome, uncertainty and UK applicability; US loss effects are not translated automatically into UK assumptions.
9. Source checks and updates
The shared Lens pipeline runs hourly and checks sources that are due, using each source’s configured interval. It archives retrieved documents and flags changes for review. A source check does not automatically import new numeric tables.
Numeric updates require reviewed source observations and explicit data periods. The previously separate ADAS watcher jobs have been superseded by this shared pipeline.
10. Validation and publication
Publication checks required fields, numeric validity, citations and shared dataset consistency. Failed source checks, unresolved conflicts and unreviewed changes retain the affected group’s previous release. New numeric imports, model mappings and changes to the Lens assessment require review. Editorial corrections preserve the original evidence provenance.