Methodology

Every figure here is computed deterministically from the official DVSA MOT record, with no scraping, sampling or estimation. If you run the same public files, you should get the same numbers.

Source

All statistics are computed from the DVSA anonymised MOT tests and results dataset, the official record of every MOT test conducted in Great Britain. We use test classes 1 and 2 (motorcycles) only. Nothing is scraped, sampled, or estimated.

Filters

Pass rates count normal tests only. Retests, partial retests and aborted tests are excluded; including them flatters pass rates, since a retest follows a repair. Rows with corrupt fields (a fraction of a percent of the source data) are excluded.

Definitions

Pass: passed first time with no defect rectified at the station.
PRS: pass after rectification: a defect was found and fixed during the visit. We count PRS separately from clean passes.
Fail: failed the test outright.
Failure groups: defects are grouped by the DVSA inspection manual section (brakes, lamps, tyres…). Advisories are not defects and are excluded from failure counts.

What the pass rate measures. The MOT is a roadworthiness test of the items in the DVSA inspection: brakes, lights, tyres, steering and suspension, structure and emissions. It is not a full mechanical inspection — engine and gearbox internals, clutch and other wear the MOT does not assess are not captured. The pass rates here are therefore a roadworthiness signal, a strong and consistent proxy for how bikes hold up in service, rather than a complete measure of mechanical reliability.

Adjusted reliability (vs expected)

A raw pass rate is confounded by age and mileage: a cherished classic ridden a few hundred miles a year will out-pass a hard-used modern commuter, even if the commuter is the better-built machine. The adjusted figure corrects for this by indirect standardisation.

Every test is placed in a stratum by age band × mileage band × test class. For each model we compute the pass rate expected for its own mix of strata — the fleet-wide pass rate in each stratum, weighted by how many of the model's tests fall in it — then report actual − expected in percentage points (pp). A positive number means the model beats comparable bikes; near zero is typical; negative trails them. Only tests with a usable age and mileage are scored, and we hold models to a minimum scored-test count before publishing. We always show the raw pass rate alongside it.

Honesty notes

MOT data measures bikes as maintained by their owners, at three-plus years old. It is a reliability signal, not a lab test. Small samples are marked and low-volume models are held to minimum test thresholds before we publish a rate. Make and model names are as recorded by testers, then canonicalised: misspellings and model-strings-entered-as-makes are folded into the real marque (e.g. "ROYAL ENFEILD" and "HARLEY-DAVIDSON VRSCDX NIGHT ROD" → Royal Enfield, Harley-Davidson), and spelling/spacing variants of a model are merged into one family.

Questions about the method? The whole pipeline is deterministic from the public dataset; if you recompute it, you should get our numbers.