How we test | Duskmile

The fixed loop

Range is measured on one 12 km circuit we ride for everything: four kilometres of 50 km/h arterial road, five of stop-start town traffic with 21 signalled junctions, and three of river path with 140 metres of cumulative climbing. Rider weight is 78 kg with a 4 kg bag. Tyres are set to the pressure printed on the sidewall. The assist level is the highest the machine offers, because that is the setting people actually use once the novelty wears off.

We ride the loop until the pack cuts out or drops below the point where assist becomes token, then divide. Cold-weather figures, when we quote them, come from the same loop between 4 °C and 8 °C.

Braking

Ten stops from 25 km/h on dry asphalt and ten on wet, measured from first lever movement to standstill, taken in sequence with no cooling pause. A machine that stops in 3.4 metres on run one and 4.9 metres on run eight has a fade problem, and the review says so. Anything sold with mechanical discs gets an extra set of runs after a week of rain, because that is when cable brakes reveal themselves.

Four weeks, not a press ride

Nothing is scored on a launch day. Every vehicle spends at least four weeks as somebody’s actual transport: locked to real racks, left out in real weather, charged from a real wall socket in a real hallway. That is where the failures that matter show up — the mudguard that rattles loose, the display that washes out in direct sun, the folding latch that needs two hands and a knee.

What the score weighs

Ride and handling — 25%Stability at speed, behaviour over broken surfaces, what the motor does when you need it to do nothing.
Measured range — 20%Our loop figure against the claim, and how honest the battery gauge is about the last 20%.
Brakes and safety — 20%Stopping distance, fade, lighting that actually lights the road rather than announcing your presence.
Build and serviceability — 20%Fastener quality, cable routing, whether a local shop can get parts, whether the battery is replaceable at all.
Value — 15%What it costs the week we publish, against what else that money buys. This is the component most likely to move a score later.

Money and independence

We buy most of what we test. When a manufacturer lends a machine we say so in the review, we keep it for the full four weeks, and we return it. We do not accept payment for a review, we do not send drafts to manufacturers, and no brand has ever seen a score before publication.

Some outbound product links may earn a commission. The commission is identical whether the score is a 4.1 or a 9.2, which is the only arrangement that makes the scores worth reading.

When we get it wrong

Corrections appear on the review itself, dated, rather than quietly. If a firmware update or a recall changes what a machine is, we re-test and move the score, and the review notes the old number.