What a First Generation Taught Us About Failure Rates
This is an uncomfortable article to publish, which is usually a sign that it is worth publishing.
The first generation of Linky had a failure rate above 20%. More than one board in five came back. Today that figure is around 5%, and the difference between those two numbers is most of what we have learned as a company.
What "failure rate" actually means
First, the honest definition, because the term gets used loosely. We count a failure as any board that needed intervention beyond normal wear — a part replaced, a repair, a return. Not scratched decks. Not worn belts, which are consumables. Anything that stopped the vehicle from doing its job.
By that definition, one in five was unacceptable. It was also, at the time, invisible to us until the boards were already in customers' hands on another continent.
The four things that went wrong
Distance hid the problem. Production was abroad. A defect that a person standing at the line would have caught in a morning travelled halfway round the world before anyone saw it. By the time a pattern was visible in the returns, several months of boards had already shipped.
Validation happened too late. Components were qualified on paper and on the bench, but the combination of them — motor, controller, battery, structure, all vibrating together on a bad road — was validated by customers. That is the most expensive test lab in the world.
The electronics carried too much. Early architectures put more into fewer, less accessible places. When something failed, the diagnosis was slow and the fix was often "replace the assembly", which is a repair strategy only if you have assemblies to spare.
We had no way to fix things quickly. This is the one that hurt most. A board that can be repaired near the customer is an inconvenience. A board that has to cross an ocean to be looked at is a lost customer.
What changed
Every one of those four has a countermeasure in Linky 2.0, and all four are structural rather than motivational.
The supply chain moved close. Mechanical parts are now made about 15 km from assembly, the wheels 13 km, the electronics 160 km. A defect found on a Tuesday is a conversation on a Wednesday, in person, next to the machine.
Testing moved earlier. The board is validated as a system, not as a parts list — including sealing to IP65 against dust and water, a homologated rider weight of 100 kg, and drop and abuse testing on the folding structure, which is the part that would fail worst and so is tested hardest.
The architecture got simpler to service. The electronics are organised so that a fault can be isolated and a part replaced, rather than a whole section condemned.
Repair became possible again. Boards are repaired in Italy. The consumables that used to strand people — belts above all — now ship two spares and a tool kit in every box, and the battery is designed to be replaced rather than to end the life of the vehicle.
Why publish the bad number
Because "reliable" is a word anyone can print, and a number is harder to fake. Because a company that only tells you about its second generation is hoping you never ask about its first. And because 5% is not zero, and we would rather say so than be caught saying otherwise.
We will keep publishing the figure as it moves. If it goes up, you will read it here.
Related reading
Share
