SNCF: 1,300 platforms of 8,700, measured from the wrong thirty years

In 2014 the French national rail operator took delivery on an order for 2,000 regional trainsVERIFIED(Verified) . The trains were built exactly to specification.

The specification was the problem.

The measurement was correct. The sample was not.

Before ordering, the operator needed to know how much room there was between a platform edge and the track. It asked the company that owned the track and the stations, Réseau Ferré de France. It received dimensions. It ordered trains that fitted them.

Most French platforms are older than thirty years, and older platforms were built closer to the track.

Nobody measured anything incorrectly. Every figure handed over was an accurate measurement of a real platform. The figures simply described a subset of the network that did not resemble the network.

The trains were slightly too wide for some of the stations they had been bought to serve. Not everywhere: 1,300 platforms out of 8,700VERIFIED(Verified) had to be modified.

The number in the headlines

This case is the reason the site has a cut register.

The 15-billion number is not an invention. It appeared in headlines, it is a real euro amount attached to a real contract, and anyone repeating it can point at a newspaper. It is still wrong for the purpose it gets used for, and no number of outlets running it makes it right, because they are all running the same confusion.

That is the difference between counting sources and counting organisations. Three outlets quoting one wire story is one source. Three outlets making the same category error is not corroboration either.

What actually failed

There was no unit mismatch here. Nobody wrote feet where metres were required. The failure is one level up: a question was asked, answered accurately, and the answer described a different population than the asker assumed.

It is the most ordinary failure on this site, and the most transferable. Every specification rests on measurements of something. The something is rarely written down.

The next case is the opposite. Four measurements, four carpenters, and all four were right.