I took the train home to Stockholm from Skövde yesterday. The train was to my very small surprise late, somehow Swedish trains have a hard time dealing with snow.
The more interesting thing was how Swedish train company SJ calculates when the late train will arrive at its destination. The train started in Gothenburg, and had continuously become more late as it approached Skövde. But according to SJ, from Skövde and on we would not only start going faster, we would actually be going faster than normal.
As I found this highly unlikely, I figured it would be fun to follow up and see if the algorithm ever picked up that trains actually were slow today?
Below is a graph of the first calculated time that SJ presented. We left Skövde 25 minutes late and according to SJ we would make up 15 minutes on the rest of the journey. This included make zero minutes stops at the stations on our way to Stockholm.
As the graph shows we actually continued losing time the rest of the journey. But just like a few project managers I’ve had, SJ’s algorithm insisted on a sometimes ridiculous make up of lost time.
This is the graph showing the next to last updated timetable. It shows that we should make the last leg in four minutes rather than the 14 minutes it normally takes.
Funnily enough, SJ marked the train as “Arrived” when we still were a minute away from the station.
I found this very analogous to the behavior I’ve seen in a few projects, where time estimates are made to please the recipient rather than to provide actual information. As if I would feel better that SJ “tried” to go faster but couldn’t.
All I know now is that what they call their “calculated” time is actually “in a Unicorn world” guess.