[endless1234]

>Given that, at peak times, some commuter services run once every ten minutes, and thus are ~always on time per this definition, it's wise to be a bit skeptical of the punctuality reporting.

I think that makes sense. If a commuter train leaves every 10 minutes, but due to some technical problems, the first train of the early morning leaves 20 minutes late which then continues through the day, does that mean every train is late? The users wouldn't know the trains they're using should have actually ran 20 minutes ago, they just know a train is to go every 10 minutes.

[iggldiggl]

The ideal measurement would be to get as good a model of day-to-day passenger flows as possible, then simulate the journey times based on both the nominal timetable, and the "actually-ran" timetable in order to also quantify the effects of missed connections.

I think the Swiss actually do something like this, whereas in Germany they've only comparatively recently started doing it, too, and even then I think they're currently only looking at long distance trains, so connections to/from regional trains don't count (and never mind local public transport).