Yeah I know you cannot track all journeys. Deutschlandticket, BahnCard 100, and similar things will be invisible. You can track the ones booked through the app though, and with a high enough sample rate you should get sufficiently close to the truth, unless there is a bias I haven't thought of yet.

You could also select random virtual journeys, I suppose. In arbitrary units you should at least be able to measure whether DB is improving or not. You could even delegate this to an independent organization. Actually, now that I think of it, isn't the API public? I'm reminded of the talk by D. Kriesel about DB data mining.