Maybe I should have said "cannot be done reliably". Even a very inaccurate and bad estimation process will yield correct estimates at least occasionally - stop

Maybe I should have said "cannot be done reliably". Even a very inaccurate and bad estimation process will yield correct estimates at least occasionally - stopped clocks and all that. But estimate-driven budget allocation is hardly useful if the result is wrong 50% of the time, and to me it feels like it's at least that unreliable. Although this story focuses on Germany, look at CrossRail in the UK for another pertinent example. It was "on time and on budget" right up until a few months before launch, when it suddenly wasn't on time or on budget anymore, and in fact, after churning through some CEOs, eventually they got one willing to admit the truth - that nobody knew how long it would take to complete. But by then they had started shutting down the project, e.g. by laying off their entire internal and external communications team. It'd have been far more effective if they'd admitted up front they didn't know exactly how long it'd take.