One City Saved $5M by Routing School Buses with an Algorithm

I have found that Google Maps (and to a lesser extent, Waze and Apple Maps) are less efficient even with near real time traffic data, than I can be -- with the caveat that I know the area and listen to local radio for traffic updates. Without that, I defer to the algorithm, but with the additional data I have outperformed the estimates provided by the computers far more often than not.

However, this is because I used the path finding algorithms in the first place; I am constantly using the systems, examining each potential path, then adding my own mental map of maybe uncaptured data: how long is this light at different times of the day? what triggers the light, a weight or magnetic or visual sensor or is it purely times? what are the chances of slowly accelerating semi trucks getting through this intersection before the timer turns the light red? if I have am turning right on a red but into the right lane and I have 300m to get to the left lane cutting across 4 through lanes, what delay will this cause if I assume I get a red or mistime oncoming traffic? is this intersection or road known for speeding or red light running or block-the-box behavior? and on, and on.

I did my university capstone on transit technologies. It was quite interesting to learn that a lot of route planning was a map and a ruler in the basement of City Hall.