Model deer collision probability feature
I am a wildlife biologist. Deer are creatures of habit, but those habits change with time of year and time of day, as well as geography. If people mapped deer sightings on the road, collisions or near-collisions in real time, you could build a model to predict likelihood of deer on the road based on the location/time/time of year. Mapping a deer sighting is not much use. But thousands of entries could build a predictive model for high likelihood hazards based on location and time. Deer crossing signs are OK, but they really only have value at certain times of the day and year. Waze has the capability, if enough people enter data, to make a much more useful feature. With each passing year it could become better and better. It would help drivers know when to be vigilant, which is valuable always, but especially when there is reduced visibility.
Just as one observer on a regular commute in Pennsylvania, I am able to predict in advance the places, time and month I most likely will encounter deer.