Road signs warning of the risk of animal-vehicle collision are undoubtedly the most widely implemented mitigation measure in the world. In fact, a person who knows anything about biogeography could tell where in the world he or she is just by looking at the animal on that sign. Unfortunately, although the cost of these signs is low, their effectiveness is also very low. Drivers, not finding a direct relationship between the warning and the real risk, tend to get used to these signs and lose their effectiveness. But this effectiveness could probably be increased if we are able to concentrate the warning only at times when a certain risk threshold is exceeded. To do this, we can take advantage of the fact that vehicle-animal collisions are concentrated in time and space. This is the idea behind a project being developed on the road network of Castilla y León (Spain) to reduce animal-vehicle collisions with roe deer, wild boar and red deer. In 13 road segments (30 signs in total) a new prototype of variable sign has been installed which receives in real time information about the probability of collision with animals from a central server. This probability is calculated on the basis of temporal models that include data on the date and time of day, the moon phase, weather conditions, hunting activity, etc. We expect this prototype, installed in 2022, to show considerable advantages over the measures currently in place to minimize animal-vehicle collision. In economic terms, it is much cheaper than other structural measures and, by focusing only on moments of real risk, it avoids driver habituation. It is a step forward in smart road and signaling projects, and can be expanded to other regions and species.