Engineering mathematics

A local numerical optimizer returns different solutions when run from different starting points on a nonconvex objective. What is a plausible explanation?