In feature scaling, the goal is to transform each variable in a dataset independently, so that afterwards, all variables are represented with a similar range of numbers.