Before we start to understand data, it worth understanding how do we usually "prepare" it. Raw data might include letters and special characters. Any algorithm/ computer/ ML model needs a numerical input. Moreover, those inputs should not be too big, in order to avoid computations overflow and gain some numerical stability. Due to those remarks we propose the following steps to "understand" data in general:
Raw -> Numerical (conversion to numerical),
Numerical -> Normalized,
Normalized -> Discrete, Continuous parts
Discrete, Continuous -> ML Model