Multiple linear regression (MLR) is best used when there is uncertainty in the slope of the linear relationship between an X and a Y, relationships between Y (output) and more than one X (input), and when we assume that the X's are independent of each other. MLR requires that the input variables, X's, are not correlated with each other to avoid issues like multicollinearity. It is also crucial for modeling relationships where multiple inputs are needed to explain the output.