especially when there isn't good theoretical hypothesis behind it. usually, one assumes certain taylor expansion which converges (i.e. coefficents of higher order terms diminishing)