The first way to try to fix this is to apply logarithm to the observations on the dependent variable. This is going to make the dependent variable with high degree of kurtosis normal.
Note that sometimes, the resulting values of the variable will be negative. Do not worry about this, as it is not a problem. It does not affect the regression coefficients, it only affects the regression intercept, which after transformation, will be of no interest.
Remember the cost cannot be a number lesser than zero. so the domain must be such that the total cost C(m) is not negative. Also the range will always be 0.5 and above provided our domain is greater than or equal to zero.
