This research contributes to the ongoing discussion on functional flexibility and theoretical consistency by comparing the empirical performance of two second order flexible functional forms - the Symmetric Generalized McFadden and the Transcendental Logarithmic. It proposes an estimation procedure to enhance the domain of applicability for the Translog by a combination of matrix decomposition, classical non-linear estimation techniques as well as bootstrapping based resampling. The validity of the proposed procedure is exemplified by applying it to a sample of small-scale farmers. The results show that the range of theoretical consistency can be crucially enhanced for the Translog functional form by maintaining its flexibility and statistical significance. Hence, beside its empirical superiority by applying the outlined procedure the Translog can also catch up with respect to the range of functional consistency.