A flexible functional form can provide a second-order approximation to an arbitrary unknown function at a single point. Except in special cases, the parameters of flexible forms will vary from one point of approximation to another. I use this property to show that, in general, if an unknown function is homogeneous then i) Euler's Theorem gives rise to linear equality constraints involving both the data and a set of observation-varying flexible form parameters, ii) the common practice of imposing homogeneity on flexible functional forms is unnecessarily restrictive, and iii) it is possible to obtain estimates of the observation-varying parameters of approximating flexible forms using a Singular Value Decomposition (SVD) estimator. Two illustrations are provided: artificially-generated data is used to estimate the characteristics of a generalised linear production function; and Canadian data is used to estimate the characteristics of a consumer demand system.