This paper extends our previous simplified approach to using group contribution methods and limited data to determine differences in solubility of sparingly soluble complex chemicals as the solvent is changed. New applications include estimating temperature dependence and the effect of adding cosolvents forming strongly nonideal aqueous mixtures and including immiscibility. The method optimally selects a 'reference solvent' from limited data, which effectively eliminates the need for pure-solute properties and minimizes the impact of their uncertainties. The technique also decreases the number of adjustable parameters to be determined by data reduction, by using an efficient experimental and mathematical regularization strategy to find their values. The approach has been established for pure and mixed solvent systems, and the resulting models have been employed in some cosolvent design problems. Though we present no new solution theory, the paper shows an especially efficient use of thermodynamic models for solvent and cosolvent selection for product formulations. Examples and discussion of applications are given. (c) 2004 Elsevier B.V. All rights reserved.