This paper describes our approach to developing input assumptions for use in mean-variance analysis. Appendices cover the weaknesses of other approaches, as well as taxable considerations.