We believe that the engineering of wastewater treatment systems would be improved if we could predict and manipulate the composition, that is, the diversity of such systems. This ability would complement our established ability to predict the size of treatment communities. The theoretically based design of diversity in wastewater treatment systems is held by some to be unnecessary and by others to be impossible. In this paper we point to important phenomena in wastewater treatment, such as acclimation, adaptation and functional redundancy that would benefit from the ability to confidently design diversity. We set out a rationale and a mathematical framework for a stochastic approach to the design of diversity and show that, in principle, this approach works with two important functional groups in wastewater treatment. The implications for design and theory are briefly discussed.