Parametric search has been widely used in geometric algorithms. Cole's improvement provides a way of saving a logarithmic factor in the running time over what is achievable using the standard method. Unfortunately, this improvement comes at the expense of making an already complicated algorithm even more complex; hence, this technique has been mostly of theoretical interest. In this paper, we provide an algorithm engineering framework that allows for the same asymptotic complexity to be achieved probabilistically in a way that is both simple and practical (i.e., suitable for actual implementation). The main idea of our approach is to show that a variant of quicksort, known as boxsort, can be used to drive comparisons, instead of using a sorting network, like the complicated AKS network, or an EREW parallel sorting algorithm, like the fairly intricate parallel mergesort algorithm. This results in a randomized optimization algorithm with a running time matching that of using Cole's method, with high probability, while also being practical. We show how this results in practical implementations of some geometric algorithms utilizing parametric searching and provide experimental results that prove practicality of the method.