The US Army frequently deals with testing and experimentation of systems, which involves binary responses (often referred to as “sensitivity testing”) where a stimulus (such as voltage or drop height) is applied as the primary input or factor. This case study is a grenade “starter slug” hotwire sensitivity test characterization that involved a number of real-world constraints such as test lab schedule and sample size. Design of experiments for generalized linear models involves certain challenges not encountered in experiments with continuous response data. For example, the data are not as information-rich as continuous measurements, the testing is often destructive, and the success of the experiment and usefulness of the response predictions are highly dependent on capturing the zone of mixed results. To deal with these difficulties, dynamic test approaches have been developed to adaptively generate data in just the right locations of the stimulus range to support effective modeling. This presentation will illustrate how ARDEC's statisticians recently adapted a binary search algorithm as the initial phase in a two-phase experiment, where the second phase used the Nonlinear Design platform in JMP to implement the Bayesian D-optimal design approach described by Gotwalt, Jones and Steinberg (Technometrics, February 2009).