A client was interested in comparing the effectiveness of four different grazing methods as a tool to increase resilience to fire and resistance to invasive annual grasses on rangelands: three of the methods require the use of fences, and the other one method does not use fences. The client considered a randomized complete block design with four treatments (the four grazing methods). However, due to the physical limitations resulting from how the fences had already been set up, in each block, the treatment that does not use fence can only be assigned to be on one end of a block—while the other three treatments can be randomized to the three fenced areas. In this talk, I will discuss this experimental design problem arising from a consulting project, and explain why this will lead to a disconnected design: one treatment will be fully confounded with the factor indicting whether a plot is at one end of a block. I will discuss the consequences of using a disconnected design and a potential remedy.