The objective of this paper is to develop a methodology for customizing pul
l control systems, replacing traditional control systems such as Kanban, Co
nwip, and Base Stock. The basic principle is to start with a pull control s
ystem that connects each stage of a given production line with each precedi
ng stage, next, the optimization of the corresponding simulation model show
s which of these potential control loops are actually implemented. The resu
lt of this methodology may be one of the traditional systems, but it may al
so be one of the following three new types: (1) the total line may be decom
posed into several segments, each with its own classic control system, (2)
the total line or its segments may combine different traditional systems, o
r (3) the line may be controlled through a new type of control system. Inde
ed, these three systems are found in a simulation experiment with a set of
twelve production lines. These twelve lines are selected through a statisti
cal design with ten factors that characterize production lines.