How do you get a thousand engineers to agree on the design a complex product in a timely way? How do you design such a product to be modular, so that portions of it can be more easily changed or upgraded at different rates? How do you ensure the flow of information among people and teams in a large organization without information overload? These are a few of the problems addressed by a technique called the Design Structure Matrix (DSM), a tool for managing complexity. To date, most of its applications have been in the fields of design and engineering management, but it holds promise for a myriad of applications to many kinds of systems where it is important to find and control patterns of relationships among the system’s elements.

What is a DSM? It is a square matrix, where cells on the upper-left to lower-right diagonal represent the elements of a system and off-diagonal cells represent the relationships among these elements. For example, in the first 6x6 DSM shown nearby, the six elements are numbered along the diagonal. (Their full names are often shown to the left or above the matrix, where there is more room.) The dots off the diagonal indicate the presence of a direct relationship (e.g., a dependency or a flow of information) among two elements. For example, the dot in row one column two could indicate a flow of information from element two to element one. Instead of a dot to indicate the mere presence of a relationship or link, it is also possible to use numbers or other symbols to indicate relationship intensity, probability, or other attributes, such as in the second, numerical version of the DSM showing the probability of a change in one element causing a change in another. Many additional extensions to the DSM are possible, such as the color coding of the elements according to their membership in an organization, subsystem, process, etc.

 

 

 

 

A DSM essentially shows the same information as a graph (q.v., graph theory), a node-link diagram, or many other possible representations. It is not necessarily a better representation, but it does have some particular visual advantages including conciseness and the highlighting of important patterns. As the number of elements grows, a node-link diagram begins to look like a large plate of spaghetti and meatballs, whereas as DSM maintains its ease of readability. DSM analyses such as clustering can group elements according to their membership in a module or team, such as in the third example DSM where three elements are assigned to each of two modules. Such modularity is helpful for controlling the emergence of system behaviors, modifying the system with minimal disruption, and increasing system robustness and resiliency, among other benefits. DSM also supports analyses of degrees of separation among elements, indirect and propagating effects, emergent behaviors, and social network analyses.

A community of researchers and practitioners from around the world has applied DSMs to models of products, processes, and organizations in a variety of industries and areas, including aerospace, automotive, real estate (as in the figure nearby), buildings, electronics, energy, and pharmaceuticals. Some insights into these efforts and results are available at the DSM community web site (www.DSMweb.org) and in a new book, Design Structure Matrix Methods and Applications. Here is a link to a recent MIT news article on DSM.

This DSM shows a model of the real estate development process, including the interactions among different functional groups, at Jones Lang LaSalle. (Image courtesy of MIT Press.)

This DSM shows a model of the real estate development process, including the interactions among different functional groups, at Jones Lang LaSalle. (Image courtesy of MIT Press.)