Applying design structure matrices

One of my learning goals for 2020 is to go deep into graph theory and its applications. As I was sorting through some books over the holidays I came across the book Design Structure Matrix Methods and Applications by Steven Eppinger and Tyson Browning at MIT. I picked up this book after reading the classic Design Rules, Vol. 1: The Power of Modularity by Carliss Baldwin and Kim Clark, one of my favourites. Baldwin and Clark made extensive use of Design Structure Matrices (DSM) in Design Rules. They looked compelling, but I am not sure that I really understood them when I first read that book.

DSMs are an application of graph theory, so I started my 2020 learning by going deeper into this framework. The best way to learn something is to apply it, and as I am a social learner I like to learn by doing things with other people, so we spun up a team at Ibbaka to explore applications. This work has expanded quickly over the last few weeks and I thought I would share some of our preliminary ideas here.

The DSM approach originates from work by Don Steward in the 1960s on dependencies in systems of equations. Basically, Steward wanted to know what order the equations should be solved in when one equation provided an input into other equations and when recursive relationships (feedback loops) are possible. This work has flowered over the following decades into a powerful way to represent, study and manage any system where there are independencies. DSMs are especially useful in identifying sequential, parallel and interdependent tasks.

Understanding where these relationships exist in a design is critical, whether the design is for a process, modular system, organization or the ways skills are connected or a market is segmented. An example of a DSM follows, from an MIT course by Professor Olivier de Weck.

There are two standard formats for DSMs. The above shows inputs in rows and output in columns. Sometimes is it convenient for layout purposes to reverse this, and have the inputs in columns and the outputs in roles. Clustering the items (the metatasks above) is an important part of the DSM method. The goal is generally to cluster items that have a lot of dependencies (or in graph theory terms, sort the items to reduce the lengths of the paths between dependent items).

We got a bit carried away with our DSM research, and are building DSMs for all of our different business processes, for the content in some of our larger implementations, and we are exploring applications in skill and capability management, market segmentation and pricing. The DSM for our market segmentation, customer targeting and pricing process has more than 100 process steps. This is a bit too many for learning and experimentation, so we are also looking at something simple that we know well, Roger Martin's Cascading Choices framework, which we have specialized for pricing work.

This is normally represented using the boxes and arrows loved by consultants, and for a simple five-step process like this a graph approach works well.

One can easily modify this for specific strategic choices like pricing.

What does this look like in a DSM format? (The below is in the rows are inputs, columns are outputs format.

The 1s are not very informative. Fortunately the DSM approach encourages one to put more context into the interaction cells. For cascading choices this could look something like this.

This is easy to translate into RDF, which is important for Ibbaka as our Skill Graph and Value Graph are both represented in this way. RDF has been described as the data language for the semantic web and connected data.

So far so good, but not really very interesting. We next asked if there are other dependencies in Cascading Choices, that the graph format glosses over. Perhaps yes.

Winning aspirations are in some cases directly dependent on ones capabilities.

OK, so the insights from DSM are limited when there are only five elements. This changes rapidly with scale though. Our current focus is using DSMs to better understand our processes and to link different parts of our platform now that we have combined two companies and have both Ibbaka Market and Ibbaka Talent (the links tell the stories of why we combined the companies from the two different perspectives).

We are starting off simply, adding teams and skills to the process model in a very simple Domain Meta Model (DMM).

We are doing this work in spreadsheets at the moment while we investigate the available DSM tools (see DSMweb.org). Eventually, DSMs and DMMs may make it into our platform and be provided to our subscribers.

Well, these are my first steps towards deepening my understanding of graph theory and its applications. To get a broader historical perspective (my learning style includes returning to the origins) I am also reading Niall Ferguson's The Square and the Tower; Networks and Power from the Freemasons to Facebook. I am still early in this book, but so far it is excellent. Ever wondered who the real Illuminati were? Read this book to find out.