Using DSM to explore the structure of communication flows in an organisation


This page explains how DSM toolbox can be used to model, and analyse, the structure of communication flows between different people in an organisation. We focus on a research laboratory, which contains enough people and different roles to be worth modelling while remaining simple enough to describe relatively quickly. The research lab being studied is the Cambridge Engineering Design Centre (EDC). The modelling, on which this page is based, was led by Steffen Schmitz as part of his Master's thesis.

The organisation

At the time of modelling, the EDC comprised 48 researchers divided into seven themes. Each theme represents a different topic area and range in size from one researcher to 14. The researchers hold different job titles: MPhil student; PhD student; Research Associate (RA); Senior Research Associate (SRA). Professors, Lecturers etc were not included in the analysis.

The modelling process

An online questionnaire was created and sent to each researcher. Each researcher was asked to select, for the 47 other researchers, the 'Intensity' and 'Frequency' of work-related communication between themselves and that individual. Frequency could be selected from: Daily Weekly Monthly Rarely/Never. Intensity could be selected from High, Medium, or Low.

4 SRAs, presumed to have a good overview of communication in the group due to their more central roles as senior members, were asked to indepently fill out marks in a 7x7 matrix, considering the frequency and intensity of interactions across research themes in the EDC.

The models

The results were anonymised and then compiled into two 48x48 DSMs and four 7x7 DSMs. The first 48x48 DSM includes only dependencies between individuals if those people spoke at least weekly and with at least medium intensity according to the survey results. The second 48x48 DSM contains all interactions identified through the survey, regardless of frequency and intensity. Both these large matrices are clustered, putting the researchers into the themes in which they belong. Where a researcher worked across two or more themes, their 'stongest affiliation' was used to create the clustering. The affiliation did not affect the connections between researchers in the matrices.

The models were created in the DSM toolbox for CAM. The workspace containing all the models can be downloaded here:


For an example of how this kind of model can be analysed, refer to this page. The same principles apply!