A method for turning expert judgments about many interrelated factors into a clear, multi-level structural model, and an interactive tool that implements it.
I built ISM Studio, a browser-based tool that implements the full ISM workflow: enter your constructs and pairwise relationships, and it computes the reachability matrix, applies transitivity, partitions the constructs into levels, and draws the resulting structural model. No coding required. The data you input stays in your browser. Nothing you enter is uploaded anywhere.
Open ISM Studio →Interpretive Structural Modeling (ISM) is a technique that transforms sophisticated and sometimes unclear articulated mental models of a complex system into a clear, hierarchically ordered structural model. Starting from a set of elements (constructs or factors) and how those elements relate to one another, ISM derives a multi-level directed graph that shows which factors are foundational drivers and which are higher-level outcomes.
ISM is especially useful for making sense of rich qualitative data. In my own work, I have used ISM to analyse laddering interview data, structuring many interrelated attributes, consequences, and values into a coherent hierarchy. This approach underpins "Micro-Level Mechanisms to Support Value Co-Creation for Design of Digital Services" (Journal of Service Research, 2023).
If you work mainly with qualitative interviews, ISM offers a structured way to turn what you hear into a clear picture of how concepts relate to each other. No computational background required.
Purpose: To find out the relationships between concepts from data, such as qualitative interview data, and to bring them into one coherent, hierarchical model.
Underlying assumption: Interviews do more than surface concepts (constructs); they also reveal how those concepts relate to one another. As your interviews can give you both the concepts and the links between them, these are input for ISM analysis.
The challenge ISM tackles: Unlike surveys, interviews are not "standardized." Across many interviews you will hear different concepts and different relationships. So a central question is: how do you synthesize findings across a large number of interviews into a conherent structure? ISM provides a systematic way to do this.
For each pair of elements i and j, one of four symbols records the direction of the relationship:
| Symbol | Meaning |
|---|---|
V | Relation from i to j (but not the reverse) |
A | Relation from j to i (but not the reverse) |
X | Relation in both directions (i↔j) |
O | No valid relation between the elements |
The SSIM is converted to a binary adjacency matrix, to which the identity matrix is added. Repeated Boolean multiplication of (A + I) until it stabilises yields the reachability matrix, which embeds the method's core assumption: transitivity: if factor A relates to B and B relates to C, then A relates to C. Level partitioning then compares each construct's reachability set with its antecedent set to assign it to a level, producing the layered structure that the final digraph visualises.
A defining choice in ISM is how the pairwise relationships in the SSIM are established.
In "Micro-Level Mechanisms to Support Value Co-Creation for Design of Digital Services" (Journal of Service Research, 2023), we applied this data-driven ISM to laddering interviews about digital services. The figure below shows the resulting structural model for an intelligent cyber-physical system (mining), built at a co-occurrence threshold of 7. By varying the threshold, we generated several ISM graphs. Lower thresholds yield more inclusive models, higher thresholds retain only the strongest relationships.
ISM organises the mechanisms into hierarchical levels (the dashed bands). Access to information sits at the foundation, driving the middle-level mechanisms — customer-orientation, context of use, and decision making / support — which in turn support the top-level outcomes: service experience, utilitarian values & goals, and reliability / credibility. The directed edges, labelled with the study's propositions (P2a, P2c, P2d, P3a, P4), show how each mechanism influences the next. The model reveals that access to information is a foundational mechanism for co-creating value — a structural insight that guides where service designers should focus.
A few honest reflections, especially if you come from a qualitative tradition and the "matrix" steps feel unfamiliar.
Coding remains central. ISM does not replace the careful analysis of your interviews. The relationship matrix that feeds the method is produced by your coding of the interview data — so thoughtful, rigorous coding is essential.
Garbage in, garbage out. The relationship matrix is the input to ISM, and the method itself is just a fixed series of computational steps. It faithfully processes whatever you give it: a well-grounded matrix yields a trustworthy model, a weak one does not. The rigour of your qualitative work determines the value of the result.
You still make the judgment calls. ISM uncovers the structure of relationships, but you decide the key choices — such as the threshold (how strong or frequent a relationship must be to count). The method supports your judgment; it does not remove it.