Yin presents case study research methodology as a linear process which includes formulate research questions, select cases, collect data, analyze, write up. It is a clean, reassuring sequence.
In practice, however, case research is non-linear, iterative, and path-dependent. Research questions change. Frameworks evolve. The case itself can select the researcher. The most important findings are often the ones that were not anticipated.
In 2002, Dubois and Gadde published article in the Journal of Business Research, the authors discuss around the abduction approach through systematic combining. They argument that case research that aimed at developing theory should be conducted through continuous matching between theory and reality. The process also involves evolving the analytical evolving throughout the study. Rather than relying on replication logic, systematic combining emphasizes on in-depth investigation of individual cases.
Relevantly, the iterative process defined by two core activities: matching and redirection. Matching is the continuous movement between theory, data, and analysis—developing categories from data rather than forcing data into pre-existing ones. Redirection occurs when unanticipated empirical findings lead the researcher to change the study’s direction, often discovering new theoretical concepts in the process.
The framework is “tight and evolving” meaning that it articulated at the outset but continuously refined through confrontation with reality. The case itself functions as both a tool for learning and a final product. The case can also select the researcher, and boundaries are set during the research process rather than predetermined. The systematic combining stands in direct contrast to the linear, staged model advocated by Yin.
Read more through the tabs below.
Dubois, A., & Gadde, L.-E. (2014). “Systematic combining”—A decade later. Journal of Business Research, 67, 1277–1284.
Systematic combining is grounded in abduction (Peirce, 1931), a logic of inquiry that differs from both induction and deduction. Abduction is about investigating the relationship between everyday language and concepts—but more importantly, it is a logic of discovery. The researcher moves continuously between the empirical world and the model world, allowing the research issues and analytical framework to be successively reoriented as they confront reality.
Matching is the process of going back and forth between framework, data sources, and analysis. It is about achieving fit between theory and reality—but not by forcing data into pre-existing categories. Rather, categories are developed from data, and the framework is refined in parallel.
Redirection is the process by which the study changes direction in response to unanticipated empirical findings or theoretical insights. In the illustrative case, the researcher initially studied outsourcing as a process but, when the outsourcing was terminated, redirected the study toward structural analysis of activity interdependencies. This redirection led to the discovery of new theoretical concepts (e.g., Richardson, 1972) and a revised research problem.
“Going back and forth between framework, data sources, and analysis. The categories are developed from data, not forced upon them.”
“The study reorients in response to unanticipated findings. In the example, the focus shifted from process (outsourcing) to structure (activity interdependencies).”
In the empirical world, there are no natural boundaries. Any expansion of the case reveals new interdependencies and new interpretations of those already identified. Boundary setting is therefore a continuous, deliberate process—not a one-time decision made at the outset of the study.
- Longitudinal depth: Waluszewski (1989) showed that the emergence of a new technique in the pulp and paper industry could only be understood when the time boundary was extended to the 1950s—narrower time frames would have missed the factors that hindered development.
- Embedded subcases: The illustrative study used four subcases (systems) within a single case company. The purpose was not to compare subcases, but to analyze variation among them in their shared context.
- Single vs. multiple: When the research problem is directed toward analysis of interdependent variables in complex structures, the natural choice is to go deeper into one case—not to increase the number of cases.
Miles and Huberman (1994) distinguished between tight and prestructured frameworks (deductive) and loose and emergent frameworks (inductive). Systematic combining offers a third path: a tight and evolving framework. Tightness reflects the degree to which the researcher has articulated preconceptions; evolution reflects the continuous refinement of concepts in response to empirical observations and theoretical insights.
Deductive. Theory and hypotheses fixed prior to data collection. Risk of blinding the researcher to important features or misreading informants’ perceptions.
Inductive (grounded theory). Risk of indiscriminate data collection and data overload. The framework emerges from data with minimal prior structuring.
Abductive. Articulated preconceptions, but open to continuous refinement. Concepts are input and output of the research process.
Yin (1984, 2009) and Eisenhardt (1989, 1991) advocate for multiple case studies based on replication logic. The argument is that multiple cases provide more compelling evidence, greater robustness, and more generalizable theory. Eisenhardt and Graebner (2007) claim that theories from multiple cases are “better grounded, more accurate, and more generalizable.”
Cases are selected to corroborate or contradict propositions. Only relationships that replicate across most or all cases are retained. The emphasis is on generality and parsimony.
Emphasizes depth, context, and thick description. The researcher focuses on one case to reveal deep structures and novel relationships. Generalization is analytical, not statistical.
- Dyer and Wilkins (1991): “Better stories, not better constructs, to generate better theory.” The problem with multiple cases is that they strip out context and miss the deep structure.
- Weick (2007): “An argument for detail, for thoroughness, for prototypical narratives.” Conceptualization creates distance from the phenomenon; richness preserves it.
- Flyvbjerg (2006): Generalization is overvalued; “the force of example” is underestimated. Single cases can provide powerful existence proofs.
- Tsang and Kwan (1999): Replication is problematic in social sciences because observations are unique in nature—a study can never be repeated exactly.
Systematic combining aims at theory development, not theory testing. The goal is to discover new variables, new relationships, and new conceptual frameworks. This is in contrast to the positivist approach, which aims to test and verify hypotheses, often through gap-spotting and replication.
- Accuracy, generality, simplicity: Thorngate (1976) argued that no theory can score high on all three. Systematic combining prioritizes accuracy and coverage (complexity) over generality.
- Trustworthiness: Lincoln and Guba (1985) proposed credibility, transferability, dependability, and confirmability as alternatives to positivist validity and reliability.
- Methodological rigor: Case researchers must provide detailed accounts of their methodology to persuade readers. However, many case studies lack such detail (Beverland & Lindgreen, 2010; Piekkari et al., 2010).
- Reflexivity: Researchers must reflect on their process—how the case selected them, how the framework evolved, and how redirections shaped the findings.