03 · What You Need to Know
How to Determine Whether a Case Study Fits Your Research Question
What Makes a Case Study a Research Design?
Case study research involves an intensive investigation of a case or a small number of cases whose boundaries and context are relevant to the inquiry.
A case may be an individual, organization, program, event, community, institutional process, or another identifiable system. The precise definition varies among methodological traditions.
Yin emphasizes investigating contemporary phenomena in their real-world contexts, particularly when the boundaries between phenomenon and context are not clearly evident. Stake places greater emphasis on understanding the particularity and complexity of the case, while Merriam and Tisdell describe qualitative case study in relation to the intensive examination of a bounded system.
These traditions share an interest in contextualized investigation, but they are not methodologically identical. Yin's approach commonly emphasizes propositions, research design logic, and the development of explanations. Stake's interpretive orientation gives particular attention to the case's meanings and particularity.
Researchers should therefore specify what they mean by case study and select a methodological tradition consistent with their research purpose.
When Is Context Essential Rather Than Incidental?
Context is essential when the phenomenon being investigated may change meaning or operate differently depending on institutional conditions, relationships, history, or local practices.
Consider a university that successfully introduces generative AI into faculty assessment practices. A researcher might initially ask whether faculty members use AI. A survey could provide useful information about frequency and reported purposes of use.
However, a different question asks how the university negotiated academic integrity concerns, faculty autonomy, student expectations, and institutional policy during implementation.
Answering that question may require examining decision-making records, policy revisions, interviews, professional development activities, and actual assessment practices.
The institutional context is not merely background information. It is part of the explanation.
A case study becomes particularly appropriate when separating the phenomenon from its setting would remove information necessary to understand it.
Which Research Questions Are Well Suited to Case Studies?
Case studies often address how and why questions, but the wording of a question alone does not determine the design.
For example, the question "How did a university develop and implement its AI assessment policy?" may support a case study because it concerns an institutional process embedded in a particular setting.
Similarly, "Why did an educational technology initiative encounter resistance despite substantial institutional investment?" may require examining leadership decisions, infrastructure, faculty beliefs, and implementation history within the case.
Descriptive questions can also be appropriate when the objective is a rich account of a bounded system. Exploratory case studies may be useful when important mechanisms, relationships, or contextual conditions are not yet sufficiently understood.
Nevertheless, a question such as "What percentage of university teachers use generative AI?" is ordinarily better addressed through an appropriate survey or administrative data analysis, even if the respondents come from one university.
The deciding issue is whether an in-depth examination of the case is necessary to answer the substantive question.
What Does It Mean for a Case to Be Bounded?
A bounded case has identifiable limits that distinguish what belongs to the case from what lies outside it.
Boundaries may concern organizational membership, geographical location, participation in a program, a particular event, or a defined period.
For example, "AI adoption in higher education" is too broad to function as a clearly specified case without further definition.
By contrast, "the development and first-year implementation of a generative AI assessment policy at one university" identifies a more defensible case.
Its boundaries might include the institution, the policy-development process, the individuals involved, and the period from initial consultation through the first academic year of implementation.
Boundaries should follow the research question rather than be selected solely for convenience. Researchers must explain how the case is bounded by person, organization, place, event, or time.
Why Is the Unit of Analysis Important?
The unit of analysis identifies what the study is fundamentally examining.
If the case is a university's AI policy implementation, the university-level implementation process may be the unit of analysis. Faculty members, students, administrators, and policy documents may provide evidence about that case without each becoming a separate case.
Confusing participants with cases can produce methodological inconsistencies. A study interviewing twenty teachers does not necessarily contain twenty cases.
The researcher should determine what constitutes the unit of analysis before selecting participants and organizing data collection.
This decision influences case boundaries, sampling, analysis, and the conclusions that can reasonably be drawn.
When Is a Single-Case Study Appropriate?
A single-case study may be justified when one case offers a particularly informative opportunity to investigate the research question.
The case may be theoretically important, unusually revealing, distinctive, or representative of a situation that deserves close examination.
For example, a university that has maintained a comprehensive record of AI policy development may provide unusually rich evidence about institutional decision-making.
However, selecting one institution merely because it is accessible does not establish the methodological value of a single-case design.
The researcher should explain why this particular case is informative and what can reasonably be learned from it.
A single-case versus multiple-case study decision should therefore follow the research purpose and theoretical rationale rather than a predetermined preference for more cases.
When Might a Multiple-Case Study Be More Appropriate?
A multiple-case study may be useful when examining how a phenomenon operates across different contexts or when comparing cases contributes to developing or refining an explanation.
For example, researchers might investigate how three universities with different governance arrangements developed policies for generative AI in assessment.
Each university would be examined as a bounded case, with analysis conducted within cases and then across cases.
The purpose is not necessarily statistical representativeness. Cases may be selected because they are expected to produce similar findings under comparable conditions or contrasting findings for theoretically meaningful reasons.
Multiple cases can provide opportunities to examine whether an explanation remains plausible across contexts. However, they also increase the demands of data collection, analysis, and maintaining sufficient depth within each case.
More cases do not automatically produce a stronger case study.
Does a Case Study Have to Be Qualitative?
No. Case study research is frequently associated with qualitative methods, but it is not inherently restricted to them.
Depending on the methodological tradition and research question, case studies may incorporate interviews, observations, documents, surveys, administrative records, or quantitative indicators.
For example, a study of a university's learning analytics implementation might examine institutional documents and interviews alongside platform usage statistics and student retention records.
The quantitative data could help establish patterns of adoption or change, while qualitative evidence could explain institutional decisions and implementation processes.
However, combining qualitative and quantitative data does not automatically make a study a mixed-methods design. Mixed-methods research requires an explicit rationale for integrating qualitative and quantitative strands to address the research question.
Similarly, collecting several types of evidence does not by itself establish a coherent case-study design. The evidence must be organized around the bounded case and its analytical purpose.
Does a Case Study Require Multiple Sources of Evidence?
Using several sources of evidence is often valuable because it allows researchers to examine different aspects of the case and compare potentially inconsistent accounts.
Interviews may reveal how participants interpret events, while policy documents can establish formal decisions and observations can show how practices unfold.
Yin particularly emphasizes multiple sources of evidence and the development of a coherent chain of evidence. Other case-study traditions may place different emphasis on how sources are selected and interpreted.
There is no universal requirement that every case study use a fixed number of data sources.
The relevant question is whether the available evidence adequately supports the claims being made. Researchers should consider whether multiple sources are necessary for the intended case-study conclusions.
When Is a Case Study Preferable to a Survey?
A survey is often more appropriate when the primary objective is to estimate distributions, compare measured characteristics across a population, or examine relationships among variables using standardized responses.
A case study is more appropriate when the objective is to understand a bounded setting in depth, particularly when institutional processes and contextual relationships are central.
| Research objective |
Potentially appropriate approach |
Reason |
| Estimate the prevalence of AI use among university teachers |
Survey |
Requires population-oriented measurement and suitable sampling |
| Understand how one university negotiated and implemented an AI policy |
Case study |
Requires contextual examination of institutional processes |
| Compare reported AI attitudes across several institutions |
Comparative survey |
Focuses on measured differences across respondents |
| Explain how different governance arrangements shaped policy implementation |
Multiple-case study |
Requires contextual analysis within and across bounded institutions |
The methods can complement each other, but they should not be treated as interchangeable. A survey conducted at one university is not automatically a case study, and a case study may legitimately incorporate survey evidence.
When Is a Case Study Preferable to Phenomenology or Ethnography?
Phenomenological research typically focuses on understanding lived experience and its meaning. Ethnography commonly investigates cultural practices, social relationships, and meaning-making through sustained engagement with a social setting.
Case study research focuses on understanding a bounded case, which may include experiences and cultural practices but is not defined exclusively by either.
Suppose the research question concerns how teachers experience anxiety when using AI in assessment. A phenomenological approach may be more appropriate if the central interest is the meaning of that experience across participants.
If the question concerns how a particular faculty developed shared norms for AI-assisted assessment, an ethnographic approach may be useful when cultural practices and sustained field engagement are central.
If the question concerns how a specific institution developed, implemented, and revised an AI assessment policy, case study may be the more appropriate design.
These distinctions are not absolute. Methodological traditions overlap, and researchers should justify their choice according to the object of inquiry rather than superficial features such as the number of interviews.
Can a Case Study Support Explanatory or Causal Claims?
Case studies can contribute to explanatory research by investigating processes, sequences, contextual conditions, and evidence relevant to proposed mechanisms.
For example, researchers may examine whether leadership decisions, professional development, and faculty participation plausibly contributed to successful AI policy implementation.
Such explanations can be strengthened through chronological evidence, competing explanations, and systematic comparison of evidence against theoretical propositions.
However, observing that an intervention preceded an outcome does not establish that it caused the outcome. Case studies must still address alternative explanations and the limits of the available evidence.
Explanatory case study is not equivalent to a randomized experiment, nor does rich description alone establish causal identification.
Can Findings From a Case Study Be Generalized?
Case studies generally do not support statistical generalization to a population merely because the case was examined intensively.
However, they may contribute to theoretical or analytical generalization by examining how findings inform, refine, or challenge concepts and explanations.
For example, an institutional AI governance case study might identify conditions under which faculty consultation influences policy implementation. The resulting explanation may inform research in other institutions, but its applicability must be assessed in relation to contextual differences.
Researchers should distinguish theoretical relevance from claims about prevalence or population-average effects.
The question of how case studies support theoretical generalization is especially important when justifying the contribution of a single-case investigation.
When Is a Case Study the Wrong Choice?
A case study may be unsuitable when the primary research question requires population prevalence estimates, statistical representativeness, or a causal effect that the available case-level evidence cannot identify.
It may also be inappropriate when the case cannot be meaningfully bounded or when the researcher has no access to the contextual information needed for an in-depth investigation.
For example, interviewing five conveniently available teachers about general AI attitudes does not necessarily justify calling the study a case study if no bounded case is being investigated.
Likewise, comparing pretest and posttest scores from one class does not become a case study simply because only one class participated.
Researchers should choose the design that answers the question most credibly, even when that means abandoning a familiar methodological label.
