Short answer: Case study research is an in-depth study of a bounded case, such as a person, organization, program, or decision, in its real-world context, using several sources of evidence. The main types are Stake's intrinsic, instrumental, and collective case studies and Yin's exploratory, descriptive, and explanatory case studies. Yin also separates single from multiple designs, and holistic from embedded ones.
Last updated: September 25, 2026
Case study research is the method most PhD students name in their proposal and the one reviewers push back on most. Usually the problem is not the case. It is the design: which type of case study you are running, whether you need one case or several, and what counts as evidence. This guide covers the types of case study, the three textbook approaches (Yin, Stake, and Merriam), single versus multiple designs, how many cases to include, how case study differs from ethnography, and how to analyze the data.
Key takeaways
- A case study is defined by a bounded case studied in context, not by a data collection technique. Interviews, documents, and observation are all fair game.
- There are two common typologies. Stake sorts cases by purpose (intrinsic, instrumental, collective). Yin sorts them by research aim (exploratory, descriptive, explanatory) and structure (single or multiple, holistic or embedded).
- Yin, Stake, and Merriam differ in epistemology. Yin is the most procedural and postpositivist-leaning, Stake the most interpretive, and Merriam sits between them.
- For theory-building multiple case studies, Eisenhardt (1989) suggests that between 4 and 10 cases usually works well. A single case needs a strong justification, for example that it is critical, extreme, or revelatory.
- Use the Case Study Design Decision Matrix below before you write your methods chapter.
What is case study research?
Case study research is a qualitative (and sometimes mixed-methods) approach for examining a contemporary phenomenon in depth, within its real-life context, especially when the line between phenomenon and context is blurred. Merriam and Stake describe the case as a bounded system: a single entity with clear limits, such as one hospital ward, one policy rollout, or one firm's AI adoption decision. If you cannot say what falls inside and outside your case, you do not yet have a case.
A widely cited overview in BMC Medical Research Methodology (Crowe et al., 2011) describes the approach as well suited to "how" and "why" questions about events the researcher cannot control. The same features that make case studies rich, such as context, multiple sources, and small numbers, are also why reviewers question their generalizability. Flyvbjerg's "Five Misunderstandings About Case-Study Research" (Qualitative Inquiry, 2006) is still the standard reply to the claim that you cannot generalize from a single case.
Types of case study
Most textbooks give two typologies. They answer different questions, so a single study can be described by both: for example, "an instrumental, explanatory, multiple-case study."
Stake's types: why you chose the case
- Intrinsic case study. The case is interesting in itself. You study one unusual school turnaround because the story matters, not to generalize.
- Instrumental case study. The case is a window onto a broader issue. You study one firm to understand how organizations adapt to remote work.
- Collective case study. Several instrumental cases studied jointly to understand a phenomenon across settings. This is Stake's version of what Yin calls a multiple-case design.
Yin's types: what the study is trying to do
- Exploratory. The phenomenon is new or poorly understood. The aim is to define questions and propositions for later research.
- Descriptive. The aim is a full account of the case in its context, often guided by a descriptive framework.
- Explanatory. The aim is to explain how or why something happened, using techniques such as pattern matching, explanation building, and ruling out rival explanations.
Structure: single vs multiple, holistic vs embedded
Yin adds a 2×2 of design structure. A single-case design studies one case; a multiple-case design studies two or more and relies on replication logic. A holistic design treats the case as one unit of analysis, while an embedded design analyzes sub-units inside it, such as departments within an organization. Baxter and Jack's guide for novice researchers (The Qualitative Report, 2008) has clear worked examples of each.
Yin vs Stake vs Merriam: which approach fits?
Yazan's comparison of the three methodologists (The Qualitative Report, 2015) is the most useful short read here. The table summarizes the differences that matter when you write your methods chapter.
| Dimension | Yin | Stake | Merriam |
|---|---|---|---|
| Epistemological leaning | Postpositivist; emphasizes validity and reliability | Constructivist and interpretive | Constructivist, with pragmatic procedures |
| Definition of the case | A contemporary phenomenon in its real-life context | A specific, complex, functioning bounded system | A bounded system; the unit of analysis defines the case |
| Role of prior theory | Strong: propositions guide data collection | Light: issues emerge in the field | Moderate: a theoretical framework informs the study |
| Design flexibility | Protocol planned in advance | Progressive focusing; design can change | Flexible, with a clear literature-based frame |
| Typical analysis | Pattern matching, explanation building, cross-case synthesis | Categorical aggregation and direct interpretation | Constant comparison and category construction |
| Good fit when | You want explanatory claims and a defensible audit trail | You want deep interpretive understanding of one case | You are in education or health and want a middle path |
Pick one approach and cite it consistently. Mixing Yin's replication logic with Stake's progressive focusing, without saying why, is one of the most common reviewer objections to case study proposals.
How many cases do you need?
There is no statistical answer, because case selection is theoretical, not random. For inductive theory building across cases, Eisenhardt's "Building Theories from Case Study Research" (Academy of Management Review, 1989) states that "while there is no ideal number of cases, a number between 4 and 10 cases usually works well." With fewer cases, the emerging theory is often thin. With more, the volume of data becomes hard to handle.
A single case is defensible when it is critical (it tests a well-formulated theory), extreme or unusual, revelatory (it gives access that was not possible before), or longitudinal. State which rationale applies. Otherwise, choose cases that you expect to show similar results (literal replication) or contrasting results for predictable reasons (theoretical replication). For a broader treatment of purposive selection, see our qualitative sampling strategies guide. For the number of interviews within each case, see our guide to data saturation.
Case study vs ethnography
Both study people in context and both can use observation and interviews, so they are often confused. The difference is what defines the study.
| Case study | Ethnography | |
|---|---|---|
| Defined by | A bounded case (organization, program, event, person) | A culture or social group, and the meanings its members share |
| Core evidence | Several sources: interviews, documents, archival records, observation | Prolonged fieldwork and participant observation |
| Time in the field | Varies; can be short and interview-led | Usually long and immersive |
| Typical output | Case description, cross-case findings, propositions | Thick description of cultural practices |
An ethnographic case study is possible. In that design, the case sets the boundary and ethnographic fieldwork is the main data source.
Designing a case study: the process
- Write the research question. "How" and "why" questions fit best. See our qualitative research question examples, which include a case study template.
- Bound the case. Specify who, where, and when, and state what falls outside.
- Choose the type and structure. Use the decision matrix below.
- Write a case study protocol. List the evidence sources, the interview guide, and the data handling rules. Semi-structured interviews are usually the core instrument.
- Collect evidence from several sources and keep a case study database, so another researcher can trace each claim back to its data. See our guide to triangulation.
- Analyze within each case first, then across cases. Code interviews with thematic analysis or open, axial, and selective coding, then compare cases. Our cross-case analysis guide covers the comparison step.
- Look for disconfirming evidence. A negative case analysis makes explanatory claims much harder to dismiss.
When you write up the design, our qualitative methods section template shows how to report case selection, bounding and evidence sources so a reviewer can follow each decision. If the case data will be archived or shared, see qualitative data sharing under GDPR.
Case Study Design Decision Matrix
The Qualitati editorial team built this matrix for PhD students and research teams. Answer each question in order. Your answers name your design.
| Question | If yes | If no |
|---|---|---|
| 1. Is the case itself the reason for the study? | Intrinsic (Stake) | Instrumental; go to Q2 |
| 2. Do you need comparison across settings to answer the question? | Multiple / collective; plan about 4–10 cases for theory building | Single case; name the rationale (critical, extreme, revelatory, longitudinal) |
| 3. Is the phenomenon poorly understood? | Exploratory | Go to Q4 |
| 4. Are you claiming how or why something happened? | Explanatory; plan rival explanations | Descriptive |
| 5. Does the case contain sub-units you will analyze separately? | Embedded | Holistic |
| 6. Do you plan propositions before fieldwork? | Follow Yin | Follow Stake (or Merriam for a middle path) |
Example output: "an instrumental, explanatory, embedded multiple-case study of six regional hospitals, following Yin."
Where Qualitati fits
Qualitati is a European qualitative research platform for universities and research companies, with GDPR and data privacy as central priorities. It does not design your case study or choose your cases. That judgment stays with the researcher. It helps with the evidence-heavy parts of the work:
- Interviews across cases. Run the same semi-structured guide as AI-moderated text or voice interviews at every site, or use Active Listener mode for real-time prompts and section tracking when you interview in person.
- Within- and cross-case analysis. ThemeLens maps codes to your research questions across up to 100 transcripts and synthesizes themes with participant-anchored quotes. QDA Workspace supports inductive and deductive coding and codebook generation, so you can apply one codebook across cases.
- Multilingual cases. Interviews and analysis work in 10 languages, which helps with cross-national case designs.
The free tier includes 30 monthly credits with no credit card. See pricing for current rates.
Limitations and methodological concerns
- Generalization is analytical, not statistical. Claim transfer to theory, not prevalence in a population.
- Case selection bias. Picking cases because they are accessible, not because the theory calls for them, weakens every later claim. Report how cases were chosen.
- Data volume. Several sources across several cases generate far more material than a single-method study. Budget time for the case database.
- AI-assisted coding needs review. Automated coding can speed up within-case analysis, but a researcher should check codes against the transcripts and keep an audit trail. See our AI coding audit checklist.
Who this is for, and when not to use it
This guide is for PhD students, supervisors, and research teams in business, education, health, and the social sciences who are planning a qualitative case study. Case study research is a poor fit when you need prevalence estimates or causal effects measured across a population. In that case, use a survey or experimental design. It is also a poor fit when the phenomenon cannot be bounded.
FAQ
What are the types of case study?
By purpose, Stake distinguishes intrinsic, instrumental, and collective case studies. By aim, Yin distinguishes exploratory, descriptive, and explanatory case studies. Yin also classifies designs as single or multiple, and as holistic or embedded.
What is a collective case study?
A collective case study examines several cases together to understand a shared phenomenon, population, or condition. Each case is instrumental, and the comparison across cases produces the findings.
How many cases should a multiple case study have?
For theory building, Eisenhardt (1989) suggests that between 4 and 10 cases usually works well. The right number depends on your question, your replication logic, and your resources.
What is the difference between a case study and ethnography?
A case study is defined by a bounded case and draws on several evidence sources. Ethnography is defined by the study of a culture through prolonged fieldwork and participant observation. An ethnographic case study combines the two.
Can a case study be quantitative?
Yes. Yin notes that case studies can include quantitative evidence, such as archival records or survey data, and many are mixed methods. Most published case studies in the social sciences are mainly qualitative.
Conclusion
Good case study research starts with a clearly bounded case and a named design. Choose the type from Stake's and Yin's typologies, pick a single or multiple design with a stated rationale, follow one methodologist consistently, and keep every claim traceable to its evidence. When your protocol is ready, start free with 30 credits to run your case interviews and analyze them within and across cases.
Human-review note: the Case Study Design Decision Matrix is an editorial tool from the Qualitati research team, not a validated instrument. Confirm design conventions with your supervisor and the methodology texts of your discipline.