Table of Contents
What a Case Study Actually Is
The term "case study" is widely misused in academic contexts to mean a descriptive account of an organisation, event, or individual. In the research methodology literature, however, a case study is a specific empirical design with its own logic of inquiry, quality standards, and epistemological claims. Robert Yin (2018) defines it as "an empirical inquiry that investigates a contemporary phenomenon in depth and within its real-world context, especially when the boundaries between phenomenon and context may not be clearly evident."
Two elements of this definition deserve attention. First, "contemporary" — case study research is about phenomena as they currently exist or have recently occurred, not historical reconstruction (which uses different designs). Second, "boundaries between phenomenon and context may not be clearly evident" — this is precisely the condition that makes case study design appropriate. When you cannot extract a variable from its context without destroying its meaning, a case study is the right tool.
A case study is a systematic inquiry with explicit design choices, data collection protocols, and analytical procedures. It is not a story about one organisation or person. If your "case study" has no research question, no explicit boundary-setting, and no triangulation, it is a descriptive account — not a case study in the methodological sense.
Yin's Design Framework
Yin's 2×2 typology classifies case study designs along two dimensions: the number of cases (single vs. multiple) and the number of units of analysis within each case (holistic vs. embedded). Understanding where your study sits in this typology allows you to justify your design explicitly and to understand what claims it can support.
| Design type | Cases | Units of analysis | When appropriate |
|---|---|---|---|
| Type 1 — Single holistic | 1 | 1 (the case as a whole) | The case is unique, extreme, or revelatory; or tests a well-formulated theory |
| Type 2 — Single embedded | 1 | Multiple subunits | The case contains meaningful internal variation worth analysing at multiple levels |
| Type 3 — Multiple holistic | 2+ | 1 per case | Replication logic — comparing across cases to confirm or disconfirm propositions |
| Type 4 — Multiple embedded | 2+ | Multiple per case | Maximum analytical power — cross-case and within-case comparison at multiple levels |
In multiple case designs, cases are not selected to be representative of a population (statistical sampling). They are selected because each case is expected to either produce similar results (literal replication) or produce contrasting results for predictable reasons (theoretical replication). This is an analogy to experimental replication, not survey sampling.
Single vs. Comparative Design
The choice between single and multiple case design is one of the most consequential in case study research. Each has advantages and vulnerabilities.
When a single case is justified
A single case is academically defensible under five rationales identified by Yin: (1) the critical case — it tests a clearly articulated theory; (2) the extreme or unique case — the phenomenon is so rare that it merits intensive study; (3) the representative case — it is explicitly typical and allows the reader to generalise vicariously; (4) the revelatory case — access to a previously inaccessible phenomenon; (5) the longitudinal case — the same case studied at two or more points in time.
When multiple cases are stronger
Multiple case designs generally produce more persuasive evidence because the replication logic reduces the risk that findings are artefacts of a single, idiosyncratic case. If two independent organisations operating in different sectors both exhibit the same governance failure pattern, the pattern carries more evidential weight than if documented in one organisation alone. The trade-off is depth: each case must be studied with sufficient rigour, which limits how many cases are feasible.
Bounding the Case
One of the most intellectually demanding tasks in case study design is specifying the boundary of the case. What is included in "the case" and what is its context? This is not merely a practical decision — it reflects the researcher's theoretical understanding of what constitutes the phenomenon under investigation.
Temporal boundary
Specify the time period the case covers. A case study of an organisational change cannot extend indefinitely — you must specify the episode you are studying and justify why that temporal boundary is meaningful.
Social/organisational boundary
Who and what is inside the case? If you are studying a school's response to a policy, does "the school" include governors, parents, and local authority officials — or only staff and students? Your theoretical framework should guide this.
Geographic boundary
Cases often have a spatial dimension. Be explicit about whether geographical context is part of the case (place matters to the phenomenon) or merely incidental (the location is a sampling convenience).
Theoretical boundary
The case must be studied as a case of something — a theoretical concept, a process, a mechanism. The theoretical focus determines which data are relevant. Without this, you are collecting everything and selecting nothing.
Data Sources and Triangulation
One of the distinctive strengths of case study research is its capacity to draw on multiple data sources. This is not a methodological luxury — it is a quality requirement. Yin identifies six primary sources of evidence for case studies.
Reports, minutes, memos, policies, publications. Stable and exact, but may be incomplete or deliberately constructed for an audience.
Quantitative records, organisational charts, survey data. Precise but may not match case-specific concepts.
Focused, semi-structured, or oral histories. Targeted and insightful, but subject to bias and poor recall.
Direct or participant observation. Covers real-time events, but may be time-consuming and alter the setting.
Technological items, tools, physical outputs. Insightful for cultural or technological studies; limited applicability.
Email threads, social media, intranet content. Increasingly important; raises privacy and completeness questions.
Triangulation is the practice of using multiple data sources to corroborate a finding. If an interview participant describes a practice, and that practice is corroborated by a policy document and observed directly, the finding is more credible than if based on the interview alone. Document where triangulation is achieved and note where it is absent — gaps in corroboration should be acknowledged as limitations.
Thick Description
Clifford Geertz's concept of thick description — developed in anthropology but widely adopted in qualitative case study research — distinguishes between a "thin" description of observable behaviour and a "thick" description that includes the meaning, context, and interpretive framework within which behaviour occurs.
Thick description allows the reader to assess the transferability of findings to their own context — they can judge whether the conditions that produced the finding in your case are sufficiently similar to their situation.
Analytical Procedures
Yin identifies five dominant analytical strategies for case study data. Selecting and applying one systematically is essential for rigour.
| Strategy | Logic | Best for |
|---|---|---|
| Pattern matching | Compare empirical pattern with predicted pattern from theory | Explanatory and confirmatory designs |
| Explanation building | Iteratively build a causal explanation through multiple data-analysis cycles | Exploratory designs where no prior theory exists |
| Time-series analysis | Examine how a phenomenon changes over specified time intervals | Longitudinal cases; process studies |
| Logic model | Stipulate complex chains of events connecting cause and outcome | Programme evaluations; policy implementation studies |
| Cross-case synthesis | Treat each case as a separate study, then aggregate findings across cases | Multiple case designs |
Generalisation Without Statistics
The most common objection to case study research — particularly single-case designs — is that findings cannot be generalised. This objection conflates two fundamentally different logics of generalisation.
Statistical generalisation uses a sample to make inferences about a population. This requires sufficient sample size and random selection. Case studies do not use this logic.
Analytical generalisation uses the case to make inferences about a theory. The case provides evidence that either supports, challenges, or refines a conceptual proposition. A single critical case can do this just as powerfully as a statistically representative sample — because you are generalising to theory, not to a population.
Structuring the Written Case Study
Case study write-ups follow two broad structural approaches, each suited to different purposes.
Linear-analytic structure
The most common academic structure: introduce the research question, review relevant literature, present methodology, present and analyse findings, draw conclusions. This structure is expected in dissertations and journal articles. It separates the empirical account from the analytical interpretation.
Narrative structure
Events are presented in the sequence in which they occurred, with analysis woven into the narrative. Used in business school cases and longitudinal process studies. More readable for practitioner audiences, but risks burying the analytical contribution in the story.
Validity and Quality Standards
Case study research has specific quality criteria that correspond to but differ from experimental validity concepts.
| Quality criterion | Case study standard | How to demonstrate it |
|---|---|---|
| Construct validity | Are you measuring what you claim to measure? | Multiple sources of evidence; member checking; chain of evidence |
| Internal validity | Are causal claims supported? (explanatory studies only) | Pattern matching; rival explanation testing; logic model |
| External validity | Can findings be applied beyond the case? | Explicit analytical generalisation; theory statement; use of replication logic |
| Reliability | Could another researcher following the same protocol reach the same findings? | Case study protocol; case study database; audit trail |