Case study methodology is a qualitative (and sometimes mixed-methods) research strategy that involves the in-depth, empirical investigation of a single bounded case — a person, group, organisation, event, programme or process — studied intensively within its real-world context. It is the method of choice when you want to answer how and why questions about a contemporary phenomenon over which you have little or no control, and where the boundaries between the phenomenon and its context are not clear-cut (Yin, 2018).
Rather than stripping a phenomenon out of its setting the way an experiment does, a case study embraces that setting. It draws on multiple sources of evidence — interviews, documents, direct observation and archival records — and converges them through triangulation to build a rich, holistic and contextually grounded account. This guide explains what a case study is, when to use it, the main types, where the data comes from, the step-by-step design process, a worked example, and how to handle its strengths and limitations with academic rigour.
What is a case study?
A case study is an empirical inquiry that investigates a contemporary phenomenon (the “case”) in depth and within its real-world context, especially when the boundaries between the phenomenon and context are not clearly evident (Yin, 2018). The defining feature is the case as a bounded system: something with limits you can specify in time, place and activity. Robert Stake (1995) frames the case study less as a methodological choice and more as a choice of what is to be studied — we study a case because we are interested in that particular case, in all its complexity.
Two authorities dominate this field, and it helps to know their emphases:
- Robert K. Yin takes a more structured, post-positivist line. He stresses a clear research design, theoretical propositions to guide data collection, multiple sources of evidence, a chain of evidence, and tests of construct, internal and external validity plus reliability.
- Robert E. Stake takes a more interpretivist, naturalistic line. He emphasises understanding the case on its own terms, the researcher’s interpretive role, and rich “thick description” that lets readers draw their own naturalistic generalisations.
Both agree on the essentials: a case study is bounded, in-depth, contextual, and built from multiple sources of evidence. It is a research strategy, not a single data-collection technique — within it you might run interviews, analyse documents, or observe behaviour. This is the most common confusion among students: they describe “a case study” as if it were a method on a par with a questionnaire, when in fact it is the overarching design inside which several methods sit. The case study tells you what you are studying and how tightly it is bounded; the methods tell you how you will gather evidence about it. Getting this distinction right early will keep your methodology chapter coherent and stop reviewers asking why your “method” section has no clear unit of analysis.
“A case study is an empirical inquiry that investigates a contemporary phenomenon (the ‘case’) in depth and within its real-world context, especially when the boundaries between phenomenon and context may not be clearly evident.” (Source: Yin, 2018)
When should you use a case study?
Yin (2018) offers a practical rule of thumb based on three conditions. A case study is the appropriate strategy when:
- Your research questions are framed as “how” or “why” questions (explanatory in nature), rather than “how many” or “how much”.
- You have little or no control over behavioural events — you cannot manipulate variables the way an experiment can.
- The focus is on a contemporary phenomenon (as opposed to a purely historical one) studied in its real-life context.
If you instead want to measure the frequency or prevalence of something across a large population, a survey is better; if you can manipulate an independent variable under controlled conditions, an experiment is stronger. The case study earns its place precisely where context is inseparable from the phenomenon — for example, why a particular school’s literacy intervention succeeded, or how a start-up navigated a crisis. Where your question concerns a purely past phenomenon with no living participants, historical research may be the more honest label.
Types of case study
Case studies are classified along two axes: by the number of cases (single vs multiple) and by the analytic purpose. Knowing which type you are running disciplines your design and your claims.
Single-case vs multiple-case designs
A single-case design studies one case in great depth. Yin argues it is justified when the case is critical (it tests a well-formulated theory), unusual/extreme, common (representative of an everyday situation), revelatory (previously inaccessible), or longitudinal (studied at two or more points in time). A multiple-case design studies several cases to allow cross-case comparison and replication logic — each case is treated like a separate experiment that either predicts similar results (literal replication) or contrasting results for predictable reasons (theoretical replication). Multiple cases generally yield more robust, generalisable findings, but cost more time and access.
Purpose-based types
The table below summarises the most commonly cited types, drawing on Yin’s and Stake’s typologies. Choose the one that matches your aim.
| Type | Core purpose | Typical question | Example |
|---|---|---|---|
| Descriptive | Describe a phenomenon in its real-life context in rich detail. | What is happening here, and what does it look like? | Documenting how one NHS ward implemented a new handover protocol. |
| Exploratory | Explore a poorly understood phenomenon to generate hypotheses for later study. | What is going on, and what questions should we ask next? | Probing how staff first react to an AI scheduling tool before designing a survey. |
| Explanatory | Explain causal links and mechanisms that are too complex for a survey or experiment. | How and why did this outcome occur? | Explaining why a corporate change programme failed despite strong funding. |
| Intrinsic (Stake) | Understand this particular case for its own sake, not to build theory. | What makes this case distinctive? | An in-depth study of one gifted child’s learning journey. |
| Instrumental (Stake) | Use the case as a means to illuminate a broader issue or refine theory. | What can this case teach us about the wider phenomenon? | Studying one inclusive classroom to understand inclusion policy generally. |
These categories are not mutually exclusive: a multiple-case study can be explanatory and instrumental. The labels matter because reviewers will expect your aim, design and claims to align — an intrinsic case study, for instance, should not over-claim generalisability.
Data sources and triangulation
A hallmark of strong case study research is the use of multiple sources of evidence, converged through triangulation so that findings rest on several independent measures rather than one. Yin (2018) identifies six common sources; the four you will most often use in a dissertation are:
- Interviews — semi-structured or in-depth interviews with key informants. Often the single richest source, capturing perceptions, explanations and meaning.
- Documents — reports, emails, minutes, policies, press releases, prospectuses. Stable, unobtrusive and reviewable; useful for corroborating and dating events.
- Direct (and participant) observation — watching behaviour and the physical setting in real time; captures what people do, not only what they say.
- Archival records — organisational charts, service statistics, budgets, survey datasets and other records, valuable for quantitative corroboration.
(The remaining two Yin sources are physical artefacts and focus-group / direct elicitation material.) For a fuller treatment of how to gather each kind of evidence, see our guide to the methods of data collection. The discipline that holds it all together is triangulation: when interviews, documents and observation point to the same conclusion, your construct validity is far stronger than any single source could provide. Storing everything in an organised case study database and maintaining a chain of evidence (so a reader can trace any finding back to its raw source) are two of Yin’s key reliability tactics.
The case study process: a step-by-step guide
Although case studies are flexible, good ones follow a disciplined sequence. Use these six steps to design and execute yours.
- Define the case and the research questions. State precisely what your case is and what it is not — its boundaries in time, place, activity and definition. Frame “how”/“why” questions that a case study can actually answer.
- Develop theoretical propositions. Derive propositions (or, for exploratory work, a clear purpose and rival explanations) from existing theory and literature. These propositions direct your attention to what should be examined and keep data collection focused.
- Choose the design. Decide single vs multiple case, and holistic vs embedded. A holistic design treats the case as one unit of analysis; an embedded design has sub-units (e.g. several departments within one organisation). This yields Yin’s four design types (single-holistic, single-embedded, multiple-holistic, multiple-embedded).
- Collect the data. Gather evidence from your multiple sources, guided by a written case study protocol (procedures, instruments, interview guide). Maintain a case study database and a chain of evidence as you go.
- Analyse the evidence. Apply a defined analytic strategy — pattern matching (compare observed patterns to predicted ones), explanation building, time-series analysis, or cross-case synthesis for multiple cases. Qualitative material is often coded using thematic analysis to surface recurring themes.
- Report the findings. Write up the case with thick, contextual description, link findings back to your propositions and the literature, state the boundaries of your claims, and discuss implications. Use tables, vignettes and a clear narrative so the reader can follow the logic from evidence to conclusion.
- Define the case & questions. The case is bounded as Northgate Components’ lean programme, on its three main production lines, over the 24 months since launch. Her question is explanatory: “How and why did the lean transformation stall after year one?” Because it is a “why” question about a contemporary event she cannot control, a case study is the right strategy.
- Develop propositions. From change-management theory she derives the proposition that change is sustained where middle managers have genuine ownership of the rollout, and stalls where it is imposed top-down. She also notes a rival explanation — that stalling is driven instead by capital under-investment — so she can test one against the other.
- Choose the design. She selects an explanatory single-case, embedded design: Northgate is the case, and its three production lines (A, B and C) are embedded sub-units that allow within-case comparison.
- Collect the data. Working from a written case study protocol, she gathers multiple sources of evidence: 14 semi-structured interviews (operators, line managers and two directors); documents — two years of board minutes, the lean rollout plan and internal memos; archival records — monthly KPI dashboards (scrap rate, throughput, downtime) and capital-spend ledgers; and three weeks of shop-floor direct observation. Everything is logged in a case study database with a chain of evidence.
- Analyse the evidence. She codes the interviews using thematic analysis and applies pattern matching, comparing the observed pattern across lines A–C against the pattern her proposition predicts. Crucially, she triangulates: interview accounts of “ownership” are checked against KPI trends and against what she observed on the floor, and the capital-spend ledgers are used to test the rival explanation.
- Report the findings. Lines A and C, where managers co-designed the rollout, sustained their year-one gains; Line B, where targets were imposed top-down, regressed to baseline — even though capital spend was comparable across all three, which lets her reject the under-investment rival explanation. The converging evidence corroborates her proposition and explains the stall.
Analytic-generalisation conclusion: The student does not claim that “most UK manufacturers stall for this reason” — that would be an illegitimate statistical leap from one case. Instead she generalises to theory: Northgate is treated as a single “experiment” that confirms and refines the middle-manager-ownership proposition, contributing a tested mechanism that future cases can replicate or challenge. The embedded within-case comparison, the triangulated sources, and the tested rival explanation are what give this single-case conclusion its analytic weight.
Strengths and limitations
Case studies trade breadth for depth, so their strengths and weaknesses are two sides of the same coin.
| Strengths | Limitations |
|---|---|
| Rich, in-depth, holistic understanding of a complex phenomenon. | Limited statistical generalisability to a wider population. |
| Preserves real-world context instead of stripping it away. | Vulnerable to researcher bias in selection and interpretation. |
| Answers “how” and “why” questions about process and mechanism. | Time- and resource-intensive; access can be hard to secure. |
| Flexible; can combine qualitative and quantitative evidence. | Large volumes of data can be difficult to analyse rigorously. |
The generalisability objection deserves a precise answer. Case studies do not aim for statistical generalisation (sample to population); they aim for analytic generalisation — generalising findings to theory, much as an experiment does (Yin, 2018). Stake similarly speaks of naturalistic generalisation, where rich description lets readers judge transferability to their own settings. Framed this way, a single, well-theorised case can legitimately advance knowledge. Some of the most influential studies in management and the social sciences are, in fact, single or small-N case studies whose value lies in the theoretical insight they generate rather than the size of their sample.
Common mistakes to avoid
- Failing to bound the case. If you cannot say where the case ends, it will sprawl. Define the unit of analysis early.
- Relying on a single source. One round of interviews is not a case study; triangulate across sources.
- Over-claiming generalisability. Claim analytic, not statistical, generalisation.
- No analytic strategy. “I read the transcripts and reported what stood out” is not analysis — name and apply a strategy (pattern matching, explanation building, cross-case synthesis).
- Letting bias run unchecked. Without explicit rigour tactics, interpretation drifts towards the conclusion you expected.
How to do a case study well: strengthening rigour
Rigour in case study research maps onto four classic tests (Yin, 2018). Build these into your design from the start:
- Construct validity — use multiple sources of evidence, establish a chain of evidence, and have key informants review your draft case report.
- Internal validity (explanatory studies) — use pattern matching, explanation building, address rival explanations, and use logic models.
- External validity — use theory in single-case designs and replication logic across multiple cases to support analytic generalisation.
- Reliability — write a case study protocol and maintain a case study database so the study could, in principle, be repeated.
Add the standard qualitative safeguards on top: keep a reflexive account of your own positionality, seek participant validation (member checking), and consider an audit trail or peer debriefing. Together these convert a vivid story into defensible research. For deciding whether a case study is even the right strategy versus a measurement-led design, our overview of quantitative vs qualitative research is a useful companion read.
Designing a case study for your dissertation?
Our academics can help you bound your case, build a rigorous protocol, and analyse your evidence with confidence.
Related methodology guides
- Triangulation in Research
- Participant Observation
