Unit of Analysis in Case Study Research 7 Boundary Decisions

Knowledge Blog
Illustration of a unit of analysis in case study research with one room highlighted in a university building

Unit of analysis in case study research means the entity your study seeks to understand and about which you will make conclusions. It might be a programme, an organisation, a team or a particular decision process. The people you interview are sources of evidence; they do not automatically define the unit.

This distinction becomes important when a proposal says the study concerns organisational change but the evidence consists entirely of individual opinions. You need to explain how those accounts illuminate the organisational process and where the evidence stops. Otherwise, the final report can move between levels without acknowledging it.

This article works through a fictional university introducing a new doctoral-supervision model. The example shows how apparently small boundary decisions change the design, evidence and conclusions.

Contents  Clarify the unit of analysis in case study research  •  Distinguish case participants and setting  •  Decision 1 Define the phenomenon  •  Decision 2 Choose the entity being explained  •  Decision 3 Set the time boundary  •  Decision 4 Establish organisational boundaries  •  Decision 5 Decide whether embedded units are needed  •  Decision 6 Match evidence to the analytical level  •  Decision 7 Bound the conclusions  •  Diagnose three common boundary problems  •  Write a boundary statement you can use

Clarify the unit of analysis in case study research

Write a provisional sentence beginning this study will explain. For example, this study will explain how a university’s doctoral-supervision model operates during its first year of implementation. That points towards the supervision model as the case.

Now compare a different sentence: this study will explore how new doctoral students experience their relationships with supervisors. That may direct the design towards individual experiences, with the university as context. Both questions can be worthwhile, but they do not imply the same analytical unit.

Baxter and Jack’s methodological discussion is a useful starting point for considering case boundaries. The decisions below are an applied way to make those boundaries visible in a proposal and protocol.

Distinguish case participants and setting

In the supervision example, the case could be the newly introduced supervision model. Participants might include students, supervisors, programme administrators and research leaders. The setting is the university in which implementation occurs.

Keeping these labels separate prevents a common counting error. Interviewing 15 students does not necessarily create 15 cases. If all interviews contribute to understanding one bounded programme, the design may remain a single-case study.

Similarly, interviewing participants from three departments does not automatically produce a multiple-case design. Departments could be embedded units within one institutional case, contextual variations or separate cases. The choice depends on the question and analytical logic.

Decision 1 Define the phenomenon

Name what you are investigating in concrete terms. Supervision can mean meeting frequency, feedback practices, research guidance, relationship quality or institutional oversight. A study cannot examine every dimension equally within a limited period.

For the illustrative project, define the phenomenon as implementation of a structured supervision model that introduces agreed milestones and documented feedback. This gives the researcher a process to follow and identifies the activities that matter.

Do not define the phenomenon solely through a tool. A new online meeting log may be part of the model, but the study concerns the way supervision is organised and experienced. The technology becomes one element of the process.

Decision 2 Choose the entity being explained

Ask which entity will appear as the subject of your main findings. Will you say the model was implemented differently across departments, students interpreted feedback differently, or supervisors adapted their practices in different ways?

The first statement concerns programme implementation. The second concerns participant experience. The third concerns professional practice. Your design can connect these levels, but it should identify the primary focus and explain the relationships.

A practical test is to inspect the planned chapter headings. If the proposal names the programme as the unit but every findings section compares personality characteristics, the design may have drifted. Revise the question or strengthen the evidence needed to examine the programme.

Decision 3 Set the time boundary

Implementation changes over time. The first month may involve training and uncertainty; later months may involve adaptation or abandonment. State whether the case covers the launch period, the full first year or a specific cohort’s experience.

For this example, select the first academic year after introduction and include supervisors and students who participated during that period. Describe whether retrospective accounts from earlier years provide context rather than equivalent evidence.

Avoid choosing dates only because they match the dissertation timetable. Practical constraints matter, but the selected period also needs to capture the process the question asks about. If implementation has barely begun, claims about settled practice would be premature.

Decision 4 Establish organisational boundaries

Specify which programmes, departments or campuses are included. A professional doctorate and a laboratory-based PhD may use different supervision arrangements even within the same university. Combining them without examining those differences can obscure the process.

One defensible boundary might include two departments using the same institutional model but exclude programmes with formally different requirements. Explain why the selected departments provide useful variation for understanding implementation.

Access alone is an incomplete justification. A friendly department may offer easy interviews but little evidence of the process you need to examine. Record what is analytically useful about the setting as well as what makes fieldwork feasible.

The Case HQ’s guide to designing a research question can help check whether the question remains aligned with these boundaries.

Decision 5 Decide whether embedded units are needed

An embedded design examines subunits within a broader case. In this example, two departments might be examined separately before interpreting the university-wide model. Each department would need enough evidence to support meaningful within-unit analysis.

Do not create embedded units simply to make the design sound more sophisticated. If interviews are too sparse to understand either department, a comparative structure may overstate what the data can support.

Equally, avoid aggregating away a consequential difference. A central feedback policy may operate differently where supervisors have very different workloads. If that variation is central to the question, collect evidence that allows it to be examined.

A published case study protocol on nurses’ transition to practice provides an example of making design decisions visible in a specific setting. Its healthcare context is different from the university example here; use it to examine how a protocol explains its case rather than importing its design unchanged.

Decision 6 Match evidence to the analytical level

To examine an institutional supervision model, interview accounts may need to be considered alongside policy documents, milestone records and meeting practices. These sources answer different questions about intended arrangements and actual delivery.

A policy shows what should happen. A meeting record shows that an event was recorded, but may reveal little about the quality of the conversation. A student account explains experience, but may not establish how the whole programme operates.

Use an evidence map to make these limits explicit:

Intended findingUseful evidenceMain caution
How the model was designedPolicy versions and design interviewsDesign intention is not implementation
How milestones were usedRecords and participant accountsCompletion does not establish learning quality
Why experiences differedInterviews and contextual recordsPerceptions require contextual interpretation

The UK Data Service overview of qualitative data illustrates the range of materials qualitative inquiry can draw upon. Select sources because they answer the question, not simply to claim variety.

Decision 7 Bound the conclusions

If the study concerns one university, explain how its findings may inform thinking elsewhere without claiming statistical representation of all universities. Describe the context sufficiently for readers to judge relevance to their own settings.

A conclusion might identify how workload, milestone interpretation and access to informal support influenced implementation in this case. That is more defensible than stating that structured supervision models improve doctoral outcomes everywhere.

Distinguish observed patterns from proposed explanations. If students reporting more useful feedback also describe stronger progress, the study still needs to consider prior preparation, supervisory expertise and other influences. A coherent case account does not eliminate alternative explanations.

Diagnose three common boundary problems

The first problem is a moving case. Fieldwork begins with one supervision programme and expands to include all research support because participants mention it. Keep relevant context, but record any genuine expansion and its implications for ethics, evidence and feasibility.

The second problem is a mismatch between unit and sample. A proposal claims to compare departments but recruits almost everyone from one. Either address the evidence gap or revise the analytical ambition transparently.

The third problem is a change in language during reporting. Individual experiences become departmental practices, then institutional outcomes. Check the subject of each claim and ask whether the evidence supports that level.

For the wider design process, see The Case HQ’s research case-writing guide. It can help turn a clear boundary decision into a coherent account for readers.

Write a boundary statement you can use

A completed statement for the example could say: the case is the implementation of the structured supervision model in two departments during its first academic year. Students and supervisors provide accounts of its operation, while policies and milestone records provide contextual and procedural evidence. The study examines departmental variation within the institutional case and does not evaluate long-term completion effects.

Give that statement to a colleague with the research question. Ask them what would be included, what would be excluded and what kind of conclusion the study could support. Their questions will expose boundaries that remain unclear.

Researchers seeking a structured learning route can explore Designing and Conducting Case Study Research. Educators helping students make these choices can consider the paid Certified Research Methodology Instructor programme. A clear unit of analysis gives every later design decision a more stable foundation.

Tags :
case boundaries,unit of analysis
Share This :

Responses

error:
The Case HQ Online
Privacy Overview

This website uses cookies so that we can provide you with the best user experience possible. Cookie information is stored in your browser and performs functions such as recognising you when you return to our website and helping our team to understand which sections of the website you find most interesting and useful.