4. Sampling, Case Selection and Research Access

Introduction: Why Qualitative Sampling Is a Design Decision

Qualitative research does not usually seek to represent a population through statistical generalisation. Its purpose is to develop deep, context-sensitive understanding of meanings, processes, relationships and practices. Consequently, the central sampling question is not simply, “How many participants are required?” It is, “Which participants, cases, settings or events can most productively illuminate the research problem?”

Sampling in qualitative inquiry is therefore an intellectual and strategic activity. It links the research question to the evidence from which interpretations will be developed. A poorly selected sample may produce abundant data but weak explanation; a carefully designed sample can produce theoretically significant insight from a relatively small number of cases.

Case selection and access are inseparable from sampling. A researcher may identify an analytically valuable organization but be unable to obtain meaningful access. Conversely, an accessible organization may offer limited variation or may encourage only officially approved accounts. Access is not merely an administrative step before data collection. It shapes what can be observed, whom the researcher can meet, what questions can be asked and which interpretations become possible.

This chapter examines purposive, theoretical, criterion, maximum-variation, snowball and deviant-case sampling. It also considers information power and saturation, organizational gatekeepers, participant recruitment, field relationships, ethical obligations and the use of digital tools in managing recruitment and access. The discussion is particularly relevant to doctoral research, where sampling decisions must be justified conceptually, methodologically and ethically.

Learning Objectives

By the end of the chapter, the reader should be able to:

1. Distinguish qualitative sampling from probability sampling and explain its relationship to research design.

2. Select and justify purposive, criterion, theoretical, maximum-variation, snowball and deviant-case sampling strategies.

3. Evaluate sample adequacy using information power and theoretically informed saturation.

4. Identify the influence of case boundaries, organizational context and access conditions on qualitative evidence.

5. Negotiate access with organizations, gatekeepers and participants in an ethical and professionally credible manner.

6. Design recruitment procedures that protect voluntariness, confidentiality and inclusion.

7. Recognise common sampling and access problems and develop appropriate responses.

Sampling as an Alignment of Question, Case and Evidence

In quantitative research, sampling is commonly discussed in terms of representativeness, sampling frames, random selection and sampling error. Qualitative research generally works with a different logic. The aim is often to obtain information-rich cases rather than a statistically representative sample. An information-rich case is one from which the researcher can learn substantially about the phenomenon under investigation.

This does not mean that qualitative sampling is arbitrary. On the contrary, it demands explicit reasoning. A doctoral researcher should be able to explain why a particular organization, occupational group, event or participant is relevant, what kind of knowledge the selection can generate and what knowledge it cannot support. Sampling decisions should follow from the research question, theoretical orientation, methodology, unit of analysis and proposed form of analysis.

For example, a study asking how middle managers interpret a new performance-management system requires access to managers who have experienced the system. A study asking why implementation differs across organizational units may require comparison between units with different outcomes. A study developing a process theory of organizational resilience may require sampling to evolve as emerging categories indicate which additional cases are theoretically useful.

The unit selected is not always the same as the unit analysed. An organization may be selected as a case, while teams, meetings and individual accounts provide embedded sources of evidence. Researchers should therefore distinguish among the sampling of cases, sites, participants, documents, events and episodes. Confusing these levels can lead to unjustified claims—for example, treating the views of a few employees as if they represented the entire organization.

Key Concepts in Qualitative Sampling

Population refers to the broader group or field to which the study is conceptually related. It need not be fully enumerated, as it would in a survey.

Sample is the set of cases, participants, documents, interactions or events included in the study.

Case is a bounded instance of a phenomenon, such as a firm, project, team, decision episode or institutional programme.

Sampling frame is a defined list or accessible structure from which selection is made. In qualitative research, it may be partial or constructed during fieldwork.

Information-rich case is a case offering substantial relevance, depth or variation for the research question.

Sampling unit is the entity selected, such as an organization, manager, team or document.

Unit of analysis is the entity about which the researcher makes an analytic claim.

Inclusion criteria specify who or what qualifies for participation.

Exclusion criteria identify cases or participants that fall outside the study boundary.

Saturation refers to a judgement that further data are unlikely to add meaningful conceptual insight to the analysis.

Information power concerns the extent to which a sample holds sufficient relevant information for the study’s aim.

Gatekeeper is a person or institution controlling, facilitating or influencing access to a setting or group.

Snowball sampling involves recruiting participants through referrals from existing participants or contacts.

Deviant case is a case that differs substantially from an emerging pattern and may refine, challenge or qualify an explanation.

Reflexivity is critical examination of how the researcher’s position, assumptions and relationships influence the research process.

A Portfolio of Sampling Strategies

Purposive Sampling

Purposive sampling is the broad principle of deliberately selecting cases because they are relevant to the research purpose. It is not one single technique but an overarching logic that includes several more specific approaches. A researcher studying ethical decision-making in investment firms might purposively select compliance officers, investment analysts and senior managers who have participated in consequential decisions.

The strength of purposive sampling is its direct connection to the research question. Its weakness is the possibility of selective visibility: researchers may choose participants who are easy to identify, articulate or supportive of the study. A rigorous design therefore states the basis of relevance and considers whose perspective may be absent.

Criterion Sampling

Criterion sampling selects all cases that meet a predetermined criterion or set of criteria. Criteria might include having worked in an organization for at least one year, having participated in a digital transformation project, or having experienced a specific organizational restructuring.

This strategy is useful when the phenomenon requires direct experience. It can improve coherence within a sample, but it may reduce variation if criteria are too narrow. Criteria should be substantively justified rather than chosen merely for convenience. Researchers should also ask whether self-reported eligibility can be verified appropriately and ethically.

Maximum-Variation Sampling

Maximum-variation sampling deliberately includes cases that differ along dimensions relevant to the research problem. These dimensions may include organizational size, sector, geography, ownership, role, professional background or degree of implementation success.

The purpose is not to estimate the effect of these differences statistically. Rather, variation can reveal common themes across contrasting contexts and identify how meanings or processes change. For example, studying employee responses to artificial-intelligence adoption across a public agency, a family-owned manufacturer and a technology firm may disclose both shared concerns and context-dependent interpretations.

Maximum variation requires disciplined selection. Including every possible difference can make the study unmanageable. The researcher should identify dimensions theoretically or empirically relevant to the question and explain why they matter.

Theoretical Sampling

Theoretical sampling is associated particularly with grounded theory. Sampling decisions are guided by concepts emerging from analysis rather than fixed entirely in advance. Early interviews may suggest that “managerial sponsorship” is less important than “local translation” in explaining implementation. The researcher may then seek participants or sites that can elaborate, compare or challenge these categories.

Theoretical sampling is iterative. Data collection, coding, comparison and sampling occur in a connected process. The researcher does not simply collect a large initial sample and analyse it later. Instead, emerging interpretations guide the next search for relevant evidence.

Theoretical sampling must not be confused with changing the research question opportunistically. Sampling changes should be documented, connected to analytic reasoning and approved through appropriate institutional procedures where required. A sampling memorandum should record the emerging category, the uncertainty it creates, the type of case sought and the rationale for selecting it.

Snowball and Chain-Referral Sampling

Snowball sampling begins with initial contacts who identify or refer other potential participants. It is particularly useful for dispersed, hidden or professionally connected populations, such as former executives, whistleblowers or specialist consultants.

Its main advantage is relational access. A trusted referral can reduce uncertainty and increase willingness to participate. Its major risk is network bias. Participants are likely to refer people who resemble themselves or who share the same professional circles. Researchers should therefore seek multiple entry points and monitor whose perspectives remain missing. Participants must never be pressured to disclose confidential information or recruit colleagues against their wishes.

Deviant-Case Sampling

Deviant-case sampling selects cases that do not fit an emerging pattern. A researcher may discover that most employees interpret a restructuring as threatening, while one unit experiences it as empowering. That unit is not necessarily an anomaly to be discarded. It may reveal a missing condition, such as unusually credible communication, strong team autonomy or a different history of labour relations.

Deviant cases can strengthen explanation by testing its boundaries. They should be investigated carefully rather than romanticised. A case may appear deviant because of measurement error, selective reporting or a different case boundary. The researcher must examine whether it genuinely challenges the explanation and specify the conditions under which the pattern holds.

Information Power and the Question of “Enough”

The adequacy of a qualitative sample cannot be determined by a universal numerical rule. Information power offers a useful alternative. A sample requires fewer participants when the study aim is narrow, participants are highly specific and knowledgeable, the theoretical framework is strong, interview dialogue is rich and the analysis is focused. A broader aim, heterogeneous sample, weakly specified phenomenon or superficial data requires greater information power through additional cases or richer evidence.

Saturation is another important concept, but it should be used precisely. Descriptive saturation occurs when no substantially new descriptive information is emerging. Code saturation occurs when major categories have been identified. Theoretical saturation, associated with grounded theory, concerns the development and refinement of categories and their relationships so that further data no longer add important properties or dimensions.

Saturation is not a mechanical stopping rule. Researchers can prematurely declare saturation when they repeatedly interview similar participants in the same network. A credible saturation judgement considers the diversity of the sample, the negative evidence encountered, the quality of interaction and the analytic purpose. Researchers should maintain a saturation or information-power log showing what new data contributed, where gaps remained and why recruitment ended.

Case Selection and the Architecture of Comparison

Case study research requires a clear case boundary. The boundary may be temporal, spatial, organizational or conceptual. A study of post-merger identity might focus on one integration project during its first eighteen months rather than on the entire history of the acquiring corporation.

Researchers may select a single critical case, a revelatory case, a typical case, multiple contrasting cases or embedded cases. A critical case can test whether an explanation is plausible under conditions where it should be especially visible. A revelatory case provides access to a phenomenon previously difficult to study. Multiple cases can support replication logic: the researcher expects similar findings for predictable reasons or contrasting findings for theoretically specified reasons.

Comparison should be designed rather than added after data collection. Cases need sufficient similarity for meaningful comparison and sufficient difference to illuminate variation. A case-selection matrix can specify dimensions such as ownership, size, institutional environment and outcome status. The matrix is a planning device, not a substitute for judgement.

Negotiating Organizational Access

Access often begins with a formal sponsor, such as a chief executive, human-resources director, public official or professional association. The sponsor may open doors but cannot ethically consent on behalf of employees. Organizational permission and participant consent are distinct.

An effective access proposal explains the research purpose in language meaningful to the organization without promising favourable findings. It should describe the time commitment, types of data requested, confidentiality arrangements, publication plans, anticipated disruption and procedures for handling sensitive information. Researchers should clarify what the organization will and will not receive. For example, an organization may receive a general findings briefing but not the opportunity to remove inconvenient results from a dissertation.

Gatekeepers occupy different positions. A formal gatekeeper has institutional authority; an informal gatekeeper may be a respected team member who influences participation. Both can facilitate or restrict access. Researchers should identify the gatekeeper’s interests and consider how their involvement may affect participant voluntariness. Recruitment messages distributed by a supervisor may be interpreted as an instruction, even when participation is described as voluntary.

A useful access negotiation addresses five questions:

• Who authorizes entry?
• Who introduces the researcher to participants?
• Who controls documents, meetings or physical spaces?
• How will privacy and withdrawal be protected?
• How will findings be communicated without allowing censorship or retaliation?

Access should be treated as provisional. Organizational restructures, leadership changes, legal concerns or disappointing preliminary findings can alter the relationship. Maintaining a written access agreement and updating it when the study changes protects both parties.

Participant Recruitment and Field Relationships

Recruitment materials should state the study purpose, eligibility, activities, expected duration, risks, benefits, confidentiality arrangements and voluntary nature of participation. They should avoid exaggerated claims and technical language that obscures the practical implications. Participants should know whether interviews will be recorded, who will see transcripts and whether quotations may appear in publications.

Field relationships require both rapport and analytic distance. Rapport encourages openness, but over-identification can lead the researcher to accept participants’ accounts uncritically. Researchers may also become incorporated into organizational politics. A manager may present the researcher as an evaluator, while employees may interpret the researcher as an ally or covert representative of senior leadership.

Reflexive fieldnotes should record not only what participants said but also how access was obtained, who was absent, how the researcher was introduced, what power relations were visible and how the researcher’s identity influenced interaction. These notes become part of the methodological audit trail.

Researchers should avoid promising absolute anonymity where organizational details make individuals identifiable. In small firms, a role, event or distinctive quotation may reveal a participant even after names are removed. Pseudonyms, aggregation, alteration of nonessential details and careful quotation selection may be necessary. Any alteration must preserve analytic integrity.

An Illustrative Case: Studying Algorithmic Scheduling in Retail

Background

The following is a fictional illustrative case. A doctoral researcher, Mira, is studying how retail employees experience algorithmic scheduling software that predicts customer demand and allocates staff shifts.

Situation and Problem

Mira initially proposes to interview twenty-five employees from one national retailer. The corporate human-resources department offers access but requests that all recruitment occur through store managers and that the company review quotations before publication.

This arrangement creates two problems. Manager-mediated recruitment may discourage employees from criticising scheduling practices, and quotation review could become a form of organizational censorship. In addition, a single corporate access route may produce a narrow sample of participants willing to speak within the company’s preferred framework.

Analysis

Mira defines the case as the implementation of algorithmic scheduling in three stores during a twelve-month period. She uses purposive and criterion sampling to recruit employees who have worked under the system for at least six months. To capture variation, she selects stores differing in size and union presence. Within stores, she seeks supervisors, full-time employees and part-time employees.

After initial interviews, a pattern emerges: employees describe schedule instability as harmful, but one store reports that the system has become acceptable because supervisors allow informal shift exchanges. Mira uses theoretical sampling to interview supervisors and employees in that store about local adaptation. She also selects a deviant case: a part-time employee who views algorithmic scheduling positively because it provides predictable hours compared with previous informal practices.

Mira negotiates revised access. The company may distribute a neutral invitation, but interested employees contact Mira directly. The organization can correct factual inaccuracies about its software but cannot veto quotations or findings. Interviews are conducted away from supervisors where possible, and participation is not reported to line managers.

Approach and Outcome

The final sample is smaller than the original target but more analytically diverse. Mira stops recruitment when additional interviews repeat established categories and no longer refine the relationships among algorithmic control, local discretion and employment insecurity. She documents this decision through an information-power and saturation memo.

The study shows that employee reactions are shaped not only by the technology but by local supervisory discretion and workers’ prior experiences of scheduling. The positive deviant case prevents an overly simple conclusion that algorithmic scheduling is uniformly experienced as harmful.

Learning Points

The case demonstrates that access conditions can shape findings. It also shows why sample size should not be fixed without reference to emerging analysis, why variation should be purposeful and why deviant cases are valuable. Most importantly, organizational permission does not replace individual consent, and a sponsor’s demand for quotation control may require principled renegotiation.

Challenges, Risks and Common Errors

A frequent error is convenience sampling presented as purposive sampling. Selecting whoever is easiest to reach may be defensible in exploratory circumstances, but it should not be disguised as theoretically meaningful selection. Another error is stopping when interviews become repetitive without asking whether the sample itself is too homogeneous.

Researchers may also over-rely on senior managers because they control access. This creates elite bias and can marginalise frontline, temporary or dissenting voices. A related danger is organizational “window dressing,” in which the researcher is shown exemplary sites while problematic settings remain inaccessible.

Other problems include unclear eligibility criteria, excessive recruitment through one social network, failure to document refusals, treating referrals as neutral, and changing the sample without recording the rationale. Access can also generate ethical risks: participants may fear retaliation, managers may expect commercially useful findings, and confidential documents may contain personal or legally protected information.

When access is blocked, the researcher should consider alternative sites, professional associations, public records, former employees, remote interviews or a revised case boundary. However, substitution should not quietly alter the research question. Any major change requires methodological and, where relevant, ethics approval.

Ethics, Governance and Digital Practice

Sampling and access are governed by respect for autonomy, informed consent, protection from harm, fairness and responsible handling of data. Institutional approval is important, but ethical practice continues throughout fieldwork. Consent may need to be renewed when interviews move into sensitive topics or when new forms of data are introduced.

Digital recruitment can widen reach through professional platforms, mailing lists and secure research portals. Digital tools can also assist with eligibility screening, contact tracking and audit trails. However, automated recruitment or scraping of online profiles raises concerns about privacy, platform terms, data security and unequal visibility. Artificial-intelligence tools should not infer sensitive characteristics or rank participants without transparent justification and appropriate safeguards.

Researchers should separate identifying information from research data, restrict access according to role, encrypt recordings and establish retention and deletion schedules. Automated transcription must be checked for errors, especially where accents, technical language or emotional speech are involved. Participants should be informed if external platforms process their data.

Current Developments and Best Practice

Contemporary qualitative research increasingly recognises that access is relational, uneven and politically structured. Researchers are paying greater attention to marginalized participants, intersectional differences and the voices excluded by formal organizational channels. Remote and hybrid research has expanded the range of accessible sites but may reduce contextual observation and exclude participants with limited digital access.

Strong practice includes the following:

• Begin with a sampling logic, not a preferred number of interviews.
• Distinguish case selection, participant recruitment and unit of analysis.
• Define inclusion and exclusion criteria explicitly.
• Combine strategies when the research question requires it, while explaining their sequence.
• Monitor missing perspectives and actively seek disconfirming evidence.
• Keep sampling, access and reflexive fieldwork memos.
• Treat organizational approval and individual consent as separate processes.
• Negotiate confidentiality, publication and quotation arrangements before data collection.
• Use information power and saturation as reasoned judgements rather than numerical formulas.
• Report boundaries, refusals, substitutions and access limitations honestly.

Chapter Summary

Qualitative sampling is a purposeful process of selecting cases and participants capable of generating relevant, deep and analytically useful evidence. Purposive sampling provides the general logic, while criterion, maximum-variation, theoretical, snowball and deviant-case strategies serve different research purposes. Theoretical sampling is iterative and particularly suited to grounded theory; deviant cases help test and refine emerging explanations.

Sample adequacy depends on information power, data richness, analytic focus and the diversity required by the research question. Saturation should be treated as a documented analytical judgement, not a universal numerical threshold. Case boundaries and comparison logic must be established before claims are made.

Organizational access is a negotiated relationship involving sponsors, formal and informal gatekeepers, researchers and participants. Ethical access requires voluntariness, confidentiality, protection from retaliation and clear limits on organizational influence over findings. Reflexivity, audit trails and transparent reporting enable researchers to show how sampling and access shaped the knowledge produced.

Key Terms

Purposive sampling: Deliberate selection of cases because they are relevant to the research purpose.

Criterion sampling: Selection of all or specified cases meeting established eligibility criteria.

Theoretical sampling: Iterative selection guided by concepts emerging during analysis.

Maximum-variation sampling: Selection of contrasting cases across dimensions relevant to the research question.

Snowball sampling: Recruitment through referrals from existing participants or contacts.

Deviant-case sampling: Selection of cases that depart from an emerging pattern.

Information power: The amount of relevant knowledge a sample contributes to the study aim.

Saturation: The point at which additional data add little meaningful conceptual insight.

Case boundary: The temporal, spatial, organizational or conceptual limits defining a case.

Unit of analysis: The entity about which the researcher makes analytical claims.

Sampling unit: The entity selected for inclusion in the study.

Information-rich case: A case offering substantial insight into the phenomenon studied.

Gatekeeper: A person or institution that facilitates, controls or influences access.

Organizational access: The negotiated permission and practical ability to enter a research setting.

Recruitment: The process of identifying and inviting eligible participants.

Reflexivity: Critical examination of the researcher’s influence on the research process.

Negative case: Evidence that contradicts or qualifies an emerging interpretation.

Access agreement: An explicit understanding concerning entry, responsibilities, confidentiality and publication.

Criterion: A condition used to determine eligibility for inclusion or exclusion.

Review Questions

Short Answer Questions

1. Why is qualitative sampling usually concerned with information-rich cases rather than statistical representativeness?

2. What is the difference between a sampling unit and a unit of analysis?

3. Define theoretical sampling and explain its iterative character.

4. What is meant by information power?

5. Why are organizational permission and individual participant consent separate?

Long Answer Questions

1. Compare criterion, maximum-variation and purposive sampling, explaining when each would be appropriate in doctoral business research.

2. Discuss the meaning and uses of saturation, including the difference between descriptive, code and theoretical saturation.

3. Explain how case boundaries and comparison logic influence the credibility of a qualitative case study.

4. Analyse the roles and potential biases of formal and informal gatekeepers in organizational research.

5. Discuss the ethical and methodological challenges of using snowball sampling in a workplace study.

Analytical / Application Questions

1. You are studying resistance to enterprise-resource-planning implementation across three organizations. Design a sampling strategy that incorporates maximum variation and deviant-case sampling, and justify the dimensions of variation you would use.

2. A human-resources director insists on recruiting all participants through line managers and reviewing every quotation before publication. How would you negotiate access while protecting voluntariness and research independence?

3. After twenty interviews, your data appear repetitive, but all participants come from one professional network and mostly support the change being studied. Should you stop recruitment? Explain your decision using information power, saturation and missing perspectives.

22. Multiple Choice Questions with Answers

Question 1

A researcher selects employees who have worked under a new scheduling system for at least six months. Which sampling strategy is most directly represented?

A. Criterion sampling

B. Deviant-case sampling

C. Snowball sampling

D. Theoretical sampling

Correct Answer: A. Criterion sampling

Question 2

What is the primary purpose of maximum-variation sampling?

A. To achieve statistical representativeness

B. To include all available participants

C. To examine commonalities and differences across relevant contexts

D. To recruit only participants with unusual experiences

Correct Answer: C. To examine commonalities and differences across relevant contexts

Question 3

In theoretical sampling, the next participant or case is selected mainly because it:

A. Is easiest to access

B. Can develop or challenge an emerging category

C. Matches the largest demographic group

D. Has already approved the researcher’s interpretation

Correct Answer: B. Can develop or challenge an emerging category

Question 4

Which situation best illustrates a deviant case?

A. A participant who meets the study’s basic eligibility criterion

B. A randomly selected employee from a large organization

C. A case that closely confirms every established pattern

D. A team that experiences a change positively when most teams experience it negatively

Correct Answer: D. A team that experiences a change positively when most teams experience it negatively

Question 5

Information power suggests that a smaller sample may be adequate when:

A. The research aim is narrow and participants provide highly relevant, rich information

B. The researcher has limited time regardless of data quality

C. All participants come from one convenient network

D. The study has no clearly defined analytic purpose

Correct Answer: A. The research aim is narrow and participants provide highly relevant, rich information

Question 6

Why can manager-mediated recruitment create an ethical concern?

A. Managers always provide inaccurate participant information

B. Employees may perceive participation as expected rather than voluntary

C. Managers cannot identify employees with relevant experience

D. It prevents the use of interview recordings

Correct Answer: B. Employees may perceive participation as expected rather than voluntary

Question 7

Which practice best protects research independence during organizational access negotiations?

A. Allowing the sponsor to remove any unfavourable quotation

B. Accepting only participants nominated by senior management

C. Allowing factual corrections while retaining researcher control over interpretation and publication

D. Agreeing not to disclose the research purpose to participants

Correct Answer: C. Allowing factual corrections while retaining researcher control over interpretation and publication

Question 8

A researcher repeatedly interviews people from the same professional network and finds no new themes. What is the main methodological risk?

A. Excessive theoretical sampling

B. Network bias and premature claims of saturation

C. Overuse of deviant-case sampling

D. Statistical sampling error

Correct Answer: B. Network bias and premature claims of saturation

Question 9

Which statement correctly distinguishes organizational permission from informed consent?

A. Organizational permission automatically constitutes participant consent

B. Participant consent is unnecessary when the study has ethics approval

C. Organizational permission authorizes entry, whereas individuals decide whether to participate

D. Only participants can authorize access to organizational documents

Correct Answer: C. Organizational permission authorizes entry, whereas individuals decide whether to participate

Question 10

Why should a researcher maintain a sampling or saturation memo?

A. To replace analysis with a numerical stopping rule

B. To document how emerging evidence and analytical reasoning informed recruitment decisions

C. To demonstrate that every member of the population was contacted

D. To ensure that gatekeepers can approve all interpretations

Correct Answer: B. To document how emerging evidence and analytical reasoning informed recruitment decisions