Studies experiences, meanings and processes in depth through words, observation and artefacts — answering how and why questions numbers cannot reach.
Research designQualitative / interpretive
What it is
Qualitative research investigates how people experience, interpret and enact the social world: how new managers make sense of authority, why a change initiative met resistance, what “work–life balance” actually means to shift workers. Data are words, images, interactions and documents rather than numbers, and the researcher — through disciplined interpretation — is the main analytical instrument.
Several traditions offer different lenses. Case study research examines one or a few bounded cases (a firm, a team, a project) in depth using multiple data sources. Grounded theory builds theory inductively through iterative coding and theoretical sampling. Phenomenology (including IPA) explores the lived experience of a phenomenon. Ethnography immerses the researcher in a setting over time. Narrative inquiry treats people's stories as the data. The tradition you choose shapes sampling, data collection and analysis, so name it and follow its logic.
Qualitative research is not “easier” than quantitative work, and it is not merely a warm-up act for surveys. Done rigorously — with systematic coding, transparent procedures and reflexivity about the researcher's own influence — it produces contextual explanation and theory that no questionnaire can. Its findings claim transferability to similar contexts, not statistical generalization to populations.
✓ When to use
The question is about meaning, experience or process — how and why, not how much
Context is inseparable from the phenomenon (organizational culture, sensemaking during crisis)
The topic is sensitive or complex, requiring trust and probing that fixed questionnaires cannot achieve
Existing theory is thin or misfits the context, and new theory needs to be built from data
You need to understand the mechanism behind a quantitative finding — why did the intervention work here and fail there?
✗ When NOT to use
The question requires quantified prevalence, magnitude or population estimates
You intend to test hypotheses about variable relationships — that is quantitative territory
Stakeholders will only accept statistically generalizable evidence, and no complementary quantitative strand is planned
Time and skill for proper fieldwork, transcription and iterative coding are unavailable — thin qualitative work is worse than none
Access to participants for sustained, in-depth engagement cannot be secured
Typical research questions
RQ — How do first-time team leaders in IT services make sense of the transition from peer to boss?
RQ — Why did a well-resourced ERP implementation meet sustained frontline resistance in one plant?
RQ — How do luxury consumers construct authenticity when brands collaborate with mass-market influencers?
RQ — What does “fairness” mean to gig workers when they evaluate platform deactivation decisions?
Key characteristics
Purpose
Understand meanings, experiences and processes in context; build theory; explain mechanisms
Typical data
Words and artefacts — interview transcripts, field notes, documents, images, recordings
Researcher control
Low over the setting; high researcher involvement — the researcher is the instrument, so reflexivity is essential
Temporal aspect
From single-interview cross-sections to months of ethnographic immersion; often iterative rather than linear
Typical sample
Small, purposive, information-rich (often 10–30 interviews; 1–5 cases; one field site) — driven by saturation, not power
Variables & measurement
Qualitative designs work with concepts and themes rather than operationalized variables. Instead of fixing measures in advance, you define sensitizing concepts from literature and let categories earn their way into the analysis from the data. “Measurement” translates into data quality: good questions, careful listening, verbatim transcription, and field notes rich enough to reconstruct context.
Rigor comes from the systematic link between raw data and claims: every theme should be traceable to coded segments, and coding decisions documented in memos. If multiple coders work on the data, negotiate a codebook and check agreement on a sample of material.
Write an interview guide of open questions and probes — then treat it as a living document
Keep reflexive memos on how your position and expectations may be shaping what you see
Define what counts as a case, an incident or a unit of analysis before coding
Plan data management early: consent for recording, transcription conventions, anonymization, QDA software (NVivo, ATLAS.ti, or manual)
Sampling approaches that fit
Purposive sampling — select participants or cases for their relevance to the question (criterion, maximum-variation, typical-case, extreme-case)
Theoretical sampling (grounded theory) — let the emerging analysis dictate who or what to sample next
Snowball sampling — for hidden or networked populations, with attention to network bias
Case selection — choose cases for theoretical reasons: revelatory, critical, polar types for comparison
Continue until saturation — new data stop generating new codes or refining categories — and report how saturation was judged
Data collection methods that fit
In-depth semi-structured interviews (commonly 45–90 minutes), recorded and transcribed verbatim
Focus groups — for norms, shared meanings and the interaction itself
Participant and non-participant observation with systematic field notes
Triangulate sources: an account that survives interviews, observation and documents is far more credible
Appropriate statistical & analytical methods
Qualitative analysis is systematic interpretation, not statistics — so this section is prose rather than Atlas links. The main families are:
If you quantify intercoder agreement on a structured codebook, Cohen's kappa (in the Statistical Methods Atlas) is the standard index — but many interpretive traditions rightly prefer dialogue and consensus over agreement statistics. Whatever the approach, show your work: code examples, theme definitions, and the audit trail from quote to claim.
Thematic analysis — the most widely used: familiarization, systematic coding, constructing themes, reviewing them against the full dataset, defining and naming, writing up (Braun & Clarke's six phases)
Grounded theory coding — open coding (fracture the data into concepts), axial coding (relate categories and subcategories), selective coding (integrate around a core category), with constant comparison and memo-writing throughout
Qualitative content analysis — systematic categorization of text, allowing category counts alongside interpretation; useful for documents and open-ended survey responses
Template analysis — start from a priori themes derived from theory and revise the template as coding proceeds
Narrative analysis — treat stories as wholes: structure, plot, characters and what the telling accomplishes
Interpretative phenomenological analysis (IPA) — idiographic, case-by-case analysis of how individuals make sense of lived experience
Cross-case synthesis — for multiple case studies: within-case analysis first, then structured comparison across cases
Transferability (≈ external validity) — thick description of context so readers can judge fit to their own settings
Dependability (≈ reliability) — an audit trail of design decisions, evolving instruments and coding; consistent, documented procedures
Confirmability (≈ objectivity) — findings grounded in data, with reflexive account of the researcher's role and assumptions
Authenticity and ethical quality — fair representation of participant voices, informed consent, and care with identifiable detail
A worked mini-example
A scholar studies how frontline bank employees experienced a chatbot deployment that automated part of their role. She conducts 24 semi-structured interviews across 6 branches (criterion sampling: 2+ years pre-deployment tenure), observes 3 branch floors for a week each, and analyses internal communications about the rollout.
Thematic analysis yields three themes: “displacement anxiety reframed as gatekeeping”, “invisible repair work” (employees quietly fixing chatbot errors), and “selective advocacy” (promoting the bot to some customers, shielding others). Member checks with eight participants refine the second theme. The study explains why satisfaction surveys showed no morale drop while informal workload rose — and generates propositions a later mixed-methods study can test.
Common pitfalls
Treating qualitative work as quotes plus opinion — no systematic coding, no audit trail
Cherry-picking vivid quotes that fit the argument while ignoring disconfirming data
Claiming statistical generalization (“most managers feel…”) from purposive samples
Skipping saturation and stopping when the calendar, not the analysis, says so
Ignoring reflexivity — pretending the researcher's presence and framing had no influence