What Are the Main Types of Qualitative Research?

When an insights team gets an ask for a qual study, the request can mean very different studies with very different outputs: a description of what an experience feels like, a model of the process that leads to a decision, or a close look at one market where something changed. Each is a distinct qualitative research design, and the design decides what the team can claim at the end, whether the data comes from interviews, diaries or documents.

What is qualitative research?

Qualitative researchers study the why behind behavior through people’s own words and observed actions. They also examine the artifacts people produce, using small, deliberately chosen samples and open-ended questioning. Quantitative research measures how many and how much; qualitative research explains why and what it means. Strella’s guide to qualitative market research covers the definition and the main methods. It also explains how qualitative and quantitative research fit together; the focus here is the research designs teams choose between.

The main types of qualitative research at a glance

The five core designs used in practice are ethnography, phenomenology, grounded theory, case study and narrative research, the same five Creswell & Poth organize Qualitative Inquiry and Research Design around. Researchers study a different thing under each design:

• Ethnography: Researchers study a group’s shared behavior and values in its own setting, including the language its members use.
• Phenomenology: Researchers study the lived experience of a specific moment or phenomenon through the people who went through it.
• Grounded theory: Researchers study a process or interaction and explain it with a theory they build from the data rather than test against one.
• Case study: Researchers study one bounded case in depth and from several sources of evidence. The case might be a market or an account, or it might concern a launch.
• Narrative research: Researchers study the stories individuals tell about their experience over time, keeping them whole rather than cutting them into themes.

The five core qualitative research designs

Take a hypothetical streaming service trying to understand why members cancel. Each design below would answer a different business question about the same cancelling member, and each pulls a different combination of methods and sample size.

Ethnography

Ethnography studies a group in its own context, through observation and conversation over time, to describe the patterns of behavior and meaning the group shares. Researchers use it to answer how people live with a product: how a household chooses what to watch on a weeknight and which services get opened or scrolled past. They also examine what the service competes with in that living room.

For the streaming service, an ethnographic study would look at households rather than individual members. Researchers would watch (in person or through recordings) the Tuesday-night decision, noting who holds the remote and how long the browse lasts before someone switches to another app. They would also record what household members say about price when the renewal email arrives. Cancellation would show up here as the end of a slow run of unopened sessions, which no exit survey can see.

Teams rarely run the classic months-long immersion today. The practical variants are digital ethnography, where researchers observe behavior through the devices people already use, and diary studies, where participants log each viewing decision over a week or two. Participants can submit photos or short videos; voice notes are another option, and the team follows up with interviews.

• Sample size: Teams match the sample to the size of the group being studied, with 25 to 50 interviews and observations a common range and a smaller number of households for diary work.
• When to choose it: An insights team chooses ethnography for questions about habits and context, especially behavior that happens somewhere analytics cannot see. The team also needs time for a diary phase before interviews start.

Phenomenology

Phenomenology is a qualitative research design in which researchers study what a specific experience is like for the people who went through it and report a description of its shared essence rather than a theory of its causes. The business question is what the moment of deciding to cancel feels like: the thought that triggers it and the emotion attached to the price. Researchers also ask whether cancellation lands as relief or loss.

Under this design the streaming team would interview a small, focused set of people who cancelled recently, all of whom have had the experience, and ask them to describe it in their own words: what they were doing when they decided and what they weighed. Researchers would also ask how it felt afterwards. The output would be a description that could be read aloud in a win-back meeting and recognized by anyone who has cancelled, and it would feed messaging and the tone of the cancellation flow rather than a retention model.

• Methods and sample: Researchers use in-depth interviews almost exclusively, often more than one per participant, with samples of roughly 5 to 25 people. They keep the number small because they analyze each account closely.
• When to choose it: An insights team chooses phenomenology for questions about meaning and emotion within a shared, specific experience. The deliverable is a description the brand and product teams can design against.

Grounded theory

Grounded theory is a qualitative research design in which researchers build an explanation of a process from the data rather than testing a hypothesis the team already holds. Researchers code interviews as they arrive, compare each new case against the categories so far, adjust the categories, and keep going until new interviews add nothing, the point called saturation. The output is a model, often a diagram, of how the process works and under what conditions.

For the streaming service, the question would be how members get from signing up to cancelling, and researchers would use grounded theory to build the paths. Researchers could find cancellation categories in a survey. As a hypothetical illustration, a survey of TV viewers might show that some cancellers say the cost exceeds the value, others cancel after finishing the programs they wanted, and others cancel after a price increase.

What researchers could not establish from the survey is the sequence: whether the price rise lands on a member already in a lull and how long the lull runs before the renewal email. They would also need to identify which path the ad-supported tier interrupts. Through grounded theory, researchers would describe that sequence as a set of named paths, each with the conditions that push a member down it.

Many teams already work this way without naming it, coding open-ended interviews and merging codes into themes. They go back to the transcripts when a theme looks thin.

• Sampling and stopping: Researchers choose the next interviews to test the emerging model and stop when no new properties appear. Samples commonly run 20 to 30 interviews, and the stopping rule matters more than the planned number.
• When to choose it: An insights team chooses grounded theory when the question is how a process works and no existing model explains it, and when leadership needs a model it can act on segment by segment.

Case study research

Case study research is a qualitative research design in which researchers examine one bounded case in depth, drawing on several sources of evidence, to explain what happened and why. A case is anything with a clear boundary in time and place: a market, an account, a launch, a quarter. The business question concerns what happened in a specific market or launch. A team can also bound the question to one segment.

In the streaming example, the case would be a market where cancellations spiked after a price change. The team would bound the case (that market, the three months after the increase), then gather interviews with members who cancelled and members who stayed, the support tickets and cancellation-survey comments from the period, the pricing emails themselves, and the churn data as context. Researchers would look for patterns that hold across the sources, such as a price message that read as a surprise or a bundle competitor launching the same month. They would also check for a local content gap. The output would explain that spike and nothing broader.

• Methods and sample: Researchers combine interviews and documents with evidence from observation. A single case typically runs 15 to 30 interviews; a comparative study across markets usually stays at 4 to 10 cases, because more cases dilute the depth that makes the design worth running.
• When to choose it: An insights team chooses case study research when the question concerns one specific event or market and evidence already exists in several forms. The business needs an explanation for that case rather than a general theory.

Narrative research

Narrative researchers study the stories people tell about their experience over time and keep each story whole and in sequence, instead of cutting it into themes. The business question is how an experience changed for a person from beginning to end: its arc and turning points, including the moment the relationship shifted.

For the streaming service, members would recount their history with the service from sign-up to cancellation: the show that brought them in, the months they watched nightly, the period they forgot it was there, the moment they noticed the charge. The researcher would restory each account into a timeline with the member’s own turning points. Laid side by side, a dozen such accounts would show where the arcs converge, which is why the design suits journey mapping; it supplies the sequence and the emotion a map needs and that viewing data cannot provide.

• Methods and sample: Researchers use long, lightly structured interviews, often two or three with the same person, plus whatever the person brings. That might include screenshots and viewing history, with emails providing further context. Classic narrative studies focus on one or two individuals; a collective version for journey work runs more, with published studies ranging up to about 24 people.
• When to choose it: An insights team chooses narrative research when the question is how something changed over time for people, and when the deliverable is a journey map or a set of member stories leadership needs to hear in full.

Research designs vs data collection methods

A design is the overall approach to a question and determines what the study can claim; a method is how the data gets collected. Many guides blur the two by listing interviews next to ethnography as if they were alternatives, which leads teams to pick a method first and discover afterwards that it cannot answer the question. The main qualitative research methods are:

• In-depth interviews: One-to-one conversations run from a discussion guide. A live researcher brings rapport and judgment to human-moderated interviews, which suits sensitive topics and executive conversations. Human moderation also suits long narrative sessions.
• AI-moderated interviews: A newer method that sits between surveys and in-depth interviews, giving interview depth at survey speed. An AI moderator holds a voice-to-voice conversation with real people, follows the researcher’s discussion guide and probes on each answer, and many interviews run in parallel. That suits concept, messaging, pricing and cross-market work needing 50 or 100 conversations fast.
• Focus groups: Moderated group discussions, used when the interaction between participants is itself the data, such as reactions to a new tier or a trailer.
• Observation: Watching behavior in context, in person or through recordings, with or without the researcher taking part.
• Diary studies: Participants logging experiences as they happen over days or weeks, which captures habits and frequency without relying on recall.
• Document and artifact analysis: Support tickets, cancellation comments, app reviews, emails and screenshots read as evidence in their own right.

Teams can use several methods within one design. In the case study of the market with the price spike, researchers would use interviews (AI-moderated for breadth across the member base, human-moderated for a handful of long-tenured members) and documents (tickets and pricing emails). They would also use observation through recordings of members working through the cancellation flow.

For ethnography, researchers combine diaries and observation, then follow up with interviews. Researchers can use the same interview method in phenomenology or grounded theory, and in narrative work as well, with a different guide and analysis each time. By choosing the design first, the team establishes which methods it needs and how many of each.

How qualitative data is analyzed

Thematic analysis is a method in which the researcher reads the transcripts, codes passages, groups codes into candidate themes, checks the themes against the full data set, names them and writes them up with verbatims. It is the standard approach, and most insights teams follow some version of those six phases whether or not they use the label. Content analysis is a method in which the researcher counts how often a code appears, which lets a team report frequency alongside meaning. Discourse analysis is a method in which the researcher studies how language is used to do something, for example how members describe the service as “not worth it” versus “too expensive”, and it suits brand and messaging questions more than process questions.

Qualitative data analysis software is the category of tool teams use to code and query transcripts. Teams use it to store codes and link them to passages, then retrieve every passage coded a certain way.

With AI-assisted synthesis, teams can go further: they receive proposed themes organized by question and objective. Each theme links back to the source interview and its supporting verbatim quotes. Teams can also count patterns across hundreds of conversations through coded open-ended answers.

The researcher still owns the interpretation. The Market Research Society’s guidance on using AI and related technologies, published in April 2025, treats AI output as a starting point that needs expert validation, human oversight and auditable records, and in practice that means checking a theme against the quotes under it before it goes into a deck. A theme a researcher can trace to a member’s own sentence is one an executive rarely argues with.

Comparison table: which qualitative design fits the question

The table puts the five designs side by side on the dimensions a team weighs when scoping a study:

How to choose the right qualitative design

The wording of the question usually points to the design before any method discussion starts. The list below works from the questions the hypothetical streaming team would be asked:

• “What is it like to…“ points to phenomenology. What is it like to decide to cancel after three years as a member?
• “How does this process work…“ points to grounded theory. How do members move from active viewing to cancellation, and what tips them?
• “What happens in this market or account…“ points to case study. What happened in the market where cancellations doubled after the price rise?
• “How do people live with…“ points to ethnography. How do households use the service on a weeknight, alongside everything else on the TV?
• “How has this changed over time for people…“ points to narrative research. How did long-tenured members’ relationship with the service change from sign-up to the day they left?

Once the wording has narrowed the field, practical constraints decide between the designs that remain:

• Timeline: Ethnography and narrative research need weeks of diaries or repeated sessions, while teams can run phenomenology or grounded theory inside a single interview wave. A case study can also fit within one wave.
• Access to participants: Phenomenological and narrative work need people who had the specific experience and will talk at length, which for cancelled members means the team’s own lapsed-customer list rather than a general panel.
• Budget: Human-moderated depth work costs a researcher’s hours per interview, so a 30-interview grounded theory study is a different commitment from a 10-interview phenomenology, and the mix of AI-moderated and human-moderated sessions changes the sum.
• Cross-market comparison: If the business needs the same question answered in five markets, the design has to hold constant across them, which favors a case study with replication or a grounded theory study run from the same guide in every market, using a method that keeps moderation consistent across languages.

If the question can be answered by one wave of interviews with people who had the experience, run phenomenology or grounded theory now. If it needs behavior seen in context or over time, choose ethnography or narrative research and budget the weeks.

Sampling and rigor

Researchers choose qualitative samples deliberately, for what each participant can tell the team, and the numbers cannot be projected to the member base. For the streaming service, that would mean recruiting members who cancelled in the last 30 days, with quotas across tenure and plan and within each market so every group is heard, instead of a random pull from the base.

The ranges in this guide come from methods texts that disagree at the edges, so the evidence on saturation is a better planning guide than any single number.

Guest, Bunce & Johnson (Field Methods, 2006) analyzed 60 in-depth interviews and found that saturation arrived within the first twelve, with the basic elements of the main themes present after six, while cautioning against assuming six to twelve will always be enough. In practice a similar segment with a narrow question reaches saturation quickly, and comparing segments or markets means reaching it within each one.

Qualitative researchers judge trustworthiness against four criteria that map onto what a leadership team asks of a qualitative deck:

• Credibility: Do the findings reflect what participants said and meant? Researchers answer this by triangulating sources and looking for cases that contradict a theme. They also take findings back to a few participants.
• Transferability: Could another market or team judge whether the findings apply to them? That requires describing who took part and in what context, in enough detail to compare.
• Dependability: Could someone retrace how the team got from transcripts to themes? Researchers keep an audit trail of guide versions and codes, with memos documenting their reasoning.
• Confirmability: Can each interpretation be traced to data rather than to the researcher’s expectations? Source-linked quotes behind every theme are the practical test.

On Strella, participant quality controls include screener fact-checking and post-call checks on completion, response quality, consistency, location, IP and timing, with human review; the team describes those checks in how Strella checks participant quality.

Guide quality belongs under rigor too. An AI moderator asks the same guide of every participant, so leading wording in the guide reaches every interview; ten tips for unbiased interview guides covers stripping it out before a study runs.

Common mistakes

Five mistakes recur in qualitative briefs, and each traces back to a decision made before the first interview:

• Confusing a method with a design: A team that writes “run 20 interviews” in its brief has chosen a method and left the design open. No one can say in advance whether researchers will produce a description or a model. Researchers might instead explain one bounded case.
• Picking a design that does not match the question: Researchers cannot identify the cancellation paths a retention lead needs through phenomenology, and they commit to slow, heavy grounded theory work if the question was only what the moment feels like.
• Stopping before saturation: A team that stops because the calendar ran out, with themes still appearing in the last interviews, produces a model with holes researchers will have to fill in the next study.
• Treating qualitative findings as statistically representative: Researchers can identify surprise at a price email as one path to cancellation through fifteen interviews; they cannot establish what share of cancellations it explains. A team that implies a share in its deck invites the first challenge in the room.
• Analysis that cannot be traced back to what participants said: Researchers cannot defend a theme with no quote under it, or a quote assigned to the wrong interview, the moment someone asks to hear the member say it.

Getting started

The sequence below moves from the business question to a study that can run this week:

1. Write the business question in the decision’s own words (why members cancel after a price rise, and what would have kept them) and note what the decision needs: a description, a model, a case explanation or a journey.
2. Choose the design with the wording test above, and write down what the output will be before anyone drafts a guide.
3. Choose the methods the design needs, and set the mix of AI-moderated and human-moderated interviews. Decide how diaries and documents will support them.
4. Plan recruiting from the right source (lapsed members for cancellation work), with quotas by tenure and plan within each market, and set the stopping rule.
5. Plan the coding approach before the first interview, including who validates themes against quotes. Decide how findings get back to the decision.

Teams run interviews on Strella and choose AI or human moderation for each study in one platform. Hybrid moderation combines both, and teams choose a level of service from self-serve to fully managed, with a real research team behind every study. For the cancellation question, that might mean AI-moderated, voice-to-voice interviews with a hundred real, recently cancelled members across markets, run overnight from the same guide with probing on every answer, alongside human-moderated sessions with a dozen long-tenured members whose stories need a live researcher’s rapport.

Synthesis begins as interviews finish, with themes by question and objective, verbatim quotes linked to the source interview, auto-coded open-ended responses and highlight reels. Ritual’s pricing tests with AI-moderated interviews drove 24% growth in net revenue per user. The Ritual case study shows how a pricing question was scoped and run.

Teams that need the why behind cancellations in hours, not weeks, can bring their research question to Strella. See Strella in action.

FAQ

What are the main types of qualitative research?

The five core designs are ethnography, phenomenology, grounded theory, case study and narrative research, and they cover nearly every question an insights team receives. Some texts add action research or historical research.

What is the difference between a qualitative research design and a method?

Interviews, focus groups, observation, diary studies and document analysis are methods, and any of them can serve several designs, while the design sets what the study can claim. A grounded theory study and a phenomenology study can both run on interviews and still produce different outputs, because researchers choose the sampling and guide to fit the design, then analyze the interviews accordingly.

Are interviews and focus groups types of qualitative research?

Interviews and focus groups are data collection methods, and a study that names only a method has not yet chosen a design. A team that says “we will run focus groups on the new tier” still has to decide whether it wants a description of how members experience the tier or a model of how they weigh it. The team might instead want a case account of one market’s reaction; researchers shape the discussion and analysis differently for each output.

How many participants does a qualitative study need?

Enough to reach saturation within each group being compared, which in a similar segment with a narrow question often arrives within six to twelve interviews, on the Guest, Bunce & Johnson (2006) evidence.

Researchers adjust the number to the design: they keep phenomenology samples small and plan for 20 to 30 interviews in grounded theory, stopping when nothing new appears. For case studies, they gather several sources per case; when comparing markets, they reach saturation in each market separately.

What is the difference between phenomenology and grounded theory?

They differ in the question they ask, the sample they use and the output they report. Phenomenology asks “what is it like to cancel?” of a small group who all cancelled and reports a description of the shared experience. Grounded theory asks “how do members get to cancellation?” of a larger, iteratively chosen sample and reports a model of the paths and the conditions behind them.

Where do AI-moderated interviews fit in qualitative research?

They are a data collection method, used within any design that runs on interviews. On Strella the AI moderator holds a live, voice-to-voice conversation with a real participant and follows the researcher’s guide. It probes on each answer, which gives interview depth across a hundred conversations at once.

Teams pick AI, human or hybrid moderation study by study, and each is an equally available choice. Human moderation suits long narrative sessions, AI moderation suits work needing many conversations fast, and hybrid moderation combines both.