How to do qualitative research
A practical walkthrough of the whole process — defining the question, choosing a method, sampling, gathering data, coding for themes, and interpreting without fooling yourself.
Qualitative research is how you understand the why behind what people do. Instead of counting how many, it asks what is going on, in whose words, and what it means. You use it when you do not yet know the right questions to put on a survey — when you are mapping a problem space, not measuring a known quantity.
This guide walks the real process end to end. It is method-agnostic: the same seven steps apply whether you run interviews, sit in on support calls, read open-ended survey answers, or analyze existing online discussion. Along the way it is honest about the limits — qualitative work tells you what people say and how they frame it, not how common any view is across a whole population. Treat the output as well-grounded hypotheses, not proof.
What qualitative research is good for
Reach for it early, when the territory is unfamiliar. It surfaces the vocabulary people actually use, the workarounds they have invented, the moments where something breaks, and the feelings attached to all of it. It is strong on depth, nuance, and discovering categories you did not know to look for.
It is weak on prevalence. A theme that shows up loudly in twenty conversations might be held by a tiny, vocal slice of your market. That is not a flaw — it is the trade. Qualitative research generates the hypotheses; quantitative methods (surveys, analytics, experiments) tell you how widely they hold. The two are partners, and the mistake is asking one to do the other's job.
The process, step by step
- 1
1. Define the research question and the decision it informs
Write the question in one sentence, then write the decision that depends on the answer. What frustrates first-time users during setup, so we know whether to rebuild onboarding or write better docs, is a research question tied to a choice. Without the decision attached, you collect data forever and act on none of it. A good question is open (it cannot be answered yes or no), bounded (a specific group, context, or moment), and modest (one thing at a time).
- 2
2. Choose a method and a source of data
Match the method to the question and to what you can realistically run. Interviews give depth but cost time and recruiting. Diary studies catch behavior in context. Open-ended survey responses scale but stay shallow. Analyzing existing online discussion — forum threads, communities, reviews — lets you study what people already said, unprompted, without scheduling anyone. There is no single best method; there is the one that fits your question, your timeline, and your access to people.
- 3
3. Sample — decide whose voices to include
Qualitative sampling is purposive, not random. You deliberately pick people or sources who have lived the experience you are studying, and often seek variety — different segments, edge cases, dissenters — so you are not just hearing one echo. Keep gathering until you hit saturation: the point where new sources stop adding new themes and start repeating ones you already have. Saturation is your signal that you have enough, not a fixed number set in advance.
- 4
4. Collect or gather the data
For interviews this means running sessions with a loose guide and recording them. For existing discussion it means assembling the relevant threads and conversations into one corpus you can work through. Capture context, not just quotes — who is speaking, in what situation, responding to what. Keep the raw material so you can return to it; your reading will change as themes emerge.
- 5
5. Code the data and find themes
Coding is labeling chunks of text with what they are about, then grouping those labels into themes. Read through once for familiarity, tag passages with short descriptive codes, then cluster related codes into a handful of themes that capture the patterns. This is the analytical heart of the work — see the thematic-analysis guide for a full method. Use a fixed, consistent vocabulary for recurring concepts so they actually group together instead of fragmenting into near-duplicates.
- 6
6. Interpret honestly — separate signal from a vocal minority
Now ask what the themes mean and how much weight each deserves. A theme that recurs across many independent voices is a stronger signal than one loud, repeated complaint from a single corner. Look actively for evidence against your favorite interpretation. Note who is missing from your data — the silent majority who never posted or never agreed to talk — because their absence shapes what you can honestly conclude.
- 7
7. Decide and document
Tie the findings back to the decision from step one. State what you now believe, how confident you are, and what you would still want to confirm quantitatively. Document the question, method, sources, themes, and the quotes that ground each theme, so a reader can trace any claim back to its evidence. Then make the call — or name the next study needed to make it.
Where rigor comes from
Qualitative work is not soft because it is interpretive — it is rigorous when it is transparent and self-critical. These habits are what separate a defensible finding from a hunch dressed up in quotes.
- Traceability: every theme points back to specific quotes a reader can inspect, so claims are checkable rather than asserted.
- Disconfirming evidence: you went looking for cases that break your interpretation, not just ones that confirm it.
- Naming your bias: you state going in what you expect to find, so you and your readers can watch for it shaping the read.
- Acknowledging absence: you say who is not in the data and what that leaves uncertain.
- Stable coding: the same idea gets the same label throughout, so patterns are real and not an artifact of loose wording.
- Holding findings as hypotheses: you frame conclusions as things to validate, not as settled fact.
A faster way to run steps 4 and 5
The slowest part of most qualitative studies is recruiting and scheduling. Interviews can take weeks to line up before a single insight appears. Analyzing existing online discussion removes that bottleneck: the conversations already happened, unprompted, in communities where people speak candidly to peers rather than to a researcher. You study what they said on their own terms. The customer-research-without-recruiting guide covers this no-recruit angle in depth.
This is observational and qualitative, with the same caveats as any qualitative method — the people who post are not a representative sample, and you are reading their words, not measuring the market. What it buys you is speed and unfiltered language at the collection-and-coding stages, so you reach a coded set of themes in a fraction of the usual time and then decide where to add quantitative confirmation.
How rawneed fits this process
rawneed is one practical way to run steps 4 and 5 on existing online discussion. You start with a plain-English question — the same kind you wrote in step one. It gathers relevant Reddit threads, then classifies what it finds along consistent dimensions: expressed pain, willingness to pay, sentiment, and the tools people mention. The output is a ranked report where every signal links back to the source thread, so you can read the original words behind any theme.
That ranking-with-links design maps directly onto the rigor habits above: themes stay traceable to evidence, and the consistent classification keeps recurring concepts grouped instead of scattered. It does not replace your judgment at steps 6 and 7 — you still interpret, weigh signal against vocal minority, and decide what to confirm with numbers. It is self-serve and meant to compress the gather-and-code work, not to hand you a verdict.
See the method written out
If you want the full picture of how the gather-classify-rank approach works and where its limits are, the methodology page lays out the steps and the honest caveats in one place.
Read the methodologyCommon mistakes to avoid
Confusing loud with common: a theme repeated by a few vocal voices is not evidence that most people feel it. Treat frequency in your data as a hint, not a measurement.
Stopping at quotes you like: cherry-picking the passages that fit your hope is the fastest way to a wrong, confident conclusion. The disconfirming-evidence pass is non-negotiable.
Skipping the decision: research with no decision attached becomes a document nobody reads. Tie every study to a choice before you start.
Treating hypotheses as proof: qualitative findings are where you form a belief, not where you bet the business on it. Carry the strong ones into a quantitative test.
Frequently asked questions
What is the difference between qualitative and quantitative research?
Qualitative research explores the why and how in people's own words — it is about depth, meaning, and discovering categories. Quantitative research measures how many and how much across a population — it is about prevalence and statistical confidence. Qualitative work generates hypotheses; quantitative work tests how widely they hold. Most strong research uses both in sequence.
How many interviews or sources do I need for qualitative research?
There is no fixed number. You gather until you reach saturation — the point where new sources stop surfacing new themes and start repeating ones you already have. For a narrow question that can happen quickly; for a broad one it takes more. Decide based on when the themes stop changing, not on a target set in advance.
Is qualitative research representative of the whole market?
No, and it does not try to be. Qualitative samples are chosen on purpose for relevant experience, not drawn at random, so they cannot tell you how common a view is across a population. Use qualitative work to find and frame the issues, then confirm prevalence with a survey, analytics, or an experiment.
How do I avoid bias in qualitative research?
You cannot remove interpretation, but you can make it disciplined: state your expectations up front, actively look for evidence that contradicts your favorite reading, keep every theme traceable to specific quotes, use consistent labels so patterns are real rather than wording artifacts, and name who is missing from your data. Transparency is what makes interpretive work defensible.
Can I do qualitative research without recruiting participants?
Yes. Analyzing existing online discussion — forum threads, communities, reviews — lets you study candid, unprompted conversations without scheduling anyone, which removes the slowest step. The trade is that posters are not a representative sample. It is a fast way to run the gather-and-code stages; you still interpret carefully and validate the strong findings with quantitative methods.
Keep reading
Write content about what your audience actually asks
Write about the questions your audience is actually asking.
Read →Qualitative research methods, compared honestly
Interviews, focus groups, ethnography, netnography, diary studies, open-ended surveys, content analysis — what each method is actually good for, what it costs, and where it falls short. A fair overview, not a sales pitch.
Read →Thematic Analysis: Coding Qualitative Data Into Themes
How to analyze qualitative data by coding it and grouping codes into themes — the standard six-phase process, codes versus themes, and where automated classification helps and where human judgment is still required.
Read →Exploratory Research
What exploratory research is, where it sits in the research-design family, and why analyzing Reddit discussion is a strong early-stage method for forming hypotheses.
Read →Netnography: ethnography for online communities
Ethnography moved online. Here is what netnography is, how to run a study, and why Reddit communities are the canonical modern field site.
Read →Customer research without recruiting a single participant
Recruiting is why most customer research never happens. Here’s what you can learn without scheduling a single participant.
Read →Reddit research tool: the honest guide to every type
Reddit is the most candid place on the internet, and the hardest to read at scale. This guide maps every type of Reddit research tool — from free keyword alerts to structured-report engines — so you can pick the one that fits the question you are actually asking.
Read →