Qualitative research techniques
Definition
Qualitative research techniques
Qualitative research techniques are methods that answer why and how people act as they do. Think interviews, observation, open questions and diary study. They use small samples, produce themes rather than numbers, and do not claim to represent a wider population.
The boundary with the sibling term matters. Quantitative research techniques answer how many and how often. Qualitative work answers why, and it cannot tell you what share of your customers feel a thing. It names the thing.
You will meet these methods under other names. A customer discovery process runs them before a product exists. A ux researcher runs them on a design that already does.
Key takeaways
- Qualitative research techniques answer why and how, and the output is a theme rather than a number.
- Samples are small and picked on purpose, so findings describe the people studied and nobody else.
- Interviews, observation, open questions and record review are the four workhorse methods.
- Ethics comes first, because consent, purpose and risk get settled before anyone is recorded.
How it works
Qualitative research techniques collect language instead of values. You recruit a small, chosen group, ask open questions, record what is said, then code the transcripts until stable themes appear. Recruiting stops when new sessions stop adding themes.
That stopping point has a name: saturation. It is a judgement, not a calculation, which is why the method needs a written protocol more than a big budget.
Sampling is purposive rather than random. You pick people for what they can tell you: the customer who cancelled, the agent on the hardest queue, the buyer who chose a rival instead of you.
A random draw would spend half your sessions on people with nothing to add. That is right when you are counting and wasteful when you are listening.
| Method | What it is good for | Typical sample |
|---|---|---|
| Depth interview | Motives, decision history, workarounds | 8 to 15 people |
| Contextual observation | What people do rather than what they say | 5 to 10 sessions |
| Focus group | Shared language and reaction to concepts | 3 to 5 groups |
| Record and ticket review | Recurring complaints already written down | 100 or more records |
| Diary study | Behaviour over weeks rather than minutes | 10 to 20 participants |
The shape of the questions separates research from a pleasant chat. A customer interview framework fixes the opening, the probes and the stop rule in advance, so two researchers running one study produce comparable transcripts.
Coding comes next. Two people read the same transcripts, tag the same passages, then argue about where they disagree — and that argument is the quality control step most teams skip.
Two habits keep the work honest. Write the analysis plan before collection starts, and keep a short memo after every session, so your thinking stays dated rather than reconstructed at the end.
The second habit is triangulation. Where a theme matters enough to spend money on, find it again in a different kind of evidence — a ticket log, a recording, a complaint file — before anyone acts on it.
Public health runs on the same logic. The United States Centers for Disease Control and Prevention published 2024 evaluation guidance saying evaluation can clarify why a program is not producing intended results, which is a qualitative question in official clothing.
Ethics sits before method, not after it. The National Research Act of 1974 created the commission that wrote the Belmont Report, setting core ethical principles for human subject research. Consent, purpose and risk are settled before recording starts.
Examples
Four settings show the methods doing real work, and none of them produces a percentage. What each one returns is a theme you can act on: a misread page, an unreported workaround, a reason people leave.
Offshore service recovery. A Manila contact centre pulls 200 escalated tickets from one quarter and codes the free text by hand. The finding is that most escalations begin with a billing page people misread. Scores never showed that; the words did.
Fixing the page cost less than the extra headcount that queue had been asking for all year.
Clinical device design. Device makers run observation sessions in hospitals before a redesign, watching nurses work an infusion pump under time pressure. Watching beats asking, because nobody volunteers their own workaround.
Ten sessions usually surface every shortcut a ward has invented — and the shortcuts are the design brief.
Leaving staff. A structured exit interview programme is qualitative research with a payroll trigger. Run it for a year and the attrition story stops being a guess. The themes are usually about a manager or a shift pattern rather than about pay.
Support language. Reading 500 chat logs tells you how customer sentiment actually gets expressed, in the customer’s words rather than your survey’s. That phrasing belongs in your help articles, because it is what people type when they search.
Related terms
- customer discovery process: the staged search for a problem worth solving, which uses these methods end to end.
- ux researcher: the role that designs the study and runs the sessions.
- customer interview framework: the fixed question structure that makes separate interviews comparable.
- voice of customer score: a numeric roll up of feedback, so it belongs on the counting side.
- customer advisory board: a standing group of customers giving recurring qualitative input.
FAQ
What do qualitative research techniques measure?
Nothing numeric. They record language and behaviour, then group both into themes that explain why something happens. The value sits in the explanation, not in any count you could run across the transcripts.
How many people do you need?
Enough to reach the point where new sessions stop adding themes. For a narrow question that is often eight to fifteen depth interviews. Add sessions until two in a row teach you nothing new, then stop recruiting.
Is a survey with open text boxes qualitative research?
Only loosely. You get words, but you cannot probe an answer, and the sample was built for counting. Use open boxes to find the questions worth asking properly, not to answer them.
When should you use quantitative research techniques instead?
When the question starts with how many, how often or how much. Qualitative work names the thing; quantitative work sizes it. Most strong teams run both, in that order.
Can qualitative findings be generalised?
No. A sample picked on purpose describes the people in it, which is why teams follow a qualitative study with a sized test before committing money.
Shortlist vetted research and back office teams in the Outsource Accelerator provider directory before you commission your next study.
Independent
