Qualitative methods: how to run expert interviews and focus groups

Learn when to use expert interviews and focus groups, how to recruit participants, write a guide, analyze the findings and turn them into a better survey.

In short: qualitative methods such as in-depth interviews and focus groups are useful when you don't yet know what reasons people have or how they view your topic, or when a survey has given you a number but no explanation. In an expert interview, you interview one person about their knowledge of a field, not about their personal experience of the issue. In a focus group, six to ten people discuss with one another, usually about something concrete such as a draft or an offer. Both methods show which reasons people give and how they explain connections. Only a survey with a suitable sample can show how widespread a reason is (Jacob et al. 2019, p. 5).

In this article, I'll answer five questions, one after the other:

• When is a conversation worth more than a survey, and what can qualitative results support?
• Expert interview or focus group: when should you use each, and whom do you invite?
• How do you build an interview guide, conduct the conversation and record it?
• How do transcripts, the written records of interviews and discussions, become findings?
• How do the conversations turn into a better questionnaire?

At the end, you'll find a checklist and short answers to frequently asked questions.

One slide in my course also says what qualitative research is not: "Not representative". Right after that, it says what it is about instead: "understanding one perspective, not all" (Marx 2026, slide 15). That sentence sums up how in-depth conversations with a few people and surveys of many people complement each other. For the conversations themselves, you need neither a studio nor a research institute: a quiet room or a video call, a recording device and a spreadsheet are enough.

When is a conversation worth more than a survey?

A conversation is worth it in three situations: before a survey, after a survey and instead of a survey.

Before the survey, you lack the questions and the answer options. A questionnaire can only ask about what you already know. If little or nothing is known about a topic, researchers therefore start with semi-structured interviews with people from selected groups. These interviews follow a prepared list of topics and help researchers understand the groups' typical perspectives (Jacob et al. 2019, p. 66).

After the survey, you have a number but no explanation. Möhring and Schlütz name both uses in one sentence: in a standardized study, semi-structured interviews or group discussions serve as a preliminary study or as a way to deepen and explain individual results afterwards (Möhring and Schlütz 2019, p. 5).

Instead of the survey, conversations are the method of choice when only a few people know something, or when the subject is too complex for closed-ended questions. That holds for processes in an organization as much as for decisions that people can't readily explain themselves. Malhotra and Birks give an example: someone who has bought an expensive sports car and is asked why will say "I got a great deal" or "My old car was falling apart". In a conversation that builds trust and probes gently, further motives can come to light (Malhotra and Birks 2007, p. 154).

In my course, research in the first two situations is described as exploratory or diagnostic: the problem is not yet precisely defined, and hypotheses are only being formed. One of the example questions there reads: "What keeps first-time buyers from buying again?" (Marx 2026, slide 13).

The example used throughout this article starts with a question like this. A language school with on-site and online courses has analyzed its annual survey: 461 of 640 respondents are satisfied or very satisfied, which is 72.0%. But the booking data show that only 372 of 1,200 learners book a follow-on course, which is 31.0%. The survey says how many are satisfied. Why satisfied people don't continue is not in any of its tables. The school decides to talk to some of them. The example and all its numbers are invented.

First, two distinctions. If you use conversations to check whether the questions in your questionnaire are understood correctly, that's a pretest, and the technique for it is called the cognitive interview. This article is about conversations in which you want to learn something about the subject itself. And whether you run a survey online, by phone or in person is yet another decision. It's covered in the article on survey methods.

Qualitative vs. quantitative research: what can the results support?

Anyone who has spent 45 minutes listening to each of eight customers faces a contradiction. On the one hand, they now know things that were not in 640 questionnaires. On the other hand, eight people are very few, and whoever reports to the team that five of eight mentioned the course times will hear that this is far too few to claim anything. Both are true. The contradiction dissolves as soon as you see that a conversation and a survey answer two different questions.

Think of a family doctor. In a consultation, she takes time for one patient. She asks follow-up questions, lets him talk, and in the end understands what is wrong with him, how long it has been going on, and how his symptoms relate to his daily life. For that purpose, one patient is enough: she has established that this condition occurs.

One thing she doesn't learn in her consultations, not even after ten patients: how many people in the city have the same condition. There are three reasons for this. Ten are few. The people who come to her are those with symptoms, not a cross-section of the city. And she doesn't examine everyone for the same thing but follows up on what the individual patient tells her. If the city's public health department wants to know how many people in the city have high blood pressure, it therefore needs a mass screening: very many people, from all across the city, and the same short measurement for everyone. The mass screening can count. It can't explain: why a particular person's blood pressure is too high is not in any series of measurements.

The two need each other. The consultations provide the clues about what a mass screening should look for. Anyone whose screening result raises a concern then has a consultation, because the number alone doesn't help them.

Conversations and surveys work in a similar way. In a conversation, you understand: which reasons people give, how they explain connections and in which words they talk about them. For that, it's enough for a reason to be described once in a way you can follow. Then you know that it exists. Whether the explanation holds is something you check against the rest of the material and against counterexamples.

Interviews cannot tell you what proportion of the population shares a reason, just as consultations cannot tell the doctor what proportion of the city's residents has a condition. The same three reasons apply.

1. They are few. With eight conversations, each single person shifts the result by 12.5 percentage points.
2. Participants are not selected to reflect the population. The people who come to a consultation are those with symptoms. The people who come to a conversation are those you have deliberately picked and who were willing. Only random selection allows you to draw conclusions about the population from the respondents and to quantify the precision of those conclusions.
3. Not everyone was asked the same question. In an open conversation, a reason comes up only if the person brings it up on their own or you happen to ask about it. So someone who doesn't mention a reason has not denied it.

A survey, the typical tool of quantitative research, counts: how many people share a reason. For that, you need exactly what a conversation lacks: many respondents, a sample that reflects the population as closely as possible, and the same question for everyone. How to arrive at such a sample is explained in the articles on sampling methods and representative samples.

Each side pays for its strength: the depth of a conversation costs generalizability, and the generalizability of a survey costs depth (Jacob et al. 2019, p. 5). That's why Malhotra and Birks explicitly list among the disadvantages of focus groups that it can be very misleading to generalize their results in the way you would generalize the results of a survey with a representative sample (2007, p. 188).

A simple rule for the report follows from this: describe what you found, and don't extrapolate to the population.

In short: conversations show which reasons exist and how the people you talked to explain them. Surveys show how often a reason occurs. A few interviews cannot give you a population percentage, because there are too few people, because they were not selected at random and because not everyone answered the same question. In return, conversations give you the list of things worth asking about in a survey.

Expert interview or focus group: when should you use each?

Both methods are conversations with a guide. They differ in whether you talk to one person or to several at once, and in the role in which you question people.

The semi-structured interview is the basic form. It does not use a fixed questionnaire, but the guiding questions are set. Their order and wording are up to the interviewer, who can adapt to the conversation in this way. That leaves room for the unexpected, and because all conversations cover the same topics, they remain comparable with one another (Möhring and Schlütz 2019, p. 5). If the main aim is to explore one person's motives, beliefs and feelings in detail, the term is in-depth interview (Malhotra and Birks 2007, p. 207).

An expert interview is a semi-structured interview with someone you interview for their knowledge of a field, rather than for their personal experience of the issue. Who counts as an expert therefore depends on your question. Meuser and Nagel show this with a teacher: if he is interviewed about pupils' difficulties in concentrating, he speaks as an expert. If the topic is the strain of teaching, the same teacher is speaking about his own experience. For Meuser and Nagel, an expert is someone who bears responsibility for designing, implementing or monitoring a solution to a problem, or who has privileged access to information about groups of people or decision processes (2002, p. 73, an English rendering of this definition is in Froschauer and Lueger 2009, p. 221).

For the language school, this means: the former learners describe their own decisions, so these are semi-structured interviews about personal experience. The program director, the course advisor at the front desk and two teachers know how transitions between courses are defined and how they work in practice, so these are four expert interviews. The difference belongs in your methods section, because it determines what you ask: you ask people with firsthand experience about what they went through, and experts about what they know about processes and about other people.

A focus group is a discussion that a trained moderator conducts in an open, natural manner with a small group. It is meant to create a setting in which the participants talk about their experiences in their own words and on their own terms (Malhotra and Birks 2007, p. 182). German textbooks mostly speak of group discussions (Möhring and Schlütz 2019, p. 5; Schumann 2019, p. 261). In this article, the two terms are used interchangeably. In a focus group, your role is to moderate the discussion, not to interview: the participants are supposed to exchange views with one another as well. That is the value of the method and at the same time its biggest drawback, because the dynamics in the group influence what is said (Möhring and Schlütz 2019, p. 5).

All page numbers in the table refer to Malhotra and Birks (2007). The typical questions are my own examples.

As a rule of thumb, choose one-on-one interviews for personal decisions, sensitive topics or individual specialists' knowledge, especially when participants have limited time. Choose a group when you need reactions to something concrete, such as a draft, an offer or a list of proposals. The language school uses both: first one-on-one interviews to understand the reasons, later two groups to discuss solutions.

If, on the other hand, specialists are supposed to arrive at a joint judgment over several written rounds, that is a method of its own. It's covered in the article on the Delphi method.

Whom do you invite, and how many?

You invite people who can contribute something to one of your open questions. This is purposive selection based on explicit criteria, rather than random selection (more on this under sampling methods). Here, that is not a flaw: the selection is not meant to mirror the whole group but to bring together cases in which differences show (Malhotra and Birks 2007, p. 160).

Experts: first the knowledge gap, then the name

A list of prominent names does not guarantee access to the knowledge you need. First write down what you don't know, and then think about who would know. A distinction by Bogner and Menz helps here. They separate three kinds of expert knowledge: technical knowledge about rules, routines and responsibilities, process knowledge about sequences of events in which the person is involved, and interpretive knowledge, that is, their points of view and explanations (Bogner and Menz 2009, p. 52).

The table also shows why rank says little about knowledge. The people at the head of an organization don't always have the knowledge that matters, and actual influence is not automatically tied to formal position (Bogner and Menz 2009, p. 55). For this reason, a study of international companies deliberately interviewed middle management: the researchers expected more knowledge of day-to-day operations there than at the top (Welch et al. 2002, cited in Littig 2009, p. 103). Also expect your list to grow. Selection is an iterative process, because after the first conversations you know better whom else you need to ask (Bogner and Menz 2009, p. 55). Often your first interviewees will name further people. Such a recommendation is useful, but consider which point of view you won't reach through this chain.

People with firsthand experience: select for relevant differences

For conversations with customers, employees or students, you look for people who differ in whatever could matter for your question. The language school selects its eight interviewees who did not book a follow-on course so that on-site and online courses, beginners and advanced learners, younger and older people, and people with fixed and with changing working hours are all represented. Where to find such people is covered in the article Where do I find respondents for my survey.

Focus group size: six to ten participants who fit together

A focus group usually has six to ten members. With fewer than six, the discussion rarely gets enough momentum. With more than ten, it gets crowded, and the group easily splits into subgroups (Malhotra and Birks 2007, p. 183). Schnell gives 8 to 12 participants for short group discussions (Schnell 2019, p. 129), and for intensive probing there are mini groups of four or five people (Malhotra and Birks 2007, p. 194).

Unlike the sample for one-on-one interviews, participants within each focus group should have relevant characteristics in common, such as age, occupation or experience with the topic. Common ground keeps side conflicts from dominating the discussion (Malhotra and Birks 2007, p. 183). You capture differences by running separate groups. In a trial group on sports offerings that Malhotra and Birks report on, the older participants hardly took the younger ones seriously. Separate groups were then held for the two age groups, and the discussions worked well (p. 191). For the selection, they recommend a short screening questionnaire that asks about familiarity with the topic, usage and a few personal details (p. 189). People who have already taken part in many focus groups don't belong in one (p. 184).

How many interviews are enough?

There is no fixed number. What matters is whether new topics are still emerging: you keep conducting interviews until the topics repeat. For focus groups, Malhotra and Birks put it this way: run further groups until the moderator can anticipate what will be said (2007, p. 192). The technical term for this point is saturation. For expert interviews, Froschauer and Lueger describe it like this: data collection and analysis continue until new material can no longer be expected to contribute further to theory development (2009, p. 225).

Two interviews with no new reasons can be a signal to stop, but they are no proof that there is no further reason. When the point is reached also depends on how different your interviewees are and how openly you ask.

In practice, what helps is therefore not a number but a stopping rule. Decide before the first interview when you will stop, for example when two interviews in a row bring nothing new. And plan time and budget for a few more interviews than you think you need. You'll only know along the way whether you need them. The textbooks this article draws on give no rule of thumb for the number of interviews. Malhotra and Birks point out that the number of in-depth interviews in a project tends to be small because of the long conversations and the high costs (2007, p. 211).

Research doesn't provide a number that holds for every project either: in one study, 92% of the keywords under which the analysis grouped similar statements had already appeared in the first twelve of 30 interviews (Guest et al. 2006, p. 66). But the authors themselves warn against turning this into a rule: they interviewed a fairly homogeneous group with the same questions for everyone, and they do not claim that their finding applies to open, exploratory interviews (pp. 63, 75, 79). The language school example, in which nothing new comes after the sixth conversation, is also invented and not a benchmark.

For focus groups, the number depends, among other things, on which types of participants you want to compare and which of them can sensibly be mixed in one group. The first group serves as a trial for the location, the guide and the composition (Malhotra and Birks 2007, pp. 191-192). The language school plans two groups of six additional participants each, one with and one without a follow-on course. Together with the eight one-on-one interviews and the four expert interviews, it talks to 24 different people.

Recruiting is one of the hardest tasks. Even after someone has agreed, call again shortly before the appointment, and compensate the participants for their time, with money, a donation or a gift (Malhotra and Birks 2007, pp. 189-190). What to keep in mind here is covered in the article on survey incentives.

How do you build a guide for interviews and focus groups?

An interview guide, or discussion guide for a focus group, lists topics, a few fully worded questions and prepared probes. It is not a questionnaire to read aloud (Malhotra and Birks 2007, p. 188). Meuser and Nagel describe its double benefit. Working on the guide makes you familiar with the topics, so that you don't come across as an unprepared interviewer. And it keeps the conversation on track without constraining it: precisely because you know the topics well, you can follow the other person. The condition is that you don't treat the guide as a fixed schedule (Meuser and Nagel 2002, pp. 77-78).

The guide moves from general questions to specific ones. Malhotra and Birks describe the first three parts (2007, p. 189). The fourth is my own suggestion:

1. Opening question. General, easy to answer and nonthreatening. It is meant to get people talking and doesn't have to matter for the analysis.
2. Transition questions. One or two questions that lead toward the topic, for instance about people's own experiences.
3. Key questions. The topics you are really after, with prepared probes.
4. Closing. Summarize and ask what is still missing.

This is what the language school's guide for the conversations with former learners looks like:

Opening: What made you want to learn the language?
Transition: Tell me about your last course with us. What was a typical week like for you during the course?
Key question 1, the end of the course: What was it like when the course ended? What were you planning to do next?
Key question 2, the decision: You didn't book another course after that. How did you reach that decision?
Key question 3, alternatives: How have you continued learning or using the language since then?
Probes for all key questions: Can you give me an example? What exactly do you mean by that? What was the deciding factor at that point?
Closing: What would have needed to be different for you to continue? Is there anything important I haven't asked about?

Three things make a guide better:

Ask about situations, not judgments. "What was it like when the course ended?" brings out a story. "How satisfied were you?" brings out one word, and you already had that from the survey.

Ask open questions, not leading ones. A question should not push anyone toward a particular answer. Lead-ins such as "I'm sure you also think that ..." are therefore to be avoided (Jacob et al. 2019, p. 136, stated there as a rule for questionnaires). In a conversation, this happens more easily than on paper, because you probe spontaneously. "Was it because of the course times?" suggests the answer. "What was the reason?" leaves it open. Probe early in the conversation so that the other person notices that precision matters to you, and more sparingly later on (Malhotra and Birks 2007, p. 187).

With experts, clarify what a statement is based on. A statement alone rarely tells you whether someone is describing a rule or offering a personal interpretation (Bogner and Menz 2009, p. 53). Three simple probes work well here: "What have you observed yourself?", "What have you been told?" and "How do you explain that?". If the course advisor says that people simply want a break after the course, that remains an important piece of information. But after the probe, you know whether her statement is based on twenty conversations at the front desk or on a guess.

How do you conduct the conversation?

Your job in the conversation is to get the other person talking and to follow what is said. You talk little.

First consider who conducts the conversation. Because an open conversation has little fixed structure, the results depend heavily on the interviewer (Malhotra and Birks 2007, p. 211). If a school employee asks former customers why they left, interviewees may soften their answers out of politeness (see social desirability bias). It's better if the questions come from someone whose work the interviewees don't have to judge. If that is not possible, say explicitly at the start that criticism is welcome.

The one-on-one interview

The following rules come from Malhotra and Birks (2007, pp. 208-209). At the start, explain the purpose, the procedure and the duration, and say what both sides get out of it. Choose a place where the other person feels comfortable. With specialists, that is usually their own office at a time that suits them. After the opening question, let the answers guide the conversation: the wording and order of your questions follow what the person says, and the guide only reminds you of what is still missing. Don't settle for a yes or a no, and note where something has remained unclear.

With experts, there is one more rule: engage with their language. Meuser and Nagel call it a cardinal error to conduct the interview in your own terms, chosen in advance, when the other person thinks in different ones (Meuser and Nagel 2002, p. 78). If the course advisor speaks of a "next-step course" and you of a "repeat booking", use her term.

The focus group

In a group, you moderate. Malhotra and Birks describe what that requires (2007, p. 194):

• Kindness with firmness. Every person should feel welcome, and nobody may take over the discussion.
• Encourage quieter participants. Address quiet participants directly.
• Probe general statements. Ask as though you are unfamiliar with the issue, and the participants will become specific.
• Watch out for unanimity. If everyone says the same thing because nobody wants to speak against the presumed group opinion, bring in a different view yourself.
• Observe and take notes. Who reacts to whom, where do people laugh, where do they fall silent?

Questions to the whole group keep the conversation among the participants going: "Who else has something that might be a bit different?" or "I see people nodding their heads; tell me about it." (p. 187). Doing both at once, moderating and taking notes, is hard. So ask a second person to take notes. The book lists the dual-moderator group as a variant of its own: one looks after the flow of the conversation, the other makes sure that all topics are covered (p. 194).

The following plan shows how 90 minutes can be divided up. It is a suggestion, not a requirement from the literature. What comes from the textbook is the order of the questions (p. 189) and the typical duration of a group of one and a half to two hours (p. 184).

If colleagues or clients want to listen in, that is possible and often instructive (Malhotra and Birks 2007, p. 186). But the participants must know that someone is listening and must have agreed (p. 200).

How to run an online focus group

Online, not everyone has to be in the same place: people in different locations can take part, and everyone sits in familiar surroundings (Malhotra and Birks 2007, pp. 201-202, said there about groups that exchange views in writing). For a group by video call, you still need a shared appointment. For such groups, here is a suggestion, not a requirement from the literature: invite four to six people rather than ten, and schedule one hour rather than two, because fewer people get a chance to speak on screen. The size roughly corresponds to the mini group (p. 194), the duration to the upper limit that the textbook gives for groups by telephone (p. 195).

How do you record conversations, and what does consent involve?

Record conversations if the participants agree, and take notes anyway.

Consent. Without the participants' consent, there is no audio or video recording and no observation. Say in advance that the conversation will be recorded and what for, and accept a no (Malhotra and Birks 2007, p. 200). Obtain consent with the invitation and once more at the start of the recording. Which details a consent form must contain and how long you may keep recordings is something to clarify with your data protection officer. This article is not legal advice. The article on data protection in online surveys gives you a starting point.

Recording. An audio recording is usually enough. Expect the recording to inhibit people at first, and expect that some important things are only said once the device is off (Malhotra and Birks 2007, p. 186). In that case, ask whether you may note the statement down and use it.

Notes. Malhotra and Birks recommend more detailed notes as soon as possible after the conversation (2007, p. 237). A practical suggestion: reserve ten minutes for this right afterwards. You capture what the recording doesn't contain: where someone hesitated and what surprised you.

Transcript. Typing up a recording is tedious and takes a lot of time, especially with groups in which people talk over one another (Malhotra and Birks 2007, p. 240). For expert interviews, Meuser and Nagel simplify the task in two ways. They consider elaborate notation rules for pauses and emphasis unnecessary, because what matters is the content. And for them, a complete transcript of the whole recording is not the default: material relevant to the research question is transcribed, and the rest is summarized (Meuser and Nagel 2002, p. 83). Programs that convert speech to text automatically save time. Check beforehand where the recording is processed and whether your consent covers that, and check every automatically generated transcript against the recording.

In the transcript, replace names with codes, for instance G1 to G8 for the conversations. How to quote responsibly later on is covered in the article on open-ended questions.

How do transcripts become findings?

Transcripts become findings in four steps: paraphrase, add headings, compare, verify. The first three follow Meuser and Nagel, who describe them for expert interviews (2002, pp. 83-88). The same steps can be applied to interviews with people who speak about their own experience. The fourth step, verifying, is described by Malhotra and Birks (2007, p. 247). If your conversations serve to prepare or explain a survey, you can end the analysis as soon as this purpose is fulfilled, that is, when the topics and hypotheses for the next steps are settled (Meuser and Nagel 2002, p. 77).

The analysis doesn't begin only after the last conversation. It runs alongside data collection from the start, because every conversation changes what you ask in the next one (Malhotra and Birks 2007, p. 236).

Verifying also includes comparing the expert interviews with the learner interviews. More on that below. If the second person reads a passage differently from you, the verification has not failed. You read the passage again together and record in the report what could be clarified and what remains open.

The second step is related to coding open-ended responses. How to form categories and write a codebook is covered in the article on open-ended questions. That article also explains how to check whether two people code in the same way and what AI assistants can do in coding.

In the example: eight conversations, eight reasons

A theme overview like this, with reasons in the rows and conversations in the columns, makes your interpretation traceable for others, and that is part of verifying (Malhotra and Birks 2007, pp. 244 and 247). On p. 247, the authors describe two further checks: you compare different data and methods with one another, and you present your conclusions to the interviewees. The first is called triangulation, defined in my course as "combining methods within a study on the same topic" (Marx 2026, slide 16).

The language school compares the expert interviews with the learner interviews. The program director says that the transition between the levels is clearly defined. Four former learners didn't know which course comes next. The two statements don't have to contradict each other. A clearly defined transition does not necessarily mean that learners know which course to choose next. The difference between the program director's account and the interviewees' experience is, for now, itself the finding. How the rule is communicated and understood is something the school checks against further material.

How do the conversations turn into a better questionnaire?

You carry three things over from conversations into a questionnaire: the answer options, the words and the hypotheses.

Answer options. The eight reasons become the answer options of a closed-ended question, worded the way a respondent would say them. Without the interviews, the school would probably have asked about price, time and satisfaction and never offered "I wasn't sure which course to take next". Always add "Other (please specify)" with a text field: it shows you later whether your list was missing something important.

Words. Use the words people used in the interviews. Nobody said "insufficient transparency of course progression". Several said: "I didn't know which course was the next one for me." Schnell names exactly this as the purpose of preliminary conversations: becoming familiar with the everyday assumptions and the language of the later respondents (Schnell 2019, p. 128).

Hypotheses. Group discussions deliver hypotheses that you then test on many people (Schumann 2019, p. 261), and the same holds for semi-structured interviews (Jacob et al. 2019, p. 66). The school suspects, for example, that the course times mainly affect people with changing working hours, and therefore includes a question about working hours. How to build the questionnaire is covered in the article on questionnaire design. Why measuring how common these reasons are still requires responses from real people is covered in the article Why surveys still matter.

Worked example: what the follow-up survey shows

The invented example shows the division of labor in numbers. The conversations delivered the list of reasons. They could not deliver the ranking of the reasons. This is clearest for the reason "I had reached my goal". An obvious explanation: someone who has reached their goal doesn't experience that as an obstacle and, in an open conversation, is more likely to talk about what bothered them. Only when everyone had the same list in front of them did a third choose this reason. That is the third of the reasons named above: not mentioned doesn't mean denied. If the school had concluded from the five mentions of the course times alone that the timetable has to be changed first, it would have overlooked that the unclear choice of course is mentioned more often among the 420 respondents. The survey percentages, too, describe only the 420 respondents at first. Whether they hold for all former learners without a follow-on course depends on the selection and on who responded (see survey response rate). A margin of error can only be stated under the assumptions of a random sample (see confidence interval).

After that, the path leads back into conversation. With the survey result in hand, the school invites two focus groups and presents three proposals to them, among them a personal course recommendation at the end of every course. Now the question is no longer about reasons but about reactions to something concrete, and a focus group is well suited to that task. In both groups, the personal course recommendation is well received. But the participants would like to get it a few weeks before the end of the course and together with the dates of the follow-on course, not in the last lesson. That, too, is a finding that helps explain people's reactions rather than measure how common they are. Whether more people book a follow-on course after the course recommendation is introduced is something the booking data will show. Whether the recommendation is the reason for it remains open.

How can QUESTIONSTAR help with interviews and focus groups?

A plug for our own tool: QUESTIONSTAR is a survey tool that can take work off your hands at three stages around the interviews. Beforehand, you can screen participants. With the invitation, you can collect sign-ups, appointment choices and consent. Afterwards, you can run the follow-up survey.

Finding participants: the screening survey. A short questionnaire clarifies who fits your criteria: experience with the topic, usage, a few personal details and the willingness to have a conversation. With skip logic, you route people who don't fit past the remaining questions to a closing page of their own, where you end the questionnaire. With quotas, you limit the number of responses from individual groups, for instance if you are looking for only a few people from each type of course. Anyone who fits and would like to talk enters their email address. Through double opt-in, the person confirms by email that you may contact them, and the address goes into a contact list of its own. You can add further answers to this list, for example the answers to your selection criteria. You distribute the link to the screening survey openly, or you invite an existing list personally by email. That works in the free plan too.

Appointment and consent: the sign-up form. The invitation to the conversation can be a second short questionnaire: the offered appointments to choose from, the email address, an explanation of the recording with a question that allows yes and no, and a field for questions. The table shows how you set it up.

Whether consent given through a form is sufficient for your purpose is something to clarify with your data protection officer.

If contact details are to be kept separate from the other answers, save them in a separate file through the privacy settings. This setting can't be undone. Leave "Anonymize IDs" off in the screening survey, because with it the email address can no longer be linked to the responses.

Counting: the follow-up survey. What the conversations suggest, you test on many people. The topics become answer options in a multiple choice question, supplemented by "Other (please specify)" with a text field. The reports show the frequencies. In the crosstab, which breaks the answers down by group, you see whether groups differ, for instance people with fixed and with changing working hours. You export the raw data as XLSX, CSV or SPSS. All of this is included in the free plan. QUESTIONSTAR doesn't offer built-in coding of the free-text answers under "Other". If QUESTIONSTAR is connected to an AI assistant through the MCP interface, the assistant can read the responses to your survey and code them according to your codebook. In that case, the assistant does the coding, not QUESTIONSTAR.

If you asked in the survey who would be willing to have a more detailed conversation, the circle closes: after the last page, QUESTIONSTAR can redirect to a second, separate survey or show a link to it (end of survey). There, interested people leave their address separately from their answers. If URL parameters are attached to your survey link, such as a participant identifier, they are passed on to the target address automatically during the redirect. If the address is to remain unconnected to the answers, use a link without such parameters.

A survey can also help with recording if you conduct an interview along a questionnaire. To do so, set up the topics or questions of your guide as a questionnaire. Whoever conducts the interview goes through it on a tablet during the conversation and notes key points. The "Audio recording" feature records the sound while the questionnaire is being filled in, and nothing more (features). As with every recording: only with the consent of the person interviewed, as described above.

Checklist: planning interviews and focus groups

1. Write down the question the conversations are meant to answer, and record what you don't know.
2. Choose the format: one-on-one interviews for decisions, sensitive topics and specialist knowledge, groups for reactions to something concrete.
3. Decide who can contribute: choose experts for their knowledge rather than rank, select people with a range of relevant personal experiences, and recruit participants with relevant characteristics in common within each focus group.
4. Write the guide: opening, transition, three to five key topics with probes, closing. Ask open questions, not leading ones.
5. Clarify who conducts the conversation, if possible someone whose work the interviewees don't have to judge.
6. Invite, limit the places per appointment, obtain consent to the recording, send a reminder shortly before the appointment, compensate people for their effort.
7. Treat the first conversation or the first group as a trial and adjust the guide afterwards.
8. Record with consent and take notes right after the conversation.
9. Decide beforehand when you will stop, and allow time and budget for additional interviews. After each interview, record what was new.
10. Analyze: paraphrase, add headings, compare, verify. Create a theme overview.
11. Report what exists and how it is connected, without percentages. Put what you want to count into a survey.

Frequently asked questions about expert interviews and focus groups

What is an expert interview, and how does it differ from a semi-structured interview?

The semi-structured interview is the form of conversation: the topics are fixed, the order and wording of the questions are not (Möhring and Schlütz 2019, p. 5). An expert interview is a semi-structured interview with a person who is interviewed for their knowledge of a field, rather than for their personal experience of the issue. An expert is someone who bears responsibility for a solution to a problem or who has privileged access to information about groups of people or decision processes (Meuser and Nagel 2002, p. 73). Who is an expert depends on the research question, not on rank.

What is a focus group?

A focus group is a moderated discussion with a small group, usually six to ten people, who talk about a given topic in their own words (Malhotra and Birks 2007, pp. 182-183). It is also called a group discussion. It is suited to reactions to something concrete, such as a draft or an offer.

How many expert interviews or semi-structured interviews do you need?

There is no fixed number. You conduct conversations until no new topics come up, and you decide beforehand how you will recognize that. In a single study with a homogeneous group and the same questions for everyone, most topics had appeared within the first twelve interviews (Guest et al. 2006, p. 66). The authors themselves warn against turning this into a rule for every study. With very different interviewees, more conversations may be needed, and the authors do not claim that their finding applies to open, exploratory interviews (pp. 75, 79).

How many participants does a focus group have?

Usually six to ten. With fewer than six, the discussion often lacks momentum. With more than ten, the group easily splits into subgroups (Malhotra and Birks 2007, p. 183). For intensive probing, there are mini groups of four or five people (p. 194). A session typically lasts one and a half to two hours (p. 184).

Are the results of interviews and focus groups representative?

No. There are few participants, they are selected deliberately, and not everyone is asked the same question. The results show which reasons and points of view exist and how the interviewees explain connections. How widespread they are in a population can only be determined with a quantitative survey based on a suitable sample (Jacob et al. 2019, p. 5).

Do I have to transcribe interviews in full?

Not always. Meuser and Nagel do not treat a full transcript as the default for expert interviews: material relevant to the research question is transcribed, and the rest is summarized (Meuser and Nagel 2002, p. 83). For a thesis, the requirements of your department apply, so ask there.

Can I run a focus group online?

Yes, a focus group can also be run by video call. Malhotra and Birks describe online groups that exchange views in writing and name as an advantage that nobody has to travel and that people in different locations can take part (2007, pp. 201-202). You need consent to the recording online just as you do in a room (p. 200).

Sources

Bogner, A. and Menz, W. (2009). The Theory-Generating Expert Interview: Epistemological Interest, Forms of Knowledge, Interaction. In A. Bogner, B. Littig and W. Menz (Eds.), Interviewing Experts (pp. 43-80). Basingstoke: Palgrave Macmillan. DOI 10.1057/9780230244276. pp. 52-55.

Brown, L. D., Cai, T. T. and DasGupta, A. (2001). Interval Estimation for a Binomial Proportion. Statistical Science, 16(2), 101-133. DOI 10.1214/ss/1009213286. pp. 107 and 115 (Wilson interval for small samples), explained in the article on the confidence interval.

Froschauer, U. and Lueger, M. (2009). Expert Interviews in Interpretive Organizational Research. In A. Bogner, B. Littig and W. Menz (Eds.), Interviewing Experts (pp. 217-234). Basingstoke: Palgrave Macmillan. DOI 10.1057/9780230244276. p. 221 (English rendering of Meuser and Nagel's definition of the expert), p. 225 (saturation).

Guest, G., Bunce, A. and Johnson, L. (2006). How Many Interviews Are Enough? An Experiment with Data Saturation and Variability. Field Methods, 18(1), 59-82. DOI 10.1177/1525822X05279903. pp. 59, 63, 66, 75, 79 (full text, accessed on November 12, 2026).

Jacob, R., Heinz, A. and Décieux, J. P. (2019). Umfrage. Einführung in die Methoden der Umfrageforschung [Surveys: an introduction to the methods of survey research] (4th ed.). Berlin/Boston: De Gruyter Oldenbourg. In German. pp. 5, 66, 136.

Littig, B. (2009). Interviewing the Elite - Interviewing Experts: Is There a Difference? In A. Bogner, B. Littig and W. Menz (Eds.), Interviewing Experts (pp. 98-116). Basingstoke: Palgrave Macmillan. DOI 10.1057/9780230244276. p. 103.

Malhotra, N. K. and Birks, D. F. (2007). Marketing Research. An Applied Approach (3rd European edition). Harlow: Pearson Education. Chapters 6 to 9, pp. 154-247.

Marx, P. (2026). Principles of Survey Research. Online course, chapter 1, slides 13, 15 and 16. QUESTIONSTAR. Slide 13, slide 15, slide 16.

Meuser, M. and Nagel, U. (2002). ExpertInneninterviews: vielfach erprobt, wenig bedacht. Ein Beitrag zur qualitativen Methodendiskussion [Expert interviews: often tried, rarely reflected on. A contribution to the debate on qualitative methods]. In A. Bogner, B. Littig and W. Menz (Eds.), Das Experteninterview. Theorie, Methode, Anwendung [The expert interview: theory, method, application] (pp. 71-93). Opladen: Leske + Budrich. In German. DOI 10.1007/978-3-322-93270-9. pp. 73, 77-78, 83-88.

Möhring, W. and Schlütz, D. (2019). Die Befragung in der Medien- und Kommunikationswissenschaft. Eine praxisorientierte Einführung [Surveys in media and communication studies: a practical introduction] (3rd ed.). Wiesbaden: Springer VS. In German. DOI 10.1007/978-3-658-25865-8. p. 5.

Schnell, R. (2019). Survey-Interviews. Methoden standardisierter Befragungen [Survey interviews: methods of standardized surveys] (2nd ed.). Wiesbaden: Springer VS. In German. DOI 10.1007/978-3-531-19901-6. pp. 128-129.

Schumann, S. (2019). Repräsentative Umfrage. Praxisorientierte Einführung in empirische Methoden und statistische Analyseverfahren [Representative surveys: a practical introduction to empirical methods and statistical analysis] (7th ed.). Berlin/Boston: De Gruyter Oldenbourg. In German. p. 261.

Welch, C., Marschan-Piekkari, R., Penttinen, H. and Tahvanainen, M. (2002). Corporate elites as informants in qualitative international business research. International Business Review, 11, 611-628. Cited in Littig (2009), p. 103.

I hope your interviewees feel ready to tell their stories, and that you have the patience to keep listening even when you think you already know the answer.

Related articles:

• Open-ended questions: how to ask and analyze them
• Questionnaire pretest: how to test your survey before launch
• Delphi method: steps, example and how to define consensus

Dr. Paul Marx

Author: Dr. Paul Marx
Dr. Paul Marx is the founder and Managing Director of QUESTIONSTAR and was a professor of marketing at the University of Siegen.