How to set up a qualitative interview properly

A qualitative interview delivers what no scale can: the reasoning behind an opinion. If you want to understand motives, attitudes, and decision logic, this method is unavoidable. This guide shows you step by step how to prepare, conduct, and analyze a qualitative interview the right way.

Qualitative or quantitative? The upfront decision

Quantitative research relies on standardized instruments and large samples. It produces statistically reliable statements about a defined target population. Qualitative research, by contrast, reconstructs behaviors, perspectives, and decision logic at the single-case level. It does not measure, it interprets in context.

Method follows research question. If you ask about frequencies, shares, or causal effects, your study belongs in the quantitative camp. If you ask about the "why," about meaning, or about lived experience, you need qualitative data.

Interview formats at a glance

Qualitative research is dominated by semi-structured formats. Structure comes from a guide that secures the topics without forcing the conversational flow. Choose the format that fits your research question:

For most use cases in market research, employee surveys, and policy consulting, the semi-structured guided interview or the expert interview is the right pick. It offers enough structure for comparability between conversations and enough openness for unexpected themes.

Building the guide with the SPSS method

The SPSS method by Helfferich (2009) is the established standard for building an interview guide. The four letters stand for four work steps:

A good guide is not a questionnaire, it is a topic map. Per block you need one story-generating opening question plus a short list of follow-ups. Keep it on a single page so you can actually use it during the conversation.

Standard structure of a guide

1. Opening: greeting, brief intro, purpose of the interview, note on recording and anonymization, ask for consent.
2. Warm-up: a low-threshold question about the person or their role to build trust.
3. Main part: three to five topic blocks, each with one opening question and prepared follow-ups.
4. Deepening: ask for concrete examples, critical moments, wishes for the future.
5. Closing: "Is there anything we haven't talked about?" as a door-opener for what you did not ask.
6. Goodbye: thank them, briefly outline how the results will be used.

Recruiting interview partners

Selection follows content relevance, not statistical representativeness (theoretical sampling). You look for people who bring maximally contrasting or particularly typical knowledge to your question.

• Snowball sampling: ask interviewees to recommend further partners. Useful when the target group is hard to reach.
• Gatekeeper sampling: approach key people in institutions who can connect you internally to the right contacts.
• Direct research: identify candidates via LinkedIn, industry directories, association registers, or events.

How many interviews are enough? The answer is theoretical saturation: stop when new conversations no longer produce substantially new themes. In practice this often means eight to twelve interviews for homogeneous groups, fifteen to thirty for heterogeneous ones.

Pre-script and post-script: the underrated helpers

A pre-script is a short note before the interview: what do you expect? Which hypotheses do you carry into the conversation? What mood are you in? The pre-script makes your own assumptions transparent and protects you from unconsciously steering the talk.

A post-script is written immediately after the interview and captures what is not on the audio: atmosphere, body language, rapport, disruptions, striking themes. These notes are gold during interpretation later on.

Running the interview

• First obtain consent for recording, then start the device.
• Begin with the story-generating opening question and give the person room. Hold silences.
• Listen actively: short confirmations, mirroring, follow-ups like "Can you give me an example?"
• Avoid leading questions and evaluations. Do not offer interpretations.
• Use the guide as a safety net, not a checklist. Adjust the order when the conversation takes its own path.
• On sensitive topics, offer pauses and emphasize the right not to answer.
• At the end, explicitly invite additions and explain how the data will be used.

Transcription: the bridge to analysis

Without a transcript there is no systematic analysis. Choose transcription depth according to your analytical aim:

• Simple transcription: full wording, smoothed grammar, no pauses or filler words. Suitable for most content-analytic work.
• Extended transcription: includes pauses, laughter, emphasis. Useful for conversation analysis or fine-grained semantic work.
• Phonetic transcription: only required for linguistic questions.

Practical rule of thumb: one hour of audio takes about five to eight hours of manual transcription. AI-powered transcription cuts this to a few minutes plus review.

Analysis with qualitative content analysis

Qualitative content analysis is the most widely used analytical procedure. It analyzes the material systematically and theory-driven. For most use cases, the content-structuring variant by Kuckartz (2016) or Mayring's (2015) approach is the right choice. Both work well with tools like MAXQDA or ATLAS.ti.

1. Initial reading: read the material, write memos, mark first observations.
2. Develop main categories: derive deductive categories from research question and guide.
3. First coding: work through the material along the main categories.
4. Inductive sub-categories: develop finer distinctions within the main categories from the material itself.
5. Second coding: code the full material with the refined category system.
6. Category-based analysis: condense, contrast, and relate content per category.
7. Presentation: build thematic matrices, profiles, typologies, or visual models.

Quality criteria for qualitative research

Classic criteria such as validity and reliability apply only to a limited extent in qualitative work. Steinke (2007) established the following criteria:

• Intersubjective traceability: document procedure, analysis, and interpretation completely.
• Indication of process: justify methodological decisions.
• Empirical grounding: back claims with quotes from the material.
• Limitation: name scope and limits of the results.
• Coherence: surface and address contradictions in the material.
• Relevance: explain the usefulness of the findings for the question.
• Reflected subjectivity: disclose your own role and assumptions.

The most common mistakes and how to avoid them

• Too many questions in the guide: one A4 page max, otherwise it becomes a hidden quantitative interview.
• Closed or leading questions: they close the narrative space instead of opening it.
• Voicing your own hypotheses: respondents pick them up and serve you the answers they think you want.
• Filling silences with your own words: silence produces the most valuable answers.
• No post-script: nonverbal impressions are lost forever.
• Starting analysis straight from the audio: transcribe first, analyze second.
• Working without a codebook: without an explicit category system the analysis is not traceable.
• Expanding the sample arbitrarily: theoretical saturation is the criterion, not gut feeling.

When AI-powered interviews are the better choice

Classic qualitative interviews scale poorly. Ten to thirty conversations per study is realistic, more is rarely feasible. Anyone who needs hundreds or thousands of voices in the same methodological depth hits limits: scheduling, travel, transcription effort, interviewer bias across many conversations.

This is exactly where AI-powered voice interviews come in. They conduct the conversation along your guide, listen actively, follow up situationally, and deliver structured transcripts. The methodological principles in this guide stay the same: story-generating opening questions, no leading, room for open answers. What changes is the scale.