# Semi-Structured Interview Guide: Template and Example

Semi-structured interview guide: definition, adaptable template, 15 open-ended questions, complete example, and traceable analysis.

- Canonical URL: https://www.harmate.com/en/blog/semi-structured-interview-guide-method-steps-and-15-examples
- Author: Harmate Team
- Published: 2026-08-14
- Updated: 2026-11-07T13:49:18.933789+00:00
- Language: en

## Content

# Semi-Structured Interview Guide: Template and Example

A semi-structured interview is a conversation guided by themes prepared in advance, with enough flexibility to explore what the participant actually says. The guide is therefore neither a questionnaire read aloud nor an improvised conversation: it is a flexible map connected to a decision or research question.

A useful guide connects four elements: an explicit objective, open-ended questions, neutral probes, and a record that preserves the context behind each conclusion. This guide provides a definition, a five-step method, an adaptable template, a complete example, and 15 question and probe formulations.

Jump to the [adaptable interview template](#a-semi-structured-interview-guide-template-you-can-adapt), the [worked example](#a-complete-example-why-a-training-survey-is-abandoned), or the [interview analysis method](#analyze-interviews-without-losing-context).

Already conducted the interviews? If the answers are in a structured CSV, JSON, or XLSX file, [import your responses into Harmate](/en/product/response-import), review the columns, and start the analysis.

## What is a semi-structured interview?

A semi-structured interview sits between two formats:

- a structured interview, which asks the same questions in the same order;
- an unstructured interview, which lets the conversation develop with very little predefined structure.

In a semi-structured interview, the interviewer prepares themes and possible questions. They can change the order, rephrase a prompt, and follow an unexpected direction when it helps explain the topic. The broad areas remain comparable across participants, while each account keeps its context.

This approach matches the definition described by [DeJonckheere and Vaughn](https://fmch.bmj.com/content/7/2/e000057): a dialogue supported by a flexible protocol, follow-up questions, and probes. Rigor does not mean preventing the conversation from moving. It means knowing why it moves and preserving what that movement teaches you.

### An open-ended question is not enough

An open-ended question gives the participant room; it does not automatically make an answer richer, more accurate, or representative. The value of an answer also depends on the situation described, the quality of the probes, the range of cases collected, and whether the analysis preserves contradictions.

Keep these ideas separate:

- the richness of an account concerns detail and context;
- the validity of an interpretation depends on the design, data, and stated limits;
- representativeness cannot be inferred from a small number of rich conversations.

A semi-structured interview helps you understand experiences, mechanisms, and differences. It does not turn a few accounts into a frequency measure by itself.

### When should you use this method?

It is useful when you need to understand:

- how someone actually experienced a situation;
- why they made a decision;
- which obstacles or trade-offs are hidden by quantitative data;
- how a process varies across people or contexts;
- which hypotheses should later be tested at a larger scale.

It is less suitable when you only need a standardized measure, ranking, or statistical comparison. In that case, a closed-ended survey will often be more efficient. To choose between exploration and measurement before writing your questions, see [the difference between qualitative and quantitative research](https://www.harmate.com/en/blog/qualitative-vs-quantitative-research-which-should-you-use-first).

For a hiring decision, use a framework suited to that context. The [structured hiring interview guide](/en/blog/structured-hiring-interview-guide-criteria-questions-and-notes) connects questions to role criteria and prepares a human review of the evidence gathered. Its purpose differs from the research interview presented here.

## A semi-structured interview guide template you can adapt

Start by writing the decision to inform, the population concerned, and the situation being studied. Then adapt the table below. Each row links a theme to a signal, a question, possible probes, and a record to preserve. Remove rows when they do not support the decision; do not add questions just to make the guide longer.

| Theme | Signal to look for | Opening question | Neutral probes | Record to preserve |
|---|---|---|---|---|
| Starting situation | Trigger, context, people involved | Tell me about the last time you [performed the activity being studied]. | When did this happen? Who was involved? | Date or period, situation, words used |
| Actual journey | Steps, detours, interruptions | Can you walk me through what happened from beginning to end? | What did you do next? Where did it change? | Sequence of events, decision made |
| Friction and trade-offs | Constraints, compromises, abandoned options | What made this situation difficult? | What options did you consider? What shaped your choice? | Concrete example, constraint, consequence |
| Difference and exception | A different case or changing condition | Can you think of a situation where it worked differently? | For whom or in what context? What changed? | Contrasting case and context |
| Next step and decision | Outcome, next action, remaining need | What did you do after that experience? | What was checked? What would need to be decided now? | Action, observed outcome, open question |

To use it:

1. replace the bracketed elements with a specific situation;
2. choose the signals that are genuinely useful for the decision;
3. keep probes as options, not as a mandatory list;
4. write down what was said, observed, and inferred in separate fields.

The template keeps interviews comparable without preventing an unexpected answer. It does not replace the interviewer’s judgment or contextual analysis.

## Before fieldwork: interview the commissioning client

The first interview in a study is often with the person who requested it. Its purpose is to clarify the decision and its boundaries, not to validate the client’s first explanation or promise that fieldwork will confirm it. If several people commissioned the work, speak with them separately when their expectations may differ, then record the differences rather than manufacturing one brief.

Use this 30–45 minute guide as a working template. Adapt it to the assignment and leave questions open when the evidence is not yet available.

| Theme | Ask the commissioning client | Follow-up if needed | Record before fieldwork |
|---|---|---|---|
| Trigger and decision | What happened that made this work necessary now? What decision will follow the study? | Who will make it, and when does the evidence need to be ready? | Decision owner, decision point, intended use |
| Situation and evidence | What have you observed directly? Which examples or existing records support that account? | What have you heard second-hand? What might contradict the current explanation? | Observation, source, interpretation, and unknowns kept separate |
| People and scope | Whose experience must be understood for this decision? Who is affected but may be hard to reach? | Which roles, locations, or situations differ in a way that could change the conclusion? | Groups to invite, inclusion logic, known absences |
| Constraints and freedom | What is fixed by law, safety, budget, timing, or authority? What can still change? | Which constraint is documented, and who can confirm it? | Non-negotiables, open choices, person able to verify each |
| Expectations and disagreement | What result do you expect? What would make you revise that expectation? | Who in the commissioning team sees the issue differently? What is the strongest alternative explanation? | Expected result, counter-view, disagreement to preserve |
| Fieldwork and use | What can participants safely be told about the purpose and use of their answers? | What should remain confidential, and what can be reported back? | Recruitment message, consent/data-use checks, reporting audience |
| Decision after the study | What evidence would be enough to compare options? What would remain uncertain? | What will you do if the study does not support a single answer? | Comparison criteria, limits, next-step options |

### Close with a pre-fieldwork decision record

Complete this sheet with the client and send it back for correction before recruitment begins:

| Item | Agreed wording |
|---|---|
| Decision to inform | [One decision and its owner] |
| Questions for fieldwork | [Questions participants can answer from experience] |
| Working assumptions | [What the commissioning team currently believes] |
| Alternative explanations | [What could also account for the situation] |
| Fixed constraints | [Constraint, source, and person who can confirm it] |
| Open choices | [What participant evidence could still change] |
| People to hear from | [Groups and reason each is relevant] |
| Evidence to preserve | [Examples, documents, excerpts, and context] |
| Out-of-scope requests | [Questions this assignment cannot answer] |
| Review point | [Who will consider the findings and what decision follows] |

Do not turn this record into a promise of representativeness or a test that participants must pass. It makes the brief inspectable and gives the interviewer permission to explore evidence that challenges the initial framing. When an assumption changes, note what changed and whether the guide or recruitment plan needs revision.

## A complete example: why a training survey is abandoned

### The framing

Decision to inform: understand why learning and development managers start preparing a training survey and then decide not to send it.

Population: managers who prepared at least one survey during the previous six months.

Research question: at what point does preparation become too costly, and which trade-off leads to abandonment?

The guide does not start with “which features are missing?” That wording already imposes an explanation. It starts with a recent experience.

### The applied guide

| Theme | Main question | Possible probes | What to preserve |
|---|---|---|---|
| Recent experience | Tell me about the last time you prepared a training survey. | What triggered the request? What happened next? | Timeline and context |
| Preparation | How did you build the questions? | What did you reuse, change, or remove? | Actions and criteria |
| Friction point | When did it start to become difficult? | What did you try to keep moving? | Specific event and perceived cost |
| Trade-off | What other options did you consider? | Why did you reject them? | Options and compromises |
| Outcome | What did you eventually do? | What would have changed the decision? | Consequence and remaining need |

### One excerpt and its analysis

Fictional excerpt, used only to demonstrate the method:

> “I started with last year’s survey. After two hours, I realized the questions no longer matched the program. I asked the trainers to review it, but no one had time. I eventually sent three broad questions on the last day.”

This account lets you distinguish several elements:

- reported fact: an older survey was reused and then changed;
- context: the program had changed and the trainers were unavailable;
- friction point: review became a bottleneck;
- trade-off: sending three broad questions rather than delaying the survey;
- limit: one account cannot show that this sequence is common.

A useful next probe would be: “What happened after you sent the three questions?” It seeks an observable consequence, not a general opinion. In analysis, this case can be compared with people who completed preparation or chose another solution.

## Build a guide in five steps

A methodological review by [Kallio and colleagues](https://onlinelibrary.wiley.com/doi/abs/10.1111/jan.13031) proposes checking whether the method is appropriate, using existing knowledge, drafting a preliminary guide, piloting it, and formalizing the complete version. Here is a practical application.

### 1. Start with the decision you need to inform

Write one sentence describing what the study must help you understand or decide.

> Understand why learning and development managers abandon survey preparation before sending it.

This is more useful than “collect feedback about our tool.” It identifies a situation, a population, and a phenomenon. It also helps you remove interesting questions that do not serve the research goal.

### 2. Turn the objective into themes

Organize the guide around four to six themes: context, recent experience, difficulties, trade-offs, previous solutions, and expected outcome.

For each theme, prepare:

- one opening question;
- two or three possible probes;
- the signal you need: a fact, behavior, constraint, perception, or decision.

This structure keeps a common thread without making the conversation artificial.

### 3. Ask for a story before an explanation

Hypothetical questions easily produce idealized answers. Start with a recent, concrete experience: “Tell me about the last time you…” Then ask what happened, in which order, who was involved, and which criteria shaped the decision.

For wording and framing risks, see our guide to [asking open-ended questions without bias](/en/blog/open-ended-questions-get-actionable-data-without-bias). An open-ended question still needs testing: if it is vague, double-barreled, or loaded with an assumption, it can produce a story that is difficult to interpret.

### 4. Pilot the guide before fieldwork

Run at least one pilot interview under realistic conditions. Check:

- whether the questions are understood without extra explanation;
- whether the order creates repetition or influences later answers;
- whether some themes remain too abstract;
- whether probes elicit concrete examples;
- whether the duration matches what you promised.

After the pilot, remove questions that do not support the objective. A shorter guide leaves more room for listening.

### 5. Plan the analysis alongside the interview

Before collecting data, choose the unit you will compare: situations, decisions, barriers, journeys, or participant profiles. Also decide how you will preserve the link between each theme and the excerpts that support it.

Planning analysis early prevents two common failures: collecting a large set of quotes with no decision path, or forcing answers too quickly into categories defined in advance.

## 15 semi-structured interview questions and probes

These are starting points, not a universal script. Replace the bracketed text with your topic and keep only the prompts that serve your objective.

### Open with experience

1. **Can you tell me about the last time you [performed the activity being studied]?**
2. **What was the context?**
3. **What triggered that process at that particular time?**
4. **What were the main steps from beginning to end?**
5. **Who else was involved, and at what point?**

### Explore without leading

6. **When you say “[word used by the participant],” what does that mean to you in practice?**
7. **Can you give me a specific example?**
8. **What happened immediately after that?**
9. **What was easiest? What was most difficult?**
10. **What information did you use to make that choice?**
11. **What other options did you consider?**
12. **Was there a moment when your view or behavior changed?**

### Test boundaries and close

13. **Can you think of a situation where it worked differently?**
14. **If you could change one thing about that experience, what would it be?**
15. **What important question have I not asked?**

Probes 6 through 13 are often more useful than adding new headline questions. They build on the participant’s own words and reduce the risk of imposing the interviewer’s framework. They do not remove the need to compare contexts and search for divergent cases.

## Conduct the interview without steering the conversation

### Set a clear frame

Before starting, explain the purpose, how responses will be used, the confidentiality offered, and the participant’s ability to decline a question. If you record the conversation, inform the participant and obtain the appropriate agreement. Data protection and research requirements depend on your jurisdiction and context; this guide is not legal or ethics advice.

### Let the participant finish

Silence is not failure. A few seconds often give someone time to clarify their thinking. Avoid completing sentences, stacking several questions together, or showing approval only when an answer confirms your hypothesis.

### Keep the structure flexible

You can change the order of themes, but note which ones were not covered. Keep the same broad areas of exploration across participants so cases remain comparable. After every interview, write a short memo covering context, surprises, emerging hypotheses, and the interviewer’s possible influence.

### Separate statements, observations, and interpretations

In your notes, distinguish what the participant said, what you observed, and what you inferred. A summary sentence should not replace the excerpt or situation that supports it.

## How many participants do you need?

There is no universal number. The appropriate sample depends on the precision of the aim, the diversity of participants, the richness of the interviews, the theoretical framework, and the analysis method.

The concept of “information power,” proposed by [Malterud, Siersma, and Guassora](https://doi.org/10.1177/1049732315617444), focuses on how much relevant information a sample contains rather than applying an automatic threshold. A tightly scoped project with directly relevant participants may require fewer interviews than a broad exploration across several contexts. Document your recruitment logic and reassess it as data collection progresses.

Do not confuse saturation, repetition, and representativeness. Similar answers may reflect a homogeneous sample; a minority case may reveal an important condition. Ask instead: which variations still need to be understood before making the planned decision?

## Analyze interviews without losing context

To compare interviews, first read each case in context, then compare cases across shared dimensions. A matrix helps connect these readings when each summary points back to its source. It organizes analysis without deciding for the analyst. One methodological paper describes cases as rows and codes as columns, with contextual summaries, to support within-case and cross-case analysis ([Gale et al., 2013](https://pmc.ncbi.nlm.nih.gov/articles/PMC3848812/)). The grid below is a working aid, not a universal method.

### Start with the cases, not the table

Reread each transcript with the research question in mind. Write a brief memo on the situation, sequence, and participant’s explanation. Note probes and passages to revisit. Choose the unit being compared — a person in a situation, a reported episode, or a whole interview — and do not mix these levels. Keep an identifier and a reference back to the transcript, such as “I1, §12.”

### Fictional example: from accounts to a matrix

These three short accounts are fictional; their references illustrate a convention, not real sources. I1 is unsure what a survey question is asking them to evaluate (I1, §12). I2 stops answering after an outside interruption late in the day (I2, §8). I3 understands the question but hesitates to answer with their name attached (I3, §15). In a real corpus, each reference should lead back to the passage and its context.

| Case / context | Understanding the question | Completion conditions | Stated identification |
|---|---|---|---|
| I1, survey on a phone | Subject may be ambiguous (I1, §12) | Phone; other conditions to check | Not explored |
| I2, late in the day | Not explored | Outside interruption before stopping (I2, §8) | Information unavailable |
| I3, named response | Question understood (I3, §15) | Information unavailable | Concern expressed (I3, §15) |

Each cell gives a summary and source reference, or explicitly says “not explored,” “information unavailable,” or “to check.” Do not leave readers to guess why a cell is empty. **Not explored does not mean absent from the person’s experience.** Recurrence in this corpus does not measure prevalence in the wider population.

Read across rows first to keep each case intact, then down columns to compare a shared dimension. Comparison assumes topics were covered in sufficiently comparable ways; a different probe or context may explain an apparent difference. Look for divergent cases and revisit transcripts before drawing a conclusion. Present an interpretation bounded by the corpus, with uncertainty and a concrete next check. Reflexivity — making explicit how your position and choices shape the reading — helps readers examine the interpretive path; a matrix cannot replace that work.

For the move from data to codes, see [DeJonckheere and Vaughn](https://fmch.bmj.com/content/7/2/e000057); the [COREQ checklist](https://www.equator-network.org/reporting-guidelines/coreq/) helps document study design and analysis. Our [thematic analysis example](/en/blog/thematic-analysis-codes-themes-and-interpretation) works through the move from excerpts to themes.

Once interviews are recorded as open-ended responses, you can [review themes and their source excerpts in Harmate](/en/product/open-questions) before choosing evidence for a decision file. For a larger corpus, our method for [analyzing 200 open responses](/en/blog/analyzing-200-open-responses-without-bias-a-method-for-actionable-decisions) shows how to keep differences and sources visible.

## Live interview, asynchronous collection, or focus group?

A live interview remains preferable when the relationship, nonverbal cues, topic sensitivity, or in-the-moment methodological judgment are central. Asynchronous collection can help reach dispersed participants, give people time to formulate an answer, or explore a shared topic with a larger group.

The formats are not interchangeable. In an asynchronous setting, probes should remain short, relevant, and limited. Participants should understand why clarification is being requested. In a focus group, interaction between participants becomes data in its own right: agreements, silences, and disagreements must be distinguished from individual opinions. See our [focus group guide and open-ended questions](/en/blog/focus-group-guide-method-agenda-and-20-open-ended-questions) if group discussion is your main design.

This is where [Harmate can complement a remote exploratory interview process](/en/use-cases/exploratory-interviews): distributing a guide, probing selected open-ended answers, and organizing themes while retaining a link to supporting quotes. It does not replace the relationship of a live interview or the researcher’s judgment. It offers a complementary format when scale or availability makes individual interviews difficult.


To connect individual exploration with a collective decision, see our [project lessons-learned method](/en/blog/project-lessons-learned-method-questions-and-an-adaptable-template). It helps organize disagreements, sources, and verifiable actions after several interviews.
## Common mistakes

### Turning the guide into a questionnaire

Reading every question in the same order can reduce listening and prevent you from following a useful detail. Keep common themes, but use the wording as a set of landmarks.

### Treating one open answer as sufficient evidence

A quote illustrates an experience. It does not prove a frequency, cause, or general preference by itself. Record what the case helps you understand and what it cannot establish.

### Adding probes that suggest the answer

“You probably found that frustrating, right?” already contains an interpretation. Ask instead: “What did you feel or do at that point?” and preserve the participant’s words.

### Removing cases that contradict the dominant theme

A divergence may point to a condition, profile, or context that differs. Look for what explains it before labeling it an irrelevant exception.

### Changing the guide without keeping a record

A guide can evolve after a pilot, but note the version, the reason for the change, and the interviews affected. Otherwise, differences between answers may come from the instrument rather than the field.

### Promising representativeness the design cannot support

Explain the recruitment scope, absences, and limitations. A precise conclusion is more useful than an overly broad generalization.

## Checklist before launching interviews

- The research question or decision fits in one sentence.
- Every theme contributes directly to that objective.
- Questions begin with concrete experiences.
- Probes are neutral and non-redundant.
- The template names the signal and the record to preserve.
- The guide has been tested under realistic conditions.
- Participants understand the frame, data use, and their rights.
- Recording, storage, and access rules are defined.
- The unit of analysis and evidence trail are planned.
- Contradictory cases will be investigated rather than discarded.
- Guide versions and adaptations will be documented.

## Frequently asked questions

### What is the difference between a semi-structured interview and a survey?

A survey generally standardizes data collection. A semi-structured interview keeps common themes but adapts the order and probes to each participant’s account. A survey with open-ended questions can use some of the same wording principles without reproducing the full relationship of a live interview.

### Should every participant receive exactly the same questions?

No. The broad themes should remain comparable, but wording, order, and follow-up questions can vary. Document major adaptations when they may influence the analysis.

### How long should a semi-structured interview last?

There is no optimal duration independent of the topic. Promise a realistic format, test the guide, and prioritize useful depth over filling a time slot.

### Can AI analyze semi-structured interviews?

Yes, as an aid for organizing, exploring, or summarizing data, provided that personal data is protected, source excerpts are preserved, and interpretations are reviewed by an accountable person. An automatically generated category is not evidence by itself.

## Key takeaway

A strong semi-structured interview is neither a disguised questionnaire nor an improvised conversation. It connects an explicit decision, a flexible template, neutral probes, and a traceable analysis. Start with a real experience, follow the participant’s words, look for cases that challenge your first assumptions, and preserve the excerpts behind every conclusion. That complete chain—not the length of the guide or the number of questions—is what makes findings useful for decisions.