ARTICLE · Research

In-depth interview questions: seven techniques and the limits of each

Boris Kaptelov · 02.08.2026 · 9 min

There is a common assumption that interview technique amounts to a set of the right questions. Ask well and you get the truth; ask badly and you get nonsense.

It works differently. The form of a question is not a channel through which ready-made knowledge flows; it is an instrument that produces a particular type of data. A free narrative, the reconstruction of an episode, a comparison, an object shown to the respondent, and a diary yield different material about the same situation. They complement one another and do not substitute for one another.

Below are seven techniques — what each one delivers and what it should not be expected to deliver. Choosing among them is the substantive part of preparing a study.

Free narrative

"Tell me how your work with suppliers is organized in general."

An open invitation hands the person control over what counts as important. That is its value: you see the respondent's structure of significance, their language and their classifications — not your own, smuggled in through the question. In the ethnographic interview, Spradley distinguished descriptive, structural, and contrast questions precisely so the conversation would proceed from the participant's own language.

What free narrative does not deliver: completeness depends on verbal fluency, memory, trust, and on what the person considers worth mentioning. Routine and socially awkward practices drop out most often — not because people hide them, but because they seem self-evident. Someone who has been reconciling stock balances by hand every evening for ten years will not bring it up: to them it is not a problem, it is just the evening.

One more thing: the causal links inside a narrative belong to the narrative, not to the events.

Episode reconstruction

"Let's take the last time you had to switch suppliers on short notice. How did it start?"

The technique anchors the conversation to a bounded event: time, place, participants, the starting point, the sequence, the forks in the road, the consequences. Flick's episodic interview method is built on the distinction between episodic knowledge, tied to the circumstances of a specific case, and semantic knowledge, expressed in generalizations.

A concrete episode usually yields more verifiable material than an abstract intention: a generalization the person will assemble on the spot, whereas a recollection can be checked against dates, documents, and system records.

An important caveat, easily lost in the excitement over a detailed account: a retrospective account remains subject to memory errors and rationalization. An episode is not retrieved from an archive; it is reassembled to serve the narrator's current purposes, and knowing the outcome rewrites the history of the decision. Verifiability comes not from the account itself but from checking it against a trace: dates, documents, entries in the system.

The limits: an episode requires that an event has occurred. For needs that have not yet triggered a search, for entirely new categories, and for routines without notable turning points, the technique works poorly.

Timeline and event history calendar

A chronology drawn together with the respondent, in which events are anchored to other events in their life or in the life of the company.

Research on event history calendars has shown that relying on temporal and thematic cues matches the way autobiographical memory is organized and, in a number of domains, improves the quality of retrospective reports compared with an ordinary question list. The effect is not universal, though: in an experimental comparison, the calendar improved reports of employment, income, and moves, but on some measures it increased the number of over-reported events.

An important property: a diagram drawn together is already a new research artifact. Its boundaries and ordering were shaped in the interaction rather than found in memory, and the timeline does not remove retrospective distortion.

Critical incidents

In Flanagan's original formulation, this is a set of procedures for collecting observations of behavior whose consequences are definite enough to judge its contribution to success or failure. "Critical" here does not mean dramatic: criticality is defined by the clarity of the consequences, not by the intensity of the emotion.

In service research the technique is applied to especially good and especially bad episodes, to recovery after a failure, and to the reasons for switching to another provider. A review of one hundred forty-one studies showed the method's wide adoption — and, along with it, considerable inconsistency in what counts as an incident.

The main price is a corpus skewed toward the memorable. Slow cumulative costs and mild recurring irritations will be underrepresented, even though in aggregate they may matter more than vivid failures.

The follow-up probe

A technique that gets underrated because it looks like the absence of technique.

Rubin and Rubin distinguish main questions, follow-up questions, and short probes that develop a topic already raised. The function of a probe is not to move on to something new but to raise specificity: an example, a sequence, the meaning of a word, a participant in the event, the basis of a comparison, a consequence.

There is a distinction here that determines the quality of all the material. Repeating the respondent's last word, staying silent, or asking them to continue introduces almost no content. Asking "Did that annoy you?" or "So it was about the price?" is simultaneously testing a hypothesis and injecting your own category. From then on, the respondent will use it as their own.

This yields a corollary that breaks the usual way interviews are judged: the depth of an answer does not by itself prove that a topic came up independently. What the person named unprompted must be distinguished from what appeared after a directed stimulus. The transcript should make that distinction visible.

A separate note on the "five whys." In an interview, this is not a self-sufficient method for finding a root cause. A repeated "why" can deepen an explanation, but it just as easily provokes rationalization and artificial linear causality: the person produces whatever explanations are available to them at the moment of the conversation. Probes about events, actions, comparisons, and consequences yield a different type of data than a request for an abstract cause.

Comparison and repertory grids

"How are these two suppliers similar to each other, and how do they differ from the third?"

Comparison shifts the task from naming properties in the abstract to distinguishing concrete objects. In Kelly's personal construct theory, meaning as such is organized through bipolar contrasts, and the repertory grid elicits those contrasts through triads.

The technique is appropriate when a ready-made list of criteria would impose the researcher's language: the respondent develops their own axes of comparison, which they would never have brought up in a free narrative.

Two limits. The form of the comparison affects the outcome: studies have directly compared triadic and dyadic elicitation and shown that the mode of comparison can change the set of constructs people formulate. And the result depends on what you put on the table: if some type of experience is missing from the alternatives being compared, the corresponding dimension simply never emerges.

Artifact, photograph, document

Packaging, a receipt, correspondence, a working document, a screen, an order history.

An object gives the conversation concreteness and holds the memory of routine details that never surface without a trace to prompt them. Harper describes the value of photography in terms of the image evoking recollections different from those that appear in response to a verbal question.

The limitation is substantial: an artifact is not a reflection of a need. It was created, kept, and selected under particular circumstances, and it usually represents the successful, official, or socially acceptable part of the practice. A procedure manual shows how things are supposed to be. A screen shows what is visible. Neither shows what a person does when things go off script.

On-site observation and the diary

Two techniques that are not interviews — and without which the interview cannot close its main blind spot.

The blind spot is the same across every technique listed above: routine. People will not talk about what they do every day and do not consider a problem — and that is usually where the bulk of the losses accumulates.

Observation in the real setting addresses this directly: you see the sequence of actions, the workarounds, the switching between systems, and what happens when exceptions arise. The logic of the contextual approach is that the researcher is present in the situation and clarifies as things unfold, rather than asking about it afterward. Observation has limits of its own: it is local, confined to the chosen period, and it changes the behavior of the person being observed.

A diary shortens the distance between the event and the record. The participant logs incidents as the week goes on, and the conversation is then built around their entries. It is the only way to reach the frequent minor episodes that, by the time of the interview, have blurred into "well, it's usually fine." The price is the burden on the participant and the incompleteness of the entries.

If the task reads "understand where we lose time and money in daily operations," the interview is secondary here. The core material will come from observation and the diary; the interview will explain what the observed material means.

How this comes together in a single project

Three things follow from everything above.

No single technique covers the task, and the choice of techniques is part of the design, not the interviewer's improvisation. The guide should state not only what to ask, but what elicits the answer and what type of data is expected.

The order of techniques changes the result. A category the interviewer introduces in minute ten will be present in the remaining fifty. So everything that sets a frame comes after the person has described the situation in their own words.

The practical takeaway for a client is simple. In a research proposal, the line "in-depth interviews" does not describe a method. Ask which techniques will be used to elicit answers and what type of data is expected from each — and you will see at once whether there is a design behind the proposal or merely a number of meetings.