{"site":{"name":"Koji","description":"AI-native customer research platform that helps teams conduct, analyze, and synthesize customer interviews at scale.","url":"https://www.koji.so","contentTypes":["blog","documentation"],"lastUpdated":"2026-07-05T12:22:26.350Z"},"content":[{"type":"blog","id":"920f34eb-5653-4a92-a614-d4fb10d1c254","slug":"in-depth-interviews-guide-2026","title":"In-Depth Interviews (IDI): The Complete Guide for 2026","url":"https://www.koji.so/blog/in-depth-interviews-guide-2026","summary":"In-depth interviews (IDIs) are one-on-one semi-structured conversations that uncover the why behind customer behavior. Teams typically need 15-30 interviews to reach saturation (20-40+ for heterogeneous audiences). Traditional IDIs are limited by human moderators who manage only 8-10 per day, but AI-moderated interviews now run hundreds in parallel at ~75% lower cost with automatic thematic analysis, giving interview depth at survey scale.","content":"**Quick answer:** An in-depth interview (IDI) is a one-on-one, semi-structured conversation designed to uncover the *why* behind a customer's behavior — their motivations, mental models, and decision triggers — in a way surveys and analytics never can. You typically need **15–30 interviews** to reach thematic saturation for a focused audience (and 20–40+ for a heterogeneous one). The catch has always been cost and time: a skilled human moderator can run only **8–10 IDIs per day**, and doubling your sample roughly doubles your budget. In 2026, AI-moderated interviews break that ceiling — running hundreds of IDIs in parallel, 24/7, with automatic analysis — so you get interview-grade depth at survey-grade scale.\n\n## What is an in-depth interview?\n\nAn in-depth interview (also called an IDI or one-on-one interview) is a qualitative research method where a single participant is guided through an open-ended conversation about a specific topic. Unlike a survey, an IDI is *adaptive*: the interviewer listens, then probes. When a participant says \"the checkout felt confusing,\" a good IDI doesn't move on — it asks \"confusing how? Walk me through exactly where you got stuck.\"\n\nIDIs sit at the depth end of the research spectrum:\n\n- **Surveys** capture *what* at scale, but stop at the first answer.\n- **Focus groups** surface *group* dynamics, but suffer from groupthink and dominant voices (see our [focus group guide](/blog/how-to-conduct-focus-group-2026)).\n- **In-depth interviews** capture the *individual why* — the richest signal, one person at a time.\n\nThey're the workhorse of customer discovery, [empathy interviews](/blog/empathy-interviews-complete-guide-2026), jobs-to-be-done research, win/loss analysis, and value-proposition testing.\n\n## When to use in-depth interviews\n\nReach for IDIs when you need depth over breadth:\n\n- **Early discovery** — you don't yet know the questions to put on a survey. IDIs help you find them. (Our [customer discovery guide](/blog/customer-discovery-interviews-startup-guide) covers the founder use case.)\n- **Sensitive or complex topics** — cancellations, pricing objections, workflow frustrations that people won't detail in a text box.\n- **Understanding a decision** — why a buyer chose you (or a competitor), the story behind a churn event, the trigger that made someone finally look for a solution.\n- **Explaining the \"what\"** — your analytics show a drop-off; IDIs tell you why it happens.\n\nIf you mostly need to quantify a known variable, a survey is cheaper. The strongest research programs combine both — see our [mixed methods guide](/blog/mixed-methods-research-guide-2026).\n\n## How many in-depth interviews do you need?\n\nThe honest answer: enough to reach **saturation** — the point where new interviews stop surfacing new themes. In practice:\n\n- **Homogeneous audience, focused question:** 15–30 interviews is the classic range, and saturation often appears around 12–20.\n- **Heterogeneous audience or multiple segments:** plan for **20–40+**, because each distinct segment needs its own path to saturation.\n\nA well-known caution from the qualitative literature: many published studies *claim* saturation they never actually reached, because the sample was too small or too narrow. The safeguard is to keep interviewing until two or three consecutive sessions add nothing new. For a deeper treatment of who to talk to, see [qualitative research sampling methods](/docs/qualitative-research-sampling-methods).\n\n## How to run a great in-depth interview\n\n**1. Write a discussion guide, not a script.** List 5–8 open-ended topics and let the conversation breathe. Use our [user interview script template](/docs/user-interview-script-template) as a starting point.\n\n**2. Open broad, then narrow.** Start with context (\"tell me about the last time you…\") before drilling into specifics. This grounds answers in real behavior rather than hypotheticals — the core lesson of [The Mom Test](/docs/mom-test-user-interviews).\n\n**3. Probe relentlessly.** The insight is almost never in the first answer. Master [probing questions](/docs/probing-questions-user-interviews) and the [laddering technique](/docs/laddering-technique-guide) to move from features to underlying motivations.\n\n**4. Avoid leading questions.** \"Wouldn't it be great if we added X?\" contaminates your data. Learn to spot and [avoid leading questions](/docs/avoiding-leading-questions).\n\n**5. Shut up and listen.** Aim for the participant talking 80% of the time. Silence is a tool — it invites elaboration.\n\n**6. Capture verbatims.** The quote *is* the deliverable. Record, transcribe, and tag every session so themes are traceable back to real words.\n\nCommon pitfalls — asking hypotheticals, pitching instead of listening, stopping at the first answer — are covered in [15 customer interview mistakes](/blog/customer-interview-mistakes-2026).\n\n## The scaling problem (and why 2026 changed it)\n\nHere's the structural weakness of traditional IDIs: **they don't scale linearly, they scale painfully.** A skilled moderator can run perhaps 8–10 interviews a day. Twenty interviews means two full days of moderation, plus scheduling, no-shows, transcription, and a week of manual synthesis. Doubling your sample doubles the cost and time. That's why so many teams settle for five interviews and call it \"research\" — and then over-generalize from a sample that never reached saturation.\n\nThe data on the alternative is striking. AI-led interviews now deliver insights up to **100× faster and at roughly 75% lower cost** than human-moderated sessions, while still capturing nuance. AI platforms can run **hundreds of interviews overnight** and return actionable insights within **24 hours**. A popular hybrid pattern: run **200 AI-moderated interviews** to map the landscape, then hand-pick 15–20 for human follow-up on the sharpest themes.\n\n## How Koji runs in-depth interviews at scale\n\nKoji is an AI-native research platform built to run true IDIs — not glorified surveys. You write your discussion guide once, and Koji's **AI moderator conducts voice or text interviews** with every participant, probing each answer in real time exactly like a skilled human researcher — asking \"compared to what?\", \"can you give a specific example?\", and \"why did that matter?\" automatically, with **no moderator bias**, no scheduling, and no fatigue on interview #200.\n\nBecause Koji supports **six structured question types** — open_ended, scale, single_choice, multiple_choice, ranking, and yes_no — you can blend the depth of open-ended probing with quantified measures (a satisfaction scale, a ranked feature list) inside the same study. Then Koji does the week-long part instantly: **automatic thematic analysis** clusters every transcript into themes, sentiment scoring flags the strongest reactions, and a **one-click report** hands stakeholders the verbatim quotes plus the distribution behind them.\n\nThe result is the thing IDIs could never offer before: **interview depth at survey scale**, from question to insight in hours, not weeks — with no research expertise required. Whether you need unmoderated reach or moderated depth, you no longer have to choose (see [moderated vs unmoderated research](/blog/moderated-vs-unmoderated-research-2026)).\n\n## Run your first in-depth interview study\n\nDraft a 5–8 topic discussion guide, then let an AI moderator conduct — and probe — every conversation for you. **[Launch your first AI in-depth interview study on Koji](https://www.koji.so)** and reach saturation in days, not months.","category":"Research","lastModified":"2026-07-04T03:16:58.334475+00:00","metaTitle":"In-Depth Interviews (IDI): The Complete 2026 Guide + Examples","metaDescription":"Everything you need to run in-depth interviews (IDIs) in 2026: what they are, when to use them, how many you need for saturation, best practices, and how to run IDIs at scale with AI moderation.","keywords":["in-depth interviews","IDI","in-depth interview guide","qualitative interviews","IDI research","one-on-one interviews","in-depth interview questions","how many interviews saturation"],"aiSummary":"In-depth interviews (IDIs) are one-on-one semi-structured conversations that uncover the why behind customer behavior. Teams typically need 15-30 interviews to reach saturation (20-40+ for heterogeneous audiences). Traditional IDIs are limited by human moderators who manage only 8-10 per day, but AI-moderated interviews now run hundreds in parallel at ~75% lower cost with automatic thematic analysis, giving interview depth at survey scale.","aiKeywords":["in-depth interviews","IDI","qualitative research","saturation","AI-moderated interviews","customer discovery","user research"],"aiContentType":"guide","faqItems":[{"answer":"An in-depth interview is a one-on-one, semi-structured qualitative conversation designed to uncover the motivations, mental models, and decision triggers behind a person's behavior. Unlike a survey, it is adaptive: the interviewer listens and probes each answer, moving past the first response to reach the underlying why.","question":"What is an in-depth interview (IDI)?"},{"answer":"For a focused, homogeneous audience, 15-30 interviews is the classic range, with saturation often appearing around 12-20. For a heterogeneous audience or multiple segments, plan for 20-40+, because each distinct segment needs its own path to saturation. Keep interviewing until two or three consecutive sessions surface nothing new.","question":"How many in-depth interviews do you need?"},{"answer":"A survey captures what at scale but stops at the first answer. An in-depth interview captures the individual why through adaptive, real-time probing. AI-moderated platforms like Koji now combine both: they run interviews at survey scale, giving you the depth of an IDI across hundreds of participants.","question":"What is the difference between an in-depth interview and a survey?"},{"answer":"Most IDIs run 30-60 minutes. Shorter sessions (15-30 minutes) work for a single focused topic, while complex discovery or win/loss conversations may run up to an hour. AI-moderated interviews let participants respond on their own schedule, which reduces drop-off and captures richer answers than a rushed live call.","question":"How long should an in-depth interview be?"},{"answer":"Yes. AI moderators ask your discussion-guide questions and probe each answer in real time with no moderator bias, no scheduling, and no fatigue. Research shows AI-led interviews can deliver insights up to 100x faster and roughly 75% cheaper than human-moderated sessions. Koji then clusters every transcript into themes and produces a one-click report automatically.","question":"Can AI moderate in-depth interviews reliably?"},{"answer":"Use IDIs when you need the individual why without group influence — for sensitive topics, personal decisions, or detailed workflows. Focus groups surface group dynamics but suffer from groupthink and dominant voices. IDIs give every participant equal space to tell their full story.","question":"When should you use in-depth interviews instead of focus groups?"}],"relatedTopics":["in-depth interviews","qualitative research","customer discovery","AI-moderated interviews","user research methods"]}],"pagination":{"total":1,"returned":1,"offset":0}}