On this page
This user research glossary covers the 32 most common terms in user research and AI interviews. Each entry gives the term and a definition, and the first sentence of each definition is a complete explanation you can quote on its own. Terms are grouped into five sections: research methods, interviews and guides, analysis and reporting, metrics and data, and AI interviews.
Method
User interview
A user interview is a qualitative research method in which a researcher talks one-on-one with a user to understand their experiences, motivations, and needs. It's good at answering “why” and “how did it happen,” but not at estimating proportions. Common formats include in-person, phone, video, and text interviews.
In-depth interview
An in-depth interview is a semi-structured interview that focuses on a few topics and uses sustained follow-up questions to understand a participant's experiences and thinking in depth. Sessions usually run longer than typical interviews. The guide sets only the topics and goals, and specific questions adapt to the answers. It suits complex questions such as motivations and decision-making processes.
Usability testing
Usability testing is a research method in which users complete set tasks while you observe how they go about it, in order to find problems with using the product. Researchers watch where users get stuck, what errors they make, and how long tasks take. Nielsen (2000) proposed that about 5 users per round of usability testing will uncover most problems; this applies only to usability testing.
Concept testing
Concept testing is a research method in which you show a product concept to target users before building the product or feature, to learn how well they understand it and how receptive they are. The stimulus can be a written description, a sketch, or a prototype. The focus is on whether users understand it, whether they need it, and in what situations they would use it.
Focus group
A focus group is a qualitative research method in which one moderator leads several participants in a group discussion on the same topic. It's useful for learning how a group thinks and talks about something, but a few people can easily dominate while others just go along. For personal experiences and sensitive topics, one-on-one interviews are usually a better fit.
Survey
A survey is a quantitative research method that uses a fixed set of questions to collect responses from a larger number of people in order to measure counts and proportions. It can answer “how many,” but it struggles to uncover reasons. Surveys often pair with interviews: interview first to find the types of reasons, then survey to estimate how common each one is.
Jobs to be Done
Jobs to be Done is a framework for understanding user needs that holds that people “hire” a product to get a job done. Closely associated with Clayton Christensen's work, it focuses on the progress users are trying to make rather than on their demographics. For example, someone buying a drill may really be trying to hang a picture on the wall.
Interviews and guides
Study plan
A research plan is a document that explains why a study is being done, what it covers, and how it will be carried out. It usually includes the business question, the decision to be made, research questions, participant criteria, sample size, timeline, and deliverables. If you can't state the decision to be made, the study often isn't ready to start.
Interview guide
An interview guide is a document that sets which topics an interview will cover, along with each topic's research goal and main questions. A good guide is organized by topic and includes follow-up directions, rather than a long list of questions read in order. In AI-moderated interviews, the guide is also what the AI uses to decide when to follow up and when to wrap up.
Follow-up
A follow-up question, or probe, is an interviewing technique in which you ask more based on a participant's previous answer to turn a vague statement into a specific one. Common directions include asking about one specific instance, the steps at the time, what they originally expected, and what they did afterward. For example, if a participant says “it's not very easy to use,” you might follow up with “What were you doing the last time it felt hard to use?”
Leading question
A leading question is one that hints at an answer or presumes a position, nudging participants to answer the way the interviewer expects. For example: “Did you stop using it because it was too expensive?” Rephrase it neutrally instead, such as “You stopped using it after a while. What was the reason?”
Open-ended question
An open-ended question offers no answer options and lets participants respond in their own words. It can surface unexpected information and is the main type of question used in interviews. A closed-ended question, by contrast, can only be answered yes or no, or by choosing from options.
Screener
A screener is a short questionnaire used before the interviews to determine whether applicants meet the participant criteria. Its questions should avoid revealing the purpose of the study so applicants don't tailor their answers to qualify. It usually also screens out people who work in the same industry and those who have recently taken part in interviews.
Participant incentives
A participant incentive is a reward, such as cash, a gift card, or account perks, given to thank participants for their time. Incentives should match the interview length and the audience: too low and recruiting is hard, too high and you may attract people who are only there for the money. Explain when and how the incentive will be paid in the invitation.
Analysis and reports
Coding
Coding is a qualitative analysis method in which you tag interview quotes with themes so they can be grouped and compared. The same quote can carry multiple codes, and the codebook keeps getting merged and adjusted during analysis. Once coding is done, you can count how many participants mentioned each theme.
Thematic saturation
Thematic saturation is the point at which adding more interviews yields few new themes, and it's commonly used to judge whether a sample is large enough. Guest, Bunce & Johnson (2006) found that in a fairly homogeneous group, few new themes emerge after about 12 interviews. When participants vary widely, assess it separately for each segment.
Affinity diagram
An affinity diagram is a synthesis method in which you write many scattered observations and quotes on cards and group them by how they relate to arrive at themes. Also called the KJ method, it was developed by Japanese scholar Jiro Kawakita. It works well for teams sorting through interview material together. Groups should emerge bottom-up from the quotes, not from categories you set in advance.
Insight
An insight is a discovery drawn from research material that explains user behavior and informs decisions. It doesn't just restate what users said; it explains the reason behind it and what it means for the product. A reliable insight traces back to quotes and states its participant count and counterexamples.
Say-do gap
The say-do gap is the mismatch between the views or intentions participants express and what they actually do. For example, a participant says they'd pay for a feature but never buys it. That's why interviews should ask more about what specifically happened in the past and less about what people would do in the future, and check against behavior data when needed. Rob Fitzpatrick discusses this problem in depth in The Mom Test (2013).
Counterexample
A counterexample is a participant statement or behavior that doesn't fit a given finding. Listing counterexamples helps readers judge how far a finding applies and avoid overgeneralizing. If counterexamples cluster in one segment, the finding often applies only to some users.
Triangulation
Triangulation is the practice of using multiple sources or methods to corroborate the same finding. For example, you might compare interview findings with survey data, behavior data, and support tickets. When several sources point to the same finding, it's more credible; when they conflict, more research is needed.
Persona
A persona is a description of a typical user, synthesized from research, that summarizes a type of user's goals, behaviors, contexts, and pain points. Good personas come from interviews and data, not from the team's imagination. Their purpose is to help the team remember exactly who they're designing for when they make decisions.
Metrics and data
Net Promoter Score (NPS)
Net Promoter Score (NPS) is a metric that measures how willing users are to recommend a product, using the question “How likely are you to recommend us to a friend or colleague?” It was introduced by Fred Reichheld in 2003. Respondents give a score from 0 to 10: 9–10 are promoters and 0–6 are detractors. NPS equals the percentage of promoters minus the percentage of detractors and ranges from -100 to 100. The score alone doesn't explain why, so the reasons behind low scores need to be explored with follow-up questions in interviews.
Customer Satisfaction Score (CSAT)
Customer Satisfaction Score (CSAT) is a metric that measures how satisfied users are with a specific service interaction or experience. A common approach is to ask users to rate it on a 1–5 scale and calculate the percentage who choose 4 or 5. It's mostly used for single interactions, such as a support conversation or a purchase.
Retention
Retention describes whether users keep using a product for a period after their first use, and it's usually expressed as a retention rate. For example, day-1 retention is the percentage of users who used the product on the first day and came back on the second. Retention data tells you how many people stay; why they stay or leave takes interviews to answer.
Churn
Churn is when users stop using a product or cancel their subscription, usually measured as the percentage of users lost over a period of time. How churn is defined varies by product, for example no login for 30 consecutive days, or a subscription that expires without renewal. Interviewing churned users is the most direct way to learn why they left.
Behavior data
Behavior data is data that records what users actually do in a product, such as which pages they visit, what they click, and when they pay. It reflects user behavior objectively but doesn't capture reasons. Used alongside interviews, it lets you select participants by behavior and test interview findings with data.
AI interviews
AI-moderated interview
An AI-moderated interview is an interview in which AI talks one-on-one with a participant following an interview guide, asks follow-up questions based on their answers, and synthesizes findings. It lets many interviews run at the same time while keeping follow-up questions in every session. Defining the research question, reviewing the guide, and reviewing findings still fall to the research lead.
Informed consent
Informed consent is the process by which participants voluntarily agree to take part in research after learning about its purpose, how it's recorded, how the data will be used, and their rights. In AI-moderated interviews, participants should also be told they're talking to AI. Consent should be obtained before the interview starts, and a record kept.
Anonymization
Anonymization is the process of removing or replacing information in research materials that could identify participants. Common practices include deleting names, phone numbers, email addresses, and street addresses, and replacing real identities with participant IDs. Quotes cited in reports should be anonymized first.
Quote traceability
Quote traceability is the ability to trace every finding in a report back to the participant quotes that support it and where they appear in the interview. It lets readers verify that findings are accurate and reveals whether AI or researchers have overgeneralized. It's one of the key things to check when evaluating AI research tools.
Invalid response
An invalid response is an interview that shouldn't be included in analysis, for example one that is off-topic, clearly low-effort, from someone who doesn't meet the participant criteria, or from a person taking part more than once. Excluding invalid responses keeps low-quality answers from skewing findings. Record the reason for each exclusion so it can be reviewed later.
How to do it in heari
heari's workflow covers most of the terms here: AI generates a research plan and interview guide from a business question, and the research lead reviews it before publishing; AI moderates interviews by text or voice and asks follow-up questions based on answers, while researchers can listen in and exclude invalid responses; every finding in the insight report comes with quotes and counterexamples and is labeled with participant counts. Connected to ThinkingAI Agentic Engine, heari lets you select participants by behavior data and validate findings with data.
FAQ
What's the difference between user interviews and surveys?
User interviews are one-on-one conversations that are good at uncovering specific experiences and the reasons behind user behavior; surveys use fixed questions to collect many responses and are good at measuring counts and proportions. Studies often start with interviews to identify the types of reasons, then use a survey to estimate how common each one is.
What is thematic saturation, and how many interviews does it take to reach it?
Thematic saturation is the point at which additional interviews yield few new themes. Guest, Bunce & Johnson (2006) found that in a homogeneous group, few new themes emerge after about 12 interviews. If your study includes several clearly different segments, assess saturation for each one separately.
How is NPS calculated?
NPS asks users to rate how likely they are to recommend you on a 0–10 scale. People who answer 9–10 are promoters and those who answer 0–6 are detractors. Subtract the percentage of detractors from the percentage of promoters to get NPS, which ranges from -100 to 100. Passives, who answer 7–8, don't factor directly into the calculation.
What's the difference between an insight and a finding?
In user research, a finding is usually a statement of what the research found, such as which step most participants struggled with; an insight goes further to explain the reason behind it and what it means for product decisions. Both should include quotes, participant counts, and counterexamples so the team can verify and trust them.
What is a leading question, and how do you avoid it?
A leading question hints at the answer, such as “Was it because it's too expensive?” To avoid one, ask only what happened and why, without offering possible reasons in the question. For example, ask instead, “You didn't keep using it. What was the reason?”
