Shuichiro Ogawa
日本語

Notes · updated 2026-11-03 · Markdown

A Checklist for Group Interviews: What to Prepare, What to Ask, and How to Report to the Client

This note divides group interviews (focus groups, round-table discussions) with people who use or buy something into stages, from planning through recruitment, confidentiality among participants, incentives, moderation, venue and observers, records, and analysis and reporting, and lists what to prepare and confirm in a…

Contents (11)
  1. How does a group interview differ from a one-to-one interview?
  2. How many groups, and how many people in each?
  3. Who should be in the same room?
  4. How should confidentiality among participants be explained?
  5. How much should the incentive be, and how should it be paid?
  6. What should the discussion guide contain, and who should moderate?
  7. How should the venue and observers be handled?
  8. How should records and data be handled?
  9. What should be counted in analysis and reporting, and how?
  10. What should be added when a screen operation check is included?
  11. How strong is the evidence for this list?

How does a group interview differ from a one-to-one interview?

A group interview (focus group, round-table discussion) brings together several people who use or buy something, has them discuss a topic under a moderator, and uses the discussion to inform decisions about what to make. Almost everything prepared for one-to-one interviews (A Checklist for User Interviews: What to Prepare, What to Ask, and How to Report to the Client) is still needed. The difference is that the data arise not from one person’s remarks but from interaction among participants. Kitzinger (1994, abstract) identified the interaction between research participants as the one feature that distinguishes focus groups from one-to-one interviews and questionnaires, and argued for the overt exploration and use of that interaction.

Interaction brings both gains and losses. In a comparison that randomized 350 African American men to focus groups or individual interviews, the two produced similar numbers of unique items per data collection event, but individual interviews produced a broader range of items per person (Guest et al. 2017, IJSRM, abstract). On the other hand, several types of sensitive and personal disclosure were more likely in the focus group setting, and some sensitive themes occurred only there. In a comparison of 55 water decision makers, for very sensitive topics participants volunteered less in groups when there seemed no compelling reason to risk stigma from other members, and more when there was an opportunity to exchange important information (Wutich et al. 2010, abstract). Whether to choose a group depends on which of these situations the questions resemble.

A group discussion cannot show whether a screen is easy to use. NN/g states that in focus groups people comment on what they are shown or remember, so focus groups cannot provide objective information about usability (Fessenden 2022). Nielsen (1997) also writes that watching a demo is fundamentally different from actually using the product. What to add when checking screen operation is summarized in the final section.

This note divides the work from receiving a commission to reporting to the client into stages, and lists what to prepare and confirm in addition to what one-to-one interviews require. The evidence comes from 18 academic works, 16 documents from public bodies, standards bodies, and industry codes, 24 materials from research firms and trade journals, and 9 first-hand accounts by practitioners. Numbers and rules were checked against the relevant passages of the full texts wherever these could be obtained, and academic works whose full texts could not be obtained are marked “abstract” after the year and were used only within the scope of the abstract. Industry codes are self-regulation imposed on members, not laws.

How many groups, and how many people in each?

There is empirical evidence on the number of groups. In an analysis of 40 focus groups on health-seeking behavior, more than 80% of all themes were discoverable within two to three groups and 90% within three to six, and three groups were enough to identify all of the most prevalent themes (Guest et al. 2017, Field Methods, abstract). Hennink et al. (2019, abstract) analyzed 10 groups and reported that four groups were sufficient for code saturation, the point at which no new issues appear, but that more groups were needed for meaning saturation, a full understanding of those issues. When groups were stratified, one group per stratum identified the issues, two groups per stratum gave a more comprehensive understanding, and more groups per stratum added little. Yet among 220 health focus group studies, only 37 (17%) attempted to explain the number of groups (Carlsen and Glenton 2011).

Guidance varies on the number of people per group. NN/g gives 6 to 9 people for 1 to 2 hours (Fessenden 2022), and the User Interviews guide gives 5 to 10 per group. Japanese research firms give 4 to 8 per session (Macromill, Cross Marketing), and an information page for participants describes round-table discussions as having at most 8 people (Intage). For online groups, the JMRA guide recommends a number the moderator can keep an eye on individually, and gives three to four people as a tip for seeing facial expressions easily (Online Qualitative Research First Step Guide). GOV.UK treats small group workshops of 3 to 8 people as different from focus groups, which measure opinions and attitudes; workshops are for finding out what works for users.

How to compose groups depends on what is to be compared. The service design book gives 4 to 12 participants (6 to 8 is often regarded as ideal), notes that focus groups often aim for homogeneity to maximize disclosure, and recommends at least a second group as a control (Stickdorn et al., This Is Service Design Doing, “Focus Groups”). A Japanese industry code requires the proposal to the client to state the group design and the number of people per group, the eligibility conditions, the date and time, the venue, and the duration (JMRA, Sales and Planning).

  • Wrote down why a group is needed (opinions arising from interaction, themes emerging in discussion) and checked whether one-to-one interviews would suffice (Kitzinger 1994 abstract, Guest et al. 2017 IJSRM abstract)
  • For sensitive topics, checked whether participants have a reason to talk in a group (Wutich et al. 2010 abstract)
  • If screen usability is to be checked, planned a separate step instead of a group (NN/g)
  • Defined strata (combinations of attributes) and planned at least two groups per stratum (Hennink et al. 2019 abstract)
  • Wrote down the number of groups and the reason (Guest et al. 2017 Field Methods abstract, Carlsen and Glenton 2011)
  • Set the number of people and the length of each group according to whether it is in person or online (NN/g, JMRA online guide)
  • Stated in the proposal the group design and size, eligibility conditions, date and time, venue, and duration (JMRA)

Who should be in the same room?

In a group, who sits with whom also shapes the data. The JMRA guideline requires care that participants do not know each other, with an exception only when acquaintance is judged useful and the client agrees (Fieldwork Management 1.1). The Government of Canada standards for public opinion research state that participants acquainted with each other may not be knowingly recruited for the same study unless they are in separate sessions. The standards also exclude anyone who attended a qualitative session within the past six months or five or more sessions in the past five years. The MRS guidance for recruiters asks recruiters to track past participation and use fresh participants, and calls pre-check calls an essential part of recruitment. Online recruitment needs further care. The same guidance asks that detailed specifications not be included when recruiting through social media, to avoid the temptation to answer to fit them. In a study that recruited parents online for focus groups, multiple submissions from the same IP address or email were common, and when the incentive was removed and only one submission per IP address was allowed, the proportion from the same IP addresses fell from 47.1% to 7.9% (Quach et al. 2013).

Recruitment should also assume that some people will not come. In an NN/g survey of UX practitioners, the average no-show rate was 11%, one in nine users (Nielsen 2003). A research firm practitioner writes that an ample over-recruit is 10 for eight to show for a responsible population, 10 for six for a busy population, and 10 for five for an irresponsible one (Schild 2003). People turned away because of over-recruiting still receive the incentive promised at recruitment (Government of Canada standards).

  • Confirmed at recruitment and at reception on the day that participants do not know each other (JMRA, Government of Canada standards)
  • Set exclusion conditions for related occupations, recent participation in qualitative research, and frequent participation (Government of Canada standards, MRS)
  • Did not put detailed eligibility specifications in social media recruitment posts (MRS)
  • For online recruitment, set a procedure to check for multiple submissions from the same IP address or email (Quach et al. 2013)
  • Over-recruited to allow for no-shows, and told people at recruitment how those turned away and late arrivals will be paid (Government of Canada standards, Schild 2003)
  • Checked in the screener for a balance of introverts and extroverts (User Interviews)

How should confidentiality among participants be explained?

In a group, the interviewer cannot promise that what is said to other participants will stay in the room. The guidance of the UC Berkeley Committee for Protection of Human Subjects states that investigators cannot guarantee that fellow participants will keep information confidential. It asks that a request not to repeat what is heard outside the group be included both in the consent form and in the instructions read before the group begins. The guideline of four German market research associations states that participants should not be addressed by their last names in the group and should be told at the beginning not to introduce themselves by last name. The JMRA online guide asks that, from the recruitment stage, participants agree not to share the URL with third parties and not to record audio, video, or screen captures, and that confidentiality be reconfirmed with everyone before the start. If unreleased products are shown, the client should understand the risk of leaks.

  • Wrote in the information sheet and consent form a request not to repeat other participants’ remarks outside, and that the interviewer cannot guarantee this (UC Berkeley CPHS)
  • Decided not to address participants by last name and not to show full names as display names (German associations’ guideline, JMRA online guide)
  • For online groups, obtained agreement not to share the URL, not to record or capture the screen, and to join from a place without other people (JMRA online guide, UC Berkeley CPHS)
  • When showing unreleased products or concepts, agreed with the client on the risk of leaks and the countermeasures (JMRA online guide)
  • Decided what to do if someone declines to be recorded (UC Berkeley CPHS)

How much should the incentive be, and how should it be paid?

Incentives are handled much as in one-to-one interviews, with some points specific to groups. A meta-research study of 80 health focus group studies from 2018 to 2024 found that a financial incentive was used in 43% of cases (Lander et al. 2025). The Government of Canada standards state that participants in qualitative sessions are typically offered incentives, and require that people turned away because of over-recruiting receive the incentive described at recruitment. The amount that a Japanese research firm shows to participants is around 8,000 yen for one to two hours (Intage, the same source as in A Checklist for User Interviews: What to Prepare, What to Ask, and How to Report to the Client).

  • Decided the amount, and told people at recruitment how those turned away and late arrivals will be paid (Government of Canada standards)
  • Decided whether the incentive covers burdens of participation such as travel to the venue (Government of Canada standards)

What should the discussion guide contain, and who should moderate?

Group moderation carries the risk that one person takes over. Eskenazi, a former director of UX research at Expedia, describes how a loud participant can lead the group in place of the moderator, and describes online formats in which participants must type their response before they see others’ responses. The service design book recommends that the moderator prevent individual participants from dominating, engage the quiet ones, and make clear that the group is not about finding a consensus but about understanding different perspectives. A research firm practitioner writes that friends recruited by accident tend to sit together and talk privately, which can be avoided by assigning seats with name cards, and that a participant who says nothing can be given a task or paired with another member (Murphy 2001).

Who moderates also changes the data. Pande, head of research at Cisco Meraki, says that when researchers were not doing the moderation, the data were not good (Portigal’s podcast). The JMRA guideline requires the moderator to be familiar with the research purpose, stimulus materials, and projective techniques before fieldwork, and allows people who have received appropriate training or know the project theme well to moderate with the client’s consent (Fieldwork Management 1.4). A study that interviewed nine researchers who had conducted focus groups in health research pointed to giving patient groups sufficient time to introduce themselves and using participant-generated cards and prioritization (Tausch and Menold 2016).

  • Wrote an opening line saying that the session is for hearing different views, not for reaching agreement (This Is Service Design Doing)
  • Decided how to handle loud and silent participants (individual input before seeing others, assigned seats with name cards, how to call on people) (Eskenazi, Murphy 2001)
  • Chose a moderator familiar with the research purpose and stimulus materials, with the client’s consent (JMRA)
  • Allowed sufficient time for self-introductions (Tausch and Menold 2016)
  • Assigned a note-taker to draw a seating chart and record group dynamics, body language, and other things not captured in the audio (User Interviews)

How should the venue and observers be handled?

The venue should be inspected in advance. The JMRA guideline requires a venue used for the first time to be inspected before use and the result recorded in writing (Fieldwork Management 1.2). If the client wants to observe, the arrangements for that are also part of the preparation. The JMRA request on observation asks that the wish to observe be communicated before recruitment begins, and that anyone who knows a participant stop observing. The guideline of the German associations requires observers to sign a declaration of commitment in advance and observation to take place in closed rooms to which only those persons have access. The MRS qualitative research guidelines (2011 edition) recommend that ideally no more than half of the groups be observed, that there ideally be only one observer per group, and that observers be told that noise or light from behind the mirror can damage group dynamics.

Observers also have their own etiquette. Research firm practitioners write that observers should arrive 45 minutes early to avoid running into participants (Henderson 1990). They should also read the guide, write down the three to five most important things they want to learn, and hold a brief debrief after each group and a longer one at the end of each day (Greenbaum 1995). A trade journal article criticizes clients who attend one group, hear one compelling story, and think they have the answer (Cooper 2010). The presence of observers can itself affect participants. In an experiment with screen tasks, the presence of two non-interactive observers led to a physiological stress response, decreased performance on some measures, and affected participants’ emotional state (Sonderegger and Sauer 2009, abstract).

Online groups need communications instead of a venue. The JMRA online guide recommends choosing a stable conferencing system, keeping client conversations out of participants’ view, dividing support into participant support, technical support, and client support, and running connection tests at least three days before. A study comparing in-person, video, and chat groups found that virtual groups offered minimal cost savings and did not recruit faster, show rates were 94% in person, 81% for video, and 69% for chat, and virtual participants were more geographically diverse (Rupert et al. 2017). The content did not differ greatly, and in one comparison sensitive topics were discussed more candidly online (Woodyatt et al. 2016, abstract).

  • Inspected a new venue in advance and recorded the result (JMRA)
  • Asked about the client’s wish to observe before recruitment began, and decided the number of observers and the declaration they sign (JMRA, German associations’ guideline, MRS)
  • Told observers to arrive early, stay quiet, step away if they know a participant, and narrow down what they want to learn (Henderson 1990, Greenbaum 1995, JMRA)
  • Arranged for the client to watch several groups so as not to conclude from one (Cooper 2010)
  • Scheduled debriefs between observers and the moderator after each group and at the end of the day (Greenbaum 1995)
  • For online groups, decided the conferencing system, the back channel, the division of roles, and connection tests (JMRA online guide)
  • Prepared a seating chart and a record of speakers so that voices can be told apart later (User Interviews)

How should records and data be handled?

Group recordings show several participants at once. The MRS guidelines (2011 edition) treat recordings of voice and appearance as personal data, and require that material handed to clients or included in reports without participants’ consent, including verbatim transcripts, be anonymized. The German associations’ guideline states that recordings must be deleted no later than three months after the client receives them. The Government of Canada standards require records to be retained for three years, and allow primary records to be destroyed after 12 months when secondary records are sufficient. Because the sources set retention differently, it should be fixed in the contract with the client.

What should be counted in analysis and reporting, and how?

Decide first whether the unit of analysis is the group or the individual. Morgan (2010, abstract) argues that although interaction is essential for producing focus group data, the analysis and reporting need not emphasize interaction, and that the research goals should decide. Numbers are hard to handle in reports. The Government of Canada standards state that qualitative results cannot be statistically projected to the broader population, and, to avoid any implication that they can, require that reports not include percentages or precise proportions. They also require the proposal to state the criteria and rationale for homogeneous or heterogeneous group composition. A collection of tips from qualitative research practitioners recommends analyzing first the items that relate directly to the study purpose and leaving more time for analysis than expected (Fuller 2000).

  • Decided from the goals whether the unit of analysis is the group or the individual, and whether to analyze interaction (Morgan 2010 abstract)
  • Left percentages and precise proportions out of the report, and stated that the results cannot be projected to the population (Government of Canada standards)
  • Wrote down the number of groups, their size, and the rationale for their composition (homogeneous or heterogeneous) (Government of Canada standards, Carlsen and Glenton 2011)
  • Distinguished one participant’s strong remark from themes that appeared across several groups (Cooper 2010)

What should be added when a screen operation check is included?

An interview may include a step in which participants operate a prototype or app and the team observes how they do it. In that case, preparation for listening must be supplemented with preparation of the thing to be operated, tasks, records, and evaluation.

When a group operates the screen, the format should be chosen first. In a group, individuals rarely get the chance to explore the product on their own (Nielsen 1997). Operation checks are therefore either run separately, one to one, before or after the group, or adapted to the group. A study that compared two people operating together while talking (constructive interaction) with individual thinking aloud found that the methods revealed similar numbers and types of problems, with only a few differences in how problems were detected, in task performance, and in participant experience (van den Haak et al. 2004, abstract). Nielsen (2007) describes having 5 to 10 people operate separate machines at the same time, and writes that this usually needs many facilitators and that the usual thinking-aloud instruction should not be given, because thinking aloud does not work in tightly packed cubicles. An article from Wales describes a format in which one member of the group uses the website and talks through what they are doing, while everyone else comments silently on sticky notes or in the chat (Quatermass, ProMo Cymru).

The thing to be operated should be prepared so that the session does not stop when it breaks. GDS practitioners advise against Lorem Ipsum and funny or lazy names in prototypes that users will see, suggest starting to use real data as the prototype is refined, and suggest keeping alternative versions to go back to when a route does not work (Lanman et al. 2015). NN/g writes that an interactive prototype needs a response set for each possible user action before the test, while with a static prototype a person familiar with the design gives the responses in real time, for example in Wizard of Oz testing (Pernice 2016).

Data and accounts should be planned so that participants do not have to use their own information. NN/g notes that qualitative studies can benefit from participants’ own information, recommends fake but realistic information as an alternative, and asks that personal information be blurred in highlight videos and in screenshots in reports (Schade 2016). MeasuringU notes that it is difficult, and often prohibited, to ask users to use their own information, and lists alternatives such as stopping before the purchase and providing prepaid cards (Sauro 2014). The UK firm dxw, testing a service that handles sensitive personal data, risked showing private data to the wrong people when only real data were available, and generated fake but realistic data sets instead. GOV.UK recommends asking participants to use their own data and documents whenever possible.

Tasks should be written so that they do not give away the answer. NN/g recommends task scenarios that are realistic, encourage an action, and do not give away how the interface should be used (McCloskey 2014). A reanalysis of nine commercial teams found no significant correlation between the percentage of problems found and the number of test users, but a significant correlation with the number of tasks used (Lindgaard and Chattratichart 2007, abstract). A researcher at note inc. writes that she always runs a rehearsal, walking through the tasks herself, once the tasks are designed (Senda 2022).

Thinking aloud needs an explanation, practice, and a decision on how to intervene. NN/g notes that most users need prompting, that a short demo video can show what is expected, and that prompts and clarifying questions can bias behavior (Nielsen 2012). Boren and Ramey (2000, abstract) pointed out that thinking aloud as practiced in usability work does not conform to its usual theoretical basis (Ericsson and Simon). In a meta-analysis of 29 studies comparing thinking aloud during and after use, Hertzum (2024, abstract) found that concurrent thinking aloud makes task time longer but total time shorter, while retrospective thinking aloud produces more explanations, problem formulations, and design recommendations.

Screen recordings and personal information need more care than recordings of talk. GOV.UK writes that screenshots from usability tests used outside the research should either contain no personal data or have any personal data blurred. The U.S. GSA guide tells the facilitator to ask participants, before they share their screen, to close anything they do not want recorded. The guidelines of Japan’s Personal Information Protection Commission give video in which a person can be identified, and audio recordings that identify a person because they contain a name, as examples of personal information. For unreleased products, a non-disclosure agreement should also be prepared. The NIST template includes a non-disclosure agreement as a discretionary appendix (NISTIR 7742).

Remote operation checks should separate the technology check from the session. Researchers at the U.S. Census Bureau report that when they emailed instructions for downloading and installing software, a participant withdrew because it seemed too complicated, so they changed the email to announce a pre-meeting “tech check” that would step participants through the process (Olmsted-Hawala and Nichols 2021). The same report says that they embedded consent, satisfaction, and debriefing into one URL, so that participants did not have to open their email while being recorded. NN/g recommends, where possible, a 15-minute online practice meeting with each participant the day before the session (Moran and Pernice 2020). An adult social care team in Wales found that video was a barrier to inclusion for people less familiar with technology, and switched to telephone sessions. Participants who use assistive technology should be asked to bring their own, because it is hard to recreate someone’s personal setup (GOV.UK).

Observers affect participants particularly in operation checks. The presence of non-interactive observers led to a physiological stress response and lowered some performance measures (Sonderegger and Sauer 2009, abstract). A team at Tabelog shares its Zoom session with colleagues through a separate Teams stream, so that the participant does not face a large audience (Kusaka 2023).

Numbers and evaluation depend on the purpose. The rule that five users find about 85% of problems assumes problems that affect 31% of users, and applies only to discovering problems, not to comparing products or estimating completion rates (Sauro 2010, Nielsen 2000). Problem severity is recommended to be rated from frequency, impact, and persistence, with each evaluator rating independently (Nielsen 1994). Even with the same method and the same interface, the average agreement between two evaluators ranges only from 5% to 65% (Hertzum and Jacobsen 2001, abstract). The NIST template lists the number of tasks completed successfully within the allotted time without assistance as one of the data to collect (NISTIR 7742).

  • Decided whether to have the group operate the screen or to run separate one-to-one operation checks (Nielsen 1997, Nielsen 2007)
  • Made the wording and names in the prototype realistic, and kept alternative versions to fall back on (Lanman et al. 2015)
  • Decided the prototype’s responses (clickable areas, responses given by a person) and walked through it (Pernice 2016, Senda 2022)
  • Prepared test accounts and dummy data (realistic names and values), and decided how to avoid using participants’ own payment or personal details (Schade 2016, Sauro 2014, dxw)
  • Wrote tasks set in realistic scenarios that do not give away the answer, and set success criteria and time limits (McCloskey 2014, NISTIR 7742)
  • Put an explanation of thinking aloud and practice time into the script, and decided how to intervene (Nielsen 2012)
  • Made screen recording a separate item on the consent form, and arranged for participants to close notifications and personal files before sharing their screen (digital.gov, GOV.UK)
  • Decided how to blur personal information on screen in reports and highlight videos (GOV.UK, Schade 2016)
  • For unreleased products, prepared a non-disclosure agreement (NISTIR 7742)
  • For remote sessions, set up a separate short technology check and a fallback to the telephone (Olmsted-Hawala and Nichols 2021, Moran and Pernice 2020, the Welsh adult social care team)
  • Arranged for users of assistive technology to take part with their own equipment and settings (GOV.UK)
  • Kept observers out of the participant’s session and showed it to them through a separate channel (Kusaka 2023)
  • Set the number of participants according to whether the purpose is to find problems or to measure values (Sauro 2010)
  • Decided how to rate problem severity, with several people rating independently (Nielsen 1994, Hertzum and Jacobsen 2001 abstract)

How strong is the evidence for this list?

There is empirical evidence for the number of groups and their size (Guest et al. 2017, Hennink et al. 2019). Both, however, come from health research, and whether the same numbers hold for design interviews has not been tested. The exclusion of acquaintances, the number of observers, the deadlines for deleting recordings, and the guideline numbers for online groups are rules in industry codes and guidance, and their effects have not been measured. No study measuring how much observers change what focus group participants say was found in this investigation. Nor was any study found that compared formats for checking screen operation in groups of three or more with one-to-one sessions. Among the items for operation checks, thinking aloud, evaluator variation, the number of tasks, and the effect of observers rest on empirical results, while the preparation of prototypes and data, remote arrangements, and the handling of on-screen information rest on guidance and practitioner experience.

Unverified items

No unverified statement remains in the main claims. The following academic works could not be obtained in full text by three or more routes (publishers refused machine access), and were used only within the scope of their abstracts: Boren and Ramey (2000), Guest et al. (2017, Field Methods), Guest et al. (2017, IJSRM), Hennink et al. (2019), Hertzum (2024), Hertzum and Jacobsen (2001), Kitzinger (1994), Lindgaard and Chattratichart (2007), Morgan (2010), Sonderegger and Sauer (2009), van den Haak et al. (2004), Woodyatt et al. (2016), and Wutich et al. (2010). Among the codes, the MRS qualitative research guidelines are the 2011 edition and the JMRA online guide is the first edition of 2021, and differences from current versions have not been checked.

References

All accessed 2026-11-03. The ledger and verification records are in papers.md and industry.md under source/review/group-interview-checklist/.

Academic

  • Boren, Ted, and Judith Ramey. 2000. “Thinking Aloud: Reconciling Theory and Practice.” IEEE Transactions on Professional Communication 43 (3): 261–278. https://doi.org/10.1109/47.867942
  • Carlsen, Benedicte, and Claire Glenton. 2011. “What about N? A Methodological Study of Sample-Size Reporting in Focus Group Studies.” BMC Medical Research Methodology 11: 26. https://doi.org/10.1186/1471-2288-11-26
  • Guest, Greg, Emily Namey, and Kevin McKenna. 2017. “How Many Focus Groups Are Enough? Building an Evidence Base for Nonprobability Sample Sizes.” Field Methods 29 (1): 3–22. https://doi.org/10.1177/1525822X16639015
  • Guest, Greg, Emily Namey, Jamilah Taylor, Natalie Eley, and Kevin McKenna. 2017. “Comparing Focus Groups and Individual Interviews: Findings from a Randomized Study.” International Journal of Social Research Methodology 20 (6): 693–708. https://doi.org/10.1080/13645579.2017.1281601
  • Hennink, Monique M., Bonnie N. Kaiser, and Mary Beth Weber. 2019. “What Influences Saturation? Estimating Sample Sizes in Focus Group Research.” Qualitative Health Research 29 (10): 1483–1496. https://doi.org/10.1177/1049732318821692
  • Hertzum, Morten. 2024. “Concurrent or Retrospective Thinking Aloud in Usability Tests: A Meta-Analytic Review.” ACM Transactions on Computer-Human Interaction 31 (3): 1–29. https://doi.org/10.1145/3665327
  • Hertzum, Morten, and Niels Ebbe Jacobsen. 2001. “The Evaluator Effect: A Chilling Fact About Usability Evaluation Methods.” International Journal of Human-Computer Interaction 13 (4): 421–443. https://doi.org/10.1207/s15327590ijhc1304_05
  • Kitzinger, Jenny. 1994. “The Methodology of Focus Groups: The Importance of Interaction between Research Participants.” Sociology of Health & Illness 16 (1): 103–121. https://doi.org/10.1111/1467-9566.ep11347023
  • Lander, Jonas, Simon Wallraf, Dawid Pieper, Ronny Klawunn, Hala Altawil, Marie-Luise Dierks, and Cosima John. 2025. “Recruiting Participants for Focus Groups in Health Research: A Meta-Research Study.” BMC Medical Research Methodology 25: 9. https://doi.org/10.1186/s12874-025-02464-x
  • Lindgaard, Gitte, and Jarinee Chattratichart. 2007. “Usability Testing: What Have We Overlooked?” Proceedings of CHI 2007: 1415–1424. https://doi.org/10.1145/1240624.1240839
  • Morgan, David L. 2010. “Reconsidering the Role of Interaction in Analyzing and Reporting Focus Groups.” Qualitative Health Research 20 (5): 718–722. https://doi.org/10.1177/1049732310364627
  • Quach, Susan, Jennifer A. Pereira, Margaret L. Russell, et al. 2013. “The Good, Bad, and Ugly of Online Recruitment of Parents for Health-Related Focus Groups: Lessons Learned.” Journal of Medical Internet Research 15 (11): e250. https://doi.org/10.2196/jmir.2829
  • Rupert, Douglas J., Jon A. Poehlman, Jennifer J. Hayes, Sarah E. Ray, and Rebecca R. Moultrie. 2017. “Virtual Versus In-Person Focus Groups: Comparison of Costs, Recruitment, and Participant Logistics.” Journal of Medical Internet Research 19 (3): e80. https://doi.org/10.2196/jmir.6980
  • Sonderegger, Andreas, and Juergen Sauer. 2009. “The Influence of Laboratory Set-up in Usability Tests: Effects on User Performance, Subjective Ratings and Physiological Measures.” Ergonomics 52 (11): 1350–1361. https://doi.org/10.1080/00140130903067797
  • Tausch, Anja P., and Natalja Menold. 2016. “Methodological Aspects of Focus Groups in Health Research: Results of Qualitative Interviews With Focus Group Moderators.” Global Qualitative Nursing Research 3. https://doi.org/10.1177/2333393616630466
  • van den Haak, M. J., M. D. T. de Jong, and P. J. Schellens. 2004. “Employing Think-Aloud Protocols and Constructive Interaction to Test the Usability of Online Library Catalogues.” Interacting with Computers 16 (6): 1153–1170. https://doi.org/10.1016/j.intcom.2004.07.007
  • Woodyatt, Cory R., Catherine A. Finneran, and Rob Stephenson. 2016. “In-Person Versus Online Focus Group Discussions: A Comparative Analysis of Data Quality.” Qualitative Health Research 26 (6): 741–749. https://doi.org/10.1177/1049732316631510
  • Wutich, Amber, Timothy Lant, Dave D. White, Kelli L. Larson, and Meredith Gartin. 2010. “Comparing Focus Group and Individual Responses on Sensitive Topics.” Field Methods 22 (1): 88–110. https://doi.org/10.1177/1525822X09349918

Public bodies, standards, guidance on laws, and industry codes

Research firms, consultancies, and trade journals

First-hand accounts by practitioners

Author: Shuichiro Ogawa (Design Researcher / Design Consultant) About me →