Focus group discussions are valuable for understanding shared language, social norms, reactions and differences in perspective. They are not designed to estimate population percentages.
What FGDs are good at
Focus group discussions (FGDs) create structured interaction among participants. That interaction is the method’s distinctive advantage: participants compare experiences, challenge one another, reveal shared language and expose points of disagreement that may not emerge in an individual interview.
FGDs are useful for concept testing, behaviour exploration, service experience, message development, product research, programme learning and formative research. They are less suitable when the main goal is to estimate prevalence or produce population-representative percentages.
Recruitment determines the quality of the room
A sophisticated discussion guide cannot rescue poor recruitment. Eligibility criteria should reflect the research question, and screening should identify participants with the relevant experience. Researchers should also avoid mixing participants whose status differences may suppress discussion, such as supervisors and direct reports or service providers and their clients in the same group.
Segmentation may be needed by age, gender, location, user type, socioeconomic group or another meaningful characteristic. The purpose is not to create endless subgroups; it is to create enough psychological safety and comparability for useful discussion.
Language and cultural context are research variables
Across African markets, language decisions often affect what participants can express. A respondent may understand an official language but communicate emotions, idioms or sensitive experiences more naturally in another language. Moderators should be selected for both language competence and cultural fluency.
Translation should preserve meaning rather than mechanically replace words. Research teams should agree terminology for difficult concepts, brief moderators on probes and document where locally meaningful language changes the interpretation of a concept.
Moderation: structure without over-control
A moderator guides the group without turning the session into eight individual interviews conducted in sequence. Good moderation opens discussion, probes examples, brings quieter participants in, controls domination and explores disagreement. The moderator should remain neutral and resist the temptation to teach, correct or defend the concept being tested.
Projective exercises, sorting tasks, stimulus materials and written reactions can deepen the discussion when they serve a clear research purpose. Activities should not be added simply to make a session feel creative.
Operational details that affect data quality
Venue accessibility, privacy, seating, sound quality, power supply, internet connection for remote observation, respondent comfort and recording setup all matter. When concepts are shown on tablets or screens, researchers should test visibility and device consistency before the session.
Consent must cover recording and any remote observation. Incentives should compensate participation appropriately without becoming coercive. Research teams also need secure processes for storing audio, video, transcripts and identifiable recruitment information.
Analysis should go beyond memorable quotes
Qualitative analysis begins with the research questions, not with the most dramatic statement in the transcript. A strong analysis compares themes across groups, identifies patterns and exceptions, examines differences by segment and links claims to supporting evidence.
Quotes can bring findings to life, but they are illustrations rather than vote counts. The final report should explain what participants meant, why the pattern matters and how the insight affects a decision. That is what turns a well-run FGD into useful research.
Decide group composition deliberately
Group composition affects what participants are willing to say and how the discussion develops. Researchers should consider which characteristics need to be similar within a group and which differences are useful for comparison. Age, gender, occupation, user status, income, language, geography or programme exposure may matter depending on the question.
Too much heterogeneity can suppress discussion because participants have little shared experience or because status differences create caution. Too much homogeneity can also narrow the range of perspectives. The recruitment plan should therefore explain why particular participants are grouped together and how the full set of groups will capture relevant diversity across the study population.
Design a guide around conversation, not a questionnaire
A discussion guide should create a logical flow from easy introductory topics to deeper exploration. Questions are usually open, neutral and broad enough for participants to describe experiences in their own language. Probes then explore reasons, examples, exceptions and points of disagreement.
Researchers should avoid packing the guide with every question the client could possibly ask. An overloaded guide encourages rushed moderation and shallow answers. It is often better to identify a small number of priority learning questions and use activities or stimuli only when they help participants think more concretely. Timing should be tested during a pilot so the moderator has enough space to pursue unexpected but relevant insights.
Manage observation and concept testing carefully
Remote observation can add value, particularly in concept, communication or product research, but it should not change the room into a performance. Participants need to know who may be observing and how recordings will be used.
Client observers should normally communicate questions through the research team rather than interrupting the session. For concept testing, exposure order, device display, language and time spent on each stimulus should be controlled so concepts receive a fair comparison. Immediate individual reactions before group discussion can also help capture first impressions that might otherwise be influenced by more vocal participants.
Create a transparent qualitative analysis process
Analysis is stronger when the team agrees a coding or thematic framework and systematically reviews all groups rather than relying on memory from the most vivid sessions. A matrix can compare themes by city, segment, concept or respondent type and record both dominant patterns and important exceptions.
Analysts should distinguish between frequency, intensity and explanatory value: an issue mentioned by only a few participants may still be strategically important if it reveals a barrier or mechanism. Findings should be supported by carefully selected quotations and by an audit trail showing how interpretations were developed. This helps qualitative evidence remain rigorous without pretending to be statistically representative.
How many focus groups are enough?
There is no universal number of FGDs that guarantees adequate insight. The required number depends on how many distinct segments, markets and research questions must be compared and on how much variation exists within the target population.
A study should continue until the planned comparison groups are covered and additional sessions are producing limited new insight on the core questions. This judgement should be made transparently rather than by claiming statistical saturation from a very small convenience sample. For multi-market work, teams should also distinguish between insights that repeat across locations and those that appear to be specific to one context.
How Surveysphere Africa can support
Surveysphere Africa supports qualitative research from recruitment and moderation to transcription, translation, thematic analysis and multi-market reporting across Africa.
