Designing for Emergence in Focus Groups
One of the most common pieces of advice in introductions to qualitative research is the need to alternate between data collection and data analysis. The common metaphor here is a cyclical rather than a linear process. Another name for this strategy is emergent design, where the original research design is only a starting point, which evolves over time. This contrasts to quantitative research, which typically does move in a straight line fashion from research design to data collection to data analysis.
There certainly are projects where a linear approach to focus groups makes perfect sense: you know what you want, and you collect and analyze the data that meets those purposes. Even so, it seems like we fall back on this approach too often, rather than following a cyclical process. There is a real temptation to set up a research design and follow through with it. Of course, problems that show up in the first group or two will lead to revisions, but as long as things go well enough, there isn’t any change in the selection of the participants, the questions that get asked, or the style of moderating. My argument is that we miss a lot of opportunities to improve the quality of our data when we don’t let the lessons that we are learning in the field influence the development of our research designs. But that still leaves a major question: How do you design for emergence? How do you plan for open-ended procedures? Hence, the goal here is to describe one basic design that moves back and forth between data collection and analysis in focus groups.
The foundation for this version of emergent design is a funnel approach. The model here is as the funnel pattern in interview guides, which moves from broad exploratory questions to semi-structured questions, before finishing with relatively narrow, researcher-oriented questions. The idea is to apply this three-part structure to a set of groups. The attachedEmergence%20Attachment.jpg diagram shows a schematic version of this design for a set of six focus groups, where the first two are very open-ended and participant oriented, followed by preliminary analyses that lead to a second set of more structured groups, which lead in turn leads to more analyses that set up the final pair of “wrap-up” groups. It should be apparent that this is a highly idealized summary that is primarily useful for illustrative purposes. For example, a truly emergent design wouldn’t call for a two-two-two division across the three parts of the funnel; instead, it would allow for either more or fewer groups in each part, according to what it takes to meet the relevant goals.
Even though the design in the diagram looks linear, it is vital to recognize that there are two places that require reflection on the existing data before moving on to further data collection. These “preliminary analyses” creates the opportunity for a considerable amount of emergence, so that the design can evolve in response to findings in the field, during the course of the overall project.
The earliest groups in this design are the least structured. They use a relatively small set of questions that are oriented to the participants’ perspective, along with a non-directive style of moderating. Like the questions at the top of a funnel, the goal here is to explore the research topic with a minimum of influence from the researcher. For the approach shown in the diagram, the plan is to collect two of these exploratory groups before conducting a preliminary analysis. The point of that analysis is to generate the likely directions for the next stage in the data collection; hence, it serves as a connecting point in the movement back and forth between data collection and analysis.
The second round of data collection can now be more structured, because it is already based on listening to the participants in the initial set of groups. That allows for a mix of questions that pursue both the participants’ perspectives and the researchers’ interests. Interestingly, one of the most likely formats for this set of interviews would be a funnel. This makes it possible to proceed from more open-ended questions to more narrowly focused queries. In particular, it is quite likely that some of the topics in the first set of interviews are less saturated than others. The need to expand on those less saturated topics makes them good candidates for the initial, more participant-oriented questions at the top of the funnel. After that, the middle of the funnel would consist of semi-structured questions that address the researchers’ agenda. The wrap-up in these interviews would serve the traditional purpose of creating a sense of closure at the end of the interview.
Once again, this data collection is followed by preliminary analysis, and again, the goal is to move from what has been learned so far, to determine the procedures for the next set of interviews. AS before, saturation is likely to serve as a criterion for which topics get carried over and which ones are considered to have sufficient coverage. In addition, it is entirely possible that new issues and interests have emerged during the second set of groups, and these would be added to the guide for the third set of interviews. Following the principle of a narrowing funnel, this final set of interviews is designed to be the most structured. The goal is thus to conclude the data collection by being as thoroughly saturated as possible across the widest range of relevant research topics.