Coding and thematic analysis - these two questions will help you develop Themes

Turning a pile of codes into usable themes gets easier once the study’s purpose is treated as the filter. The core move is to ask why the research is needed and what change the findings are meant to support—because those answers determine which patterns in the data actually belong in the thematic story and which ones can be set aside.

A first checkpoint is rationale: what problem the study is trying to address, and why that problem matters in the relevant context. Closely tied to that is the study’s aims and implications—who is expected to benefit and what practical or policy outcomes might follow. When those pieces are clear, theme selection becomes less about “keeping everything” and more about choosing the codes that help build the narrative needed to address the problem. For example, if the study focuses on teaching and feedback, the rationale for improving teaching practices and the hoped-for benefits (better guidelines, better policies) become the yardstick for deciding which candidate themes are worth developing.

The transcript then uses a hypothetical study about why ex-offenders struggle during job interviews to show how this filter works in practice. The obvious starting point is a theme about struggles—barriers, obstacles, and challenges that show up in interview responses. But the data might also include accounts of strategies used before interviews, such as meditation or researching the workplace. Those details can still matter even if they don’t directly answer “why they struggle,” because the long-term aim may be to help more people succeed in interviews. In that case, what helps some candidates become successful is relevant to the broader goal of reducing struggle.

The same logic applies to information that seems off-target at first glance. If participants describe what they do during the interview—how they dress, how they act, or recommendations they make about prison training—those elements may not be the direct mechanism behind “struggling,” but they can illuminate why current preparation falls short and what interventions could prevent future problems. Recommendations about improving training, for instance, can both connect back to the “why” (preparation isn’t adequate) and support the “so what” (what should change).

The central warning is that focusing only on the narrow wording of the research question can lead to themes that are too limited, even when the dataset contains richer material. The goal isn’t to guess what others might notice; it’s to build a strong contribution to the real-world issue the study targets. Once the rationale and implications are pinned down, codes can be organized into themes that reflect the narrative the study needs to tell—without losing the parts of the data that are essential for impact.