In qualitative research, “coding” means reading what people said and tagging it with themes, such as “price worries” or “trust in the brand,” so patterns become visible. Traditionally this happens after the study ends. Fieldwork closes, answers are exported, researchers code for days, and the readout is a scramble. This post explains a better rhythm: code while the board is still live, and use what you see to steer the study.
The old loop and the new loop
The old loop is field, export, code, and panic before the client meeting. The new loop is field while themes accumulate, probe what looks thin, and walk into the meeting with a draft narrative already forming. The research is the same, but problems surface when you can still fix them.
What “AI themes” should mean
It should not mean a magic slide that appears from nowhere. A good AI theme view is a living map of coded themes across respondents that researchers can question, with quick access to the raw answers, photos, and transcripts underneath. The researcher stays in charge. The AI does the first pass of sorting so people can spend time on judgment.
Three tests of a trustworthy theme view
- Traceability: you can click from a theme to every quote, image, and transcript line that supports it.
- Correctability: you can rename, merge, split, or reject a theme without fighting the tool.
- Honesty about weight: it shows how many respondents support a theme, not just that it exists.
Why this matters during the study
- Spot missing stories early enough to extend a day, add a follow-up, or send a nudge.
- Brief stakeholders mid-field without building a temporary deck.
- Notice if one theme is driven by only two loud respondents before you build a story on it.
- Cut the post-field crunch that burns out senior researchers.
- Show a client real progress instead of only a completion percentage.
What to watch out for
Early themes are hypotheses, not findings. On Day 1 you have a few respondents and the most eager voices. Treat early patterns as prompts for probing, and expect them to shift as quieter people answer and as later days reveal behavior that contradicts what people first said. Also keep in mind that automated coding can over-weight vivid language, so spot-check against the raw data.
A practical routine for fieldwork
- 1After the first day, review the early themes and compare them with your hypotheses.
- 2Open the raw answers behind any theme that surprises you.
- 3Check media and transcripts for stories the text answers do not capture.
- 4Add a probing question for the next day if a topic looks thin.
- 5Before close, mark the themes you trust and the ones you still doubt.
| Fieldwork moment | What the theme view helps you decide |
|---|---|
| End of Day 1 | Are people understanding the tasks? Is any topic missing entirely? |
| Mid-study | Which segments differ? What deserves a deeper probe on later days? |
| Day before close | Which themes are solid enough to headline, and which need one more question? |
| After close | How to structure the story, not how to start coding from zero |
Example: on Day 2 of a four-day board about meal planning, early themes show that “time” dominates but “budget” barely appears, which the team did not expect. They add a Day 3 question on weekly spending and discover a hidden split between two audience groups, all before the study closes.
FAQ