You coded everything and still have no theory

The problemThe least documented step in qualitative method

That gap has a name in the methods literature, and it is not a personal failing. Köhler, Lê and Smith call the missing step the domain of “advanced analytic techniques”, and open their 2026 paper by observing that resources for it “are largely absent from textbooks, research methods syllabi, and methods sections in empirical papers” (p. 2). What the tools and templates cover well is the lower half of the work: import, transcribe, code, retrieve, count. The step that makes a study worth reading, the conceptual leap “from the mass of words … to an abstract and explicit set of concepts, relations and explanations” (Klag & Langley 2013, p. 150, quoted on p. 2), is the least documented part of qualitative method. Their explanation is uncomfortable: the field’s push for standardized templates has made researchers hide their actual discovery process, because stumbling and leaping look less scientific than following a procedure (pp. 3, 25–26).

ChallengingTake the surprise seriously

Theorizing starts when something in the material refuses to fit. An unexpected observation, a divergent case, a questioning comment during a member check: the anomaly “draws into doubt what the researcher thought they knew” and produces what the paper calls generative doubt, a motivated state of not knowing (p. 6). In their first worked example, a research team noticed managers repeatedly using the term “end-to-end management” for coordination, a usage that felt unusual to them as management scholars. Instead of coding past it, they made the odd term itself the object: retrieved every mention, preserved context and temporal order, and let the anomaly restructure the analysis (p. 12).

The practical move: when the data surprises you, write the surprise down before you resolve it, together with the explanation it just defeated. A surprise you code over is gone by the next session.

SeeingRun an analytic experiment

Seeing is “new ways of making sense of some aspect of existing social worlds” (Klag & Langley 2013, quoted on p. 7), and it is produced by moves. The paper’s examples are concrete. Arranging the end-to-end data month by month exposed that the concept was evolving for the participants themselves (p. 13). In the second example, coding meeting transcripts thematically produced only superficial differences between German and American teams; building activity logs, minutes spent per activity in order, made the patterns “literally jump out”: different rhythms, content, and participation, replicated across independent tasks (pp. 17–18). In the third, temporal bracketing and visual maps of six firms’ international expansion turned a mass of archival data into comparable phases, and the “squiggles” where expansion stalled became the finding (pp. 20–22).

None of these moves is exotic. Each is an experiment: a conjecture about what might be going on, an arrangement of the material that would show it, and a result that confirms, complicates, or kills the conjecture. The authors’ word for choosing moves is methodological bricolage, “the combining of analytic moves for the purpose of solving a problem or problems tailored to one’s own research project” (Pratt et al. 2022, quoted on p. 7). There is no fixed sequence to comply with; there is a fit between question, setting, data, and move. My epistemic notes have a flow, and that flow is the analytic experiment.

ArticulatingWrite the claim so it can be examined

The third process is getting the insight out of your head in a form others can inspect: writing, visualizing, presenting, iterating (p. 10). In the expansion study this took years, several rejected framings, and reviewer pressure before “sensegiving followed by sensemaking” emerged as the explanation that held (pp. 22–23). Articulation is analysis: each attempt to state the claim exposes where it exceeds the material.

# Conceptual leap — [title]

## Observation        — what you saw; quote it, with source
## Surprise           — why it stopped you; what you expected instead
## Existing explanation — the theory that should have covered this, and where it falls short
## Generative doubt   — the open question, written as a question
## Conjecture         — what might be going on (dated, signed speculation)
## Analytic move      — what you did to test it, and on which data
## Result             — confirmed, complicated, or killed; say which
## Alternative explanations — what you considered and did not adopt, and why
## Theoretical consequence  — what you can now claim, and where the claim stops

Set it upWith your agent, or with epistemic notes

If you keep your analysis in plain files, paste this to your AI agent:

I keep my qualitative analysis in plain markdown files. Create a file called
"Conceptual leap note template.md" in my templates folder with these sections:
Observation, Surprise, Existing explanation, Generative doubt, Conjecture,
Analytic move, Result, Alternative explanations, Theoretical consequence.
Add a one-line comment under each saying what belongs there.

Do not fill in any section for me, and never suggest conjectures, explanations,
or theory: a leap note only counts if I wrote it. After I have written one, your
job is limited to checking structure: flag any empty section, check that
"Analytic move" names data I can point to, and check that "Alternative
explanations" contains at least one explanation I did not adopt. Report
what you find; I decide what it means.

Or use my Epistemic Notes system, where each stage of the leap already has a state:

Most qualitative software assumes the value lies in organising data. My vault assumes the value lies in tracking epistemic movement. The real contribution comes from the movement between observation, doubt, imagination, and theory, what they call challenging, seeing, and articulating.

Related guidesRead next

ReferenceThe paper

• Köhler, T., Lê, J., & Smith, A. (2026). Advanced qualitative analyses for theorizing: Balancing deep data engagement with imaginative theorizing. The Journal of Applied Behavioral Science (open access).