Both methods you suggested are possible in MAXQDA, but they support different workflows.

Method 1 (Memo-based condensation, then coding)

Highlight raw text = Meaning Unit

Memo attached to the segment = Condensed Meaning Unit

Assign code to the segment = Code

Organize codes into a hierarchy = Sub-categories → Categories

Themes via Code Sets or broader categories

Pros:

Keeps raw data and your interpretation distinct (transparent analytic audit trail).

Supports reflexivity: you can go back to memos to see why you condensed the unit the way you did.

MAXQDA’s memo system is built for exactly this kind of layered note-taking.

👉 Cons:

Can get memo-heavy and harder to keep “big picture” visible unless you are disciplined with memo organization.

Slightly more work up front (writing memos for every condensed unit).

Method 2 (Everything as a hierarchical code tree)

Lowest-level codes = meaning units (possibly very fine-grained).

Second-level codes = condensed meaning units or initial codes.

Higher levels = sub-categories, categories, themes.

Pros:

One unified system: all analysis lives in the code system.

Easy to export frequency tables, code matrices, or visualizations directly tied to hierarchical levels.

Less reliance on memos (fewer moving parts).

Cons:

Risk of “over-coding” — your code system may balloon into hundreds of very fine-grained codes that are hard to manage.

Less clear separation between data, condensed interpretation, and abstraction, which can make transparency weaker.

You may lose nuance of your condensed meaning unit unless you embed it in a paraphrase code label (which isn’t ideal).

3. What Most Researchers Do in MAXQDA

In practice, most qualitative researchers using MAXQDA (or NVivo/Atlas.ti) combine elements of both:

Use coding for abstraction (codes, sub-categories, categories, themes).

Use memos for interpretive notes (condensed meaning units, analytic reflections, theoretical links).

That means:

Don’t try to make every meaning unit a separate code. Instead, highlight the meaning unit, write your condensation as a memo (anchored to that segment), and then assign an appropriate code.

Build your code hierarchy for sub-categories → categories → themes.

Use memos for the “middle layer” (your interpretive condensation).

I am not sure what you mean by building "a single, hierarchical code tree." In any Qualitative Data Analysis program, you are likely to produce a set of code families, each of which has differing levels of sub-codes. Perhaps you mean working top-down and breaking up larger codes as you go, versus bottom up where you aggregate smaller codes as you go.

One of the strengths of MAXQDA is its overall ability to work with hierarchical sets of codes and sub-codes, so that you can easily create new sub-codes under exiting codes or group existing codes under larger headings. You don't have to do one or other; instead, you can work in both directions as your coding system develops.

Hello Sanam,

Though I recognise both approaches are technically feasible in MaxQDA, to me what is missing is a clear sense of how you want to interpret your data and under what kind of framework. In other words, what analytic and epistemological/ontological positions inform your analysis (e.g. positivist, critical realist, constructivist)?

Once you establish that, you can choose an appropriate process (e.g. Qualitative Content Analysis, Thematic Analysis, Grounded Theory, Framework Analysis) and deploy that in MaxQDA.

In short, I would say choose a position and approach first, then tend to the technical details. Many approaches can be suitable, but will depend on your research question and philosophical orientations.

Louis Moustakas Thank you for the information.

The study will use a basic qualitative study design and an inductive content analysis approach. The process for analyzing the qualitative data will follow the steps suggested by Graneheim and Lundman.

Right. Then technically, I think distinguishing between 'meaning units' and 'condensed meaning units' and then 'codes' is tricky based on that process. I think it makes more sense to have codes closer to what they call 'sub-categories' and then give each code a definition (via a code memo). Within the code, you will then see a list of all coded segments anyway (meaning units). If you really create individually inductive codes for each 'meaning unit' (as Graneheim and Lundman seem to do), your coding tree in MaxQDA will be unruly.

I hope that made sense?