How to Code Qualitative Data: Inductive, Deductive and Saldaña’s Cycles

What does coding mean in qualitative research?

In qualitative research, a code is, in Saldaña’s widely used definition, most often a word or short phrase that symbolically assigns a summative, salient, essence-capturing attribute to a portion of data. Coding is the process of attaching such labels systematically to your interview transcripts, field notes or documents so that you can retrieve, compare and organise what participants said.

Coding is not the analysis itself, but it is the foundation of almost every qualitative method: thematic analysis, qualitative content analysis, grounded theory and framework analysis all start with coding, even if they use different terms. What differs is how open the coding is, whether codes are fixed in advance and what you do with the codes afterwards.

Running example: you interviewed 12 first-generation students about belonging at university. Your research question is: how do first-generation students develop (or fail to develop) a sense of belonging in their first year?

Inductive vs deductive coding

Inductive coding (also data-driven or bottom-up) means you develop codes from the material itself, without a predefined list. It suits exploratory questions and approaches such as grounded theory and conventional content analysis. Deductive coding (concept-driven or top-down) means you start with a list of codes derived from theory, previous studies or your interview guide and assign data to them. It suits directed content analysis, framework analysis and studies that test or extend existing models.

In practice most student projects use a hybrid approach: you start with a few deductive main codes from your interview guide and add inductive subcodes as new aspects appear. Say so explicitly in your methodology chapter.

First cycle coding methods (Saldaña)

Saldaña’s *Coding Manual for Qualitative Researchers* describes more than 30 coding methods. You do not need all of them. Choose one or two first cycle methods that fit your research question and name them in your methods section. The most common in dissertations are listed below.

Second cycle coding: from codes to categories

After first cycle coding you may have 100 to 300 codes. Second cycle coding reorganises them into a smaller number of categories, themes or concepts. Saldaña describes methods such as pattern coding (grouping similar codes into meta-codes), focused coding (selecting the most frequent or significant codes, as in Charmaz’s grounded theory), axial coding (relating categories to subcategories and specifying their properties) and theoretical coding (integrating categories around a core concept).

At this stage you compare codes, merge duplicates, split codes that are too broad and build a hierarchy of parent and child codes. Our free coding tool at /codieren supports this with parent and child codes and a code matrix that shows which codes appear in which interviews.

• Merge synonyms: “paid work” and “part-time job” become one code
• Group related codes under a parent code: “Time poverty” with “paid work”, “commuting”, “caring responsibilities”
• Look for relationships: does time poverty lead to missing informal social events, which in turn weakens social belonging?
• Check negative cases: participants whose accounts contradict the pattern

Analytic memos: thinking on paper

An analytic memo is a short note in which you reflect on a code, a participant, a pattern or your own role. Memos are where interpretation happens. Saldaña and Charmaz both treat memo writing as essential, not optional. Write memos while you code, date them and link them to the relevant codes or segments.

Good memos ask questions: why does P04 frame asking for help as a sign of failure? Is this pattern the same for students who live at home? What do I, as a researcher, expect to find, and how could that shape my coding? Later, your memos become the first drafts of paragraphs in the findings chapter.

Practical decisions: units, lumping, splitting, rounds

Before you start, decide on your coding unit. Most students code meaningful segments (a sentence to a paragraph) rather than single words or entire answers. Code enough context that a quote still makes sense when retrieved on its own.

Saldaña describes two styles: lumpers apply one code to a larger passage, splitters code line by line with many fine-grained codes. Splitting produces richer but more numerous codes; lumping is faster but can miss nuance. Both are legitimate; be consistent.

• Round 1: code all interviews with first cycle methods
• Round 2: recode with the revised code list and check consistency across interviews
• Round 3: second cycle coding to categories or themes
• Throughout: update your codebook with definitions and examples for every code
• Your supervisor or department may prescribe a specific coding procedure: follow your university’s guidelines (you can upload them to Professor Digital)

How to describe your coding in the methodology chapter

Examiners look for transparency: what did you do, in what order and why? A concise description might read: “Coding followed a hybrid approach. Deductive codes were derived from the interview guide; inductive codes were developed through descriptive and in vivo coding (Saldaña, 2021). In a second cycle, pattern coding was used to group 164 initial codes into 18 categories. Analytic memos documented decisions throughout.” Then provide the codebook and an example of a coded transcript excerpt in the appendix.