How to code qualitative data: open, axial and selective coding explained

A well-coded dataset produces findings that can be confidently defended to any stakeholder, whereas a poorly coded one produces superficial themes that may not hold up to stakeholder scrutiny. Developed by Strauss and Corbin as a systematic refinement of grounded theory coding, the three-phase framework - open, axial, and selective – is among the most reliable qualitative data coding methods used across consumer insights, UX, and market research teams worldwide. 

TL;DR 

• What it is: Qualitative coding is the systematic process of applying conceptual labels to meaningful segments of text (sentences, paragraphs, or exchanges) to transform raw language into a structured, searchable, and auditable evidence base.
  
  

• Open Coding: Breaks down raw data into discrete, unstructured labels to create an inventory of what the text contains.
  
  

• Axial Coding: Groups initial labels into broader conceptual categories and maps the relationships between them.
  
  

• Selective Coding: Integrates all categories around a single, central core finding that explains the entire dataset.
  
  

• AI Integration: Modern automation speeds up the generation of labels/ categories, but cannot replace human judgment required for rigorous synthesis. 

Key phases of qualitative data coding explained   

Phase 1: Open coding 

The open coding qualitative phase involves breaking down text into distinct, labelled segments without forcing it into a predetermined structure. Transcripts are read line-by-line and descriptive tags are attached to every meaningful thought. For example, a participant in a consumer study says, "I always intend to buy the healthier option, but by the time I get to the checkout, I just grab what I always get." You might apply multiple open codes to this single segment: 

• Intention-behaviour gap 

• Habitual purchasing 

• Checkout point-of-purchase friction 

You can also employ in vivo coding, wherein you use the participant's exact words ("grab what I always get") as the code label to preserve their precise phrasing. 

Phase 2: Axial coding 

During axial coding research, the researcher stops creating new labels. Instead, they begin connecting existing labels. The focus shifts from simple description to systemic explanation. Independent open codes are taken from phase one and clustered into structured code categories based on relationships, causes and contradictions observed. 

Continuing from the above example, you might combine the codes "habitual purchasing" and "checkout friction" into a broader category: cognitive default under pressure. 

You then link this to a separate category tracking pre-shop intentions to explain exactly why and how their initial plans failed on reaching the shelf. 

Phase 3: Selective coding 

Selective coding is the final integration phase. You look across the entire dataset to identify a single core category that accounts for the variation in your data. You are answering one definitive question: What is this entire study actually about? 

In the shopping study, your core category might emerge as the effortful exception problem - the reality that healthy purchasing requires an active conscious override carrying a cognitive cost that tired consumers refuse to pay at checkout. 

At a glance:

How to code interview data using AI  

Modern qualitative research methods incorporate Computer Assisted Qualitative Data Analysis Software (CAQDAS) to manage large-scale data streams. Artificial intelligence (AI) fits cleanly into this workflow. AI tools excel at handling high-volume, open coding qualitative data. An algorithm can quickly scan hundreds of transcript pages, generate thousands of labels, and cluster them by similarities. This pattern-matching capability accelerates the transition into the next phase, i.e., axial sorting. 

However, software cannot execute the last phase, i.e., conceptual synthesis required for selective integration. Identifying a core finding is an interpretive act based on study objectives and the researcher's knowledge of the category / brand / client context. 

Key Takeaways 

• Separate the phases: Use open coding to capture raw details, axial coding to build structural categories, and selective coding to isolate your primary takeaway.
  
  

• Avoid codebook duplication: Build your codes inductively from the text, not deductively from the discussion guide.
  
  

• Sanitise code lists: Regularly merge redundant tags during axial review to keep your codebook clean and functional.
  
  

• Verify automated labels: Treat AI coding outputs as initial sorting suggestions that always require human analytical verification against the source data. 

FAQs 

What is the definition of coding in qualitative research? 

It is the process of applying conceptual or descriptive labels to text data to systematically organise and analyse qualitative data. 

How do open coding and thematic coding differ? 

Open coding is a specific phase of grounded theory, designed to build unstructured inventories from scratch. On the other hand, thematic coding can be inductive or deductive and focuses primarily on identifying recurring patterns across data, without necessarily building a formal theory. 

Can AI handle the entire qualitative data coding process? 

No. AI can manage high-volume open-label generation and basic pattern clustering, but the interpretive synthesis needed for axial and selective steps requires human judgment. 

What is in vivo coding? 

It is a variation of analysis where the researcher uses a participant’s exact words or phrasing as the code label to preserve the authentic language of the participant's cohort. 

Why shouldn't I build my codebook directly from my interview guide? 

Coding strictly from the guide creates a closed loop that forces participant statements into pre-existing categories, causing you to miss unexpected insights that emerge outside your original questions. 

Sources 

1. Strauss, A. & Corbin, J. (1990). Basics of Qualitative Research: Grounded Theory Procedures and Techniques. Sage Publications.
Referenced for the three-phase coding framework (open, axial, selective) and the definition of selective coding as "the process of selecting the central or core category, systematically relating it to other categories."
https://www.iier.org.au/iier16/moghaddam.html 

2. Koji Research Guides (2026). Open, Axial, Selective Coding Guide.
Confirms the three-phase approach as the most widely used coding framework in qualitative research in 2026; source for the AI coding limitation finding (JMIR, 2025): "AI is strongest at the breadth-and-clustering end of coding, weakest at the depth-and-judgment end."
https://www.koji.so/docs/open-axial-selective-coding 

3. Wikipedia (2026). Axial Coding.
Definition of axial coding as "the process of relating codes to each other via a combination of inductive and deductive thinking" per Strauss and Corbin (1990, 1998).
https://en.wikipedia.org/wiki/Axial_coding 

4. Vollstedt, M. & Rezat, S. (2019). An Introduction to Grounded Theory with a Special Focus on Axial Coding and the Coding Paradigm. Springer.
Confirms that open, axial, and selective coding procedures are not rigidly sequential and do not "easily define phases that chronologically come one after the other." https://link.springer.com/chapter/10.1007/978-3-030-15636-7_4