Qualitative research generates rich, detailed material that offers deep insights for any research team, but that detail does not organise itself. Interviews, focus groups, diaries, and open-ended research often have lengthy passages of text, shaped by context, memory, interaction, and background data. Traditionally, in qualitative research, coding is the method that qualitative researchers use to work through bodies of research, breaking it into meaningful parts so that patterns can be examined without losing sight of the original data.
Coding is not a single, standardised step in qualitative analysis. In practice, researchers use different types of coding depending on what they are trying to learn, what themes start to come up, and where they are in the analytic process. Some approaches remain deliberately open and descriptive, supporting early familiarisation with the data. Others are more selective or relational, helping to structure interpretation and theory building at later stages.
This article focuses on the core types of qualitative coding that have shaped research practice for decades. These approaches form the foundation of how qualitative data has traditionally been analysed, whether manually or with the support of early qualitative analysis software.
More recent AI-assisted approaches build on these same methods. They change how coding work is carried out, shifting effort away from manual organisation and towards interpretation and judgement.
What is qualitative coding in research methodology?
In research methodology, qualitative coding refers to the process researchers use to organise and interpret non-numerical data systematically. Coding involves assigning labels to segments of text so ideas, experiences, or actions can be examined across a dataset rather than remaining tied to individual accounts.
This coding allows researchers to work analytically with qualitative material, moving from detailed transcripts or documents towards patterns that can be compared, interpreted, and explained. The way data is coded reflects the focus of the research and shapes how findings are developed, making coding a central part of qualitative analysis rather than a preparatory step.
What are the different types of qualitative coding?
The main types of qualitative coding used in research methodology are often described in terms of first-cycle, or initial coding, and later-stage analytic coding. These labels reflect when and how different coding approaches tend to be applied during analysis, rather than fixed steps that must be followed in sequence. Some types of coding are used more frequently than others, and many qualitative studies combine several approaches as understanding of the data develops.
First-cycle coding
First-cycle coding takes place once data collection is complete, whether that data comes from interviews, focus groups, open-ended survey responses, or observational notes. At this stage, the researcher works directly with the raw material, reading and rereading transcripts to begin labelling segments of text in a way that captures what is being said, felt, or done. The purpose is familiarity rather than explanation, allowing ideas, actions, emotions, and language to surface without forcing them into predefined categories.
1) Structural coding
Structural coding is used straight after transcripts have been collected to organise qualitative data according to research questions, interview topics, or sections of a discussion guide. Rather than capturing meaning within the data, structural coding indexes content based on where it sits within the structure of the study.
For example, all responses to a question about daily routines may receive the same structural code, regardless of what is said within those responses. This allows researchers to quickly retrieve and compare data across participants by topic.
Structural coding is applied early in analysis, particularly in large datasets or team-based research, as it supports systematic organisation before deeper interpretive coding begins.
2) Open coding
Open coding is used as a starting point in qualitative research. Researchers work through the data, line-by-line, and assign labels to segments of text without applying a fixed structure in advance; thus, the “open” element of the coding.
At this stage, the codes are typically broad and descriptive. Researchers sometimes use the participants’ own words or short phrases that capture an idea, reaction or theme. At this stage, where the codes are typically vague, there may be overlap. The key here is to get full coverage of the data, rather than granular insight.
How this may look in practice would be on a statement such as:
“I kept checking my phone because I didn’t want to miss anything, but it was making me feel more anxious.”
This could be open-coded as checking the phone frequently, fear of missing information, or increased anxiety. These codes are not final categories. They act as markers that help the researcher return to similar moments across the dataset during later analysis.
3) Descriptive coding
Descriptive coding involves assigning simple, low-inference labels to segments of qualitative data to summarise what is being discussed. It is different from open coding as the focus is organisational rather than exploratory, with codes used to capture the topic of a passage rather than interpret its underlying meaning.
Coding labels here are small and factual. This type of coding is commonly used to help researchers build an overview of what the data contains and identify where specific topics recur across the dataset.
Within the types of qualitative coding, descriptive coding provides structure without pushing interpretation too early. It creates a practical map of the data that can support later, more interpretive analysis, including thematic or pattern-based approaches.
4) In vivo coding
In vivo coding uses the Latin phrase ‘within the living’ and takes the participants’ own words as codes. Short phrases and terms are lifted directly from the data and applied to preserve the language people use to describe their experiences.
This approach is best suited for when the wording itself carries meaning, where voice particularly matters. Using the participants’ own language helps keep the analysis grounded in how people express themselves, rather than translating it into researcher-defined terms too quickly.
This particular approach stays close to the data. It will typically be combined with other coding approaches as the patterns between the participants’ language begins to emerge.
5) Process coding
Process coding focuses on actions, interactions, or changes over time. Codes are typically written as verbs or short action phrases and are used to capture what is happening in the data as situations unfold.
This approach is applied when the research question is concerned with behaviour and decision-making. Instead of grouping ideas or meanings, process coding tracks sequences, responses, and transitions, such as how people adapt, respond, or move through a particular experience.
6) Emotion coding
Emotion coding focuses on identifying feelings that are expressed or implied within qualitative data. These emotions may be stated directly, such as frustration or anxiety, or inferred from the way participants describe experiences, decisions, or relationships.
This approach is often used in research concerned with lived experience, identity, stress, wellbeing, or motivation. By coding for emotional responses, researchers can begin to see how people feel about events and situations, not just what they do or think.
Emotion coding is rarely used on its own. It is typically layered onto other first-cycle methods, helping add emotional depth to descriptive or process-focused analysis and supporting later thematic development.
7) Values coding
Values coding captures beliefs, attitudes, and value systems that shape how participants interpret their experiences. This includes statements that reflect what people see as important, fair, acceptable, or worthwhile.
For example, references to independence, security, fairness, or responsibility may not describe a specific action but reveal underlying priorities that influence behaviour. Values coding is particularly useful in studies exploring culture, ethics, decision-making, or organisational norms.
Within qualitative research, values coding helps move beyond surface description by highlighting the principles that sit beneath actions and opinions, which can later inform broader thematic or interpretive analysis.
8) Evaluation coding
Evaluation coding records judgements made by participants about people, processes, systems, or outcomes. These judgements may be positive, negative, or mixed, and often appear in the form of opinions, appraisals, or assessments.
This approach is commonly used when understanding how something is perceived or assessed is a core research aim.
Evaluation coding works well alongside descriptive and values coding, helping researchers distinguish between what participants experienced and how they evaluated those experiences.
Second-cycle coding
Second-cycle coding happens after the initial round of coding, once the dataset has been broken down into a large number of first-cycle codes. Here, the focus shifts from labelling to analysis. Codes are compared, grouped, merged, or discarded as the researcher looks for patterns, relationships, and underlying meaning. This is the stage where broader categories and themes are developed, helping transform detailed observations into insights that can support interpretation, theory building, or research conclusions.
1) Pattern coding
Pattern coding is used to condense large numbers of first-cycle codes into fewer, more meaningful categories. It helps researchers identify explanatory groupings that capture trends, similarities, or recurring configurations within the data.
This method supports the move from detailed labelling to higher-level insight. Multiple codes that relate to similar behaviours or experiences may be brought together under a single pattern that reflects a broader issue or dynamic.
2) Thematic coding
At this stage of analysis, thematic coding builds on earlier coding by reviewing existing labels to identify repetition, similarity, and connection. Related codes are brought together and developed into themes that represent broader concepts present across the data. For example, earlier codes such as checking the phone frequently or fear of missing an important email may be grouped into a theme such as work life boundaries erosion. These themes show how individual statements and viewpoints cluster into shared experiences.
Thematic coding differs from pattern coding in its analytic intent. Pattern coding is primarily concerned with consolidation and organisation, reducing a large set of codes into more manageable groupings. Thematic coding goes a step further by interpreting what those groupings mean. Themes bring related codes together in a way that helps explain how experiences or perspectives are shared across the dataset.
The goal of thematic coding is to develop a clear understanding of common experiences and concerns within the data. It supports interpretation rather than theory construction.
3) Axial coding
Axial coding focuses on examining relationships between existing codes, looking at how ideas connect across the dataset. At this point in the process, a qualitative researcher looks at the difference between cause and effect, be that through different conditions, actions, or responses and consequences. Codes that may have initially appeared to be unrelated may start to build clusters around broader categories.
A researcher can start to bring in more structure, without pigeonholing or closing down interpretation. It moves the data from description towards explanation, while still maintaining a data-based grounding.
4) Selective coding
Selective coding focuses on a smaller set of codes that become central to the overall shape of the analysis. At this stage, the researcher looks across the dataset as a whole and begins to prioritise the material that contributes most directly to the emerging explanation or narrative.
Some codes are refined, merged, or set aside as the analysis narrows. The emphasis shifts from exploring everything the data contains to clarifying what matters most in relation to the research question.
The purpose of selective coding is consolidation. It brings multiple analytic threads into a more focused account that can be communicated clearly in reports, papers, or other research outputs.
How AI-assisted coding changes the process in qualitative research
AI-assisted tools such as Beings handle much of the mechanical coding work internally. allowing researchers to move straight into analysis rather than manual labelling. Beings uses internal coding structures to identify patterns, cluster related material, and surface themes across a dataset almost immediately.
When transcripts or recordings are uploaded, Beings can highlight recurring themes and organise the data around them. These themes are not treated as conclusions. They act as entry points that researchers can explore and refine.
For example, Beings’ analysis assistant, Aida, might surface a theme such as work-life boundaries across multiple interviews and present the exact transcript excerpts and timestamps where that theme appears. This keeps themes directly linked to the source material and allows researchers to assess how well they are grounded in participant accounts.
What remains unchanged is analytical responsibility. Researchers still decide which themes matter and how they should be interpreted. Beings reduces the manual effort involved in organising and revisiting qualitative data, creating more space for interpretation and judgement.
If you want to explore how this approach works in practice, you can try Beings for free.
