From Codes to Themes: Practicing Thematic Analysis

Introduction

Thematic analysis (TA) according to the Braun & Clarke model includes six consecutive phases, each requiring systematic analytical work. In this article, each phase is considered in detail with practical examples, typical mistakes, and recommendations for implementation.

Phase 1: Familiarization with the Data

The goal of the first phase is deep immersion in the data. The researcher must “know” their data so well that they freely navigate its content.

Key Actions

• Multiple readings of transcripts — at least two or three full readings of each interview or document
• Note-taking (memos) — recording first impressions, interesting fragments, preliminary ideas
• Active reading — not passive perception of the text, but seeking meanings, patterns, and contradictions
• Checking transcript quality — comparing the transcript with the audio recording if necessary

Techniques for Immersion

1. Annotating — writing comments in the margins of the transcript: “The participant describes a conflict between work and family,” “Emotional reaction when discussing the supervisor”
2. Creating a brief summary for each interview (1–2 paragraphs): main topics, tone, key quotes
3. Keeping a reflexive journal — recording the researcher’s own reactions and assumptions

Sample memo: “Interview 3: the female participant repeatedly returns to the topic of injustice. She uses the metaphor ‘glass ceiling.’ Contrast between formal equality policy and actual experience.”

Phase 2: Initial Coding

Coding is the systematic labeling of data fragments that are relevant to the research question. A code is a short tag describing the content or meaning of a fragment.

Approaches to Coding

• Line-by-line coding — assigning a code to each meaningful line or sentence. Provides maximal detail
• Open coding — generating codes without pre-existing categories, “bottom-up” from the data
• Paragraph coding — assigning codes to larger semantic units

Example of Coding an Interview Fragment

Data fragment: “When I come home after work, I have no strength to spend time with my children. I feel guilty for not spending enough time with them, but I am so exhausted that I simply cannot.”

Codes:

• Physical exhaustion after work
• Work–family conflict
• Parental guilt
• Lack of time with children
• Inability to combine roles

Coding Recommendations

• Code all data, not just what matches expectations
• One fragment may have several codes
• Use descriptive codes (what is said) and interpretive codes (what is implied)
• Keep a codebook — a list of all codes with definitions and examples

Phase 3: Searching for Themes

In this phase the researcher transitions from working with individual codes to broader patterns— themes.

The Process of Grouping Codes

1. Write down all codes on separate cards or in a table
2. Look for connections — which codes are meaningfully related?
3. Group codes into potential themes (clusters)
4. Create a thematic map — a visual scheme displaying the links between themes and subthemes

Grouping Example

Thematic Map

A thematic map is a diagram showing the hierarchy of themes (main themes and subthemes) and links between them. It helps visualize the structure of the analysis at this stage and serves as a working tool to be reviewed in subsequent phases.

Phase 4: Reviewing Themes

The aim is to check how well the identified themes reflect the data. The review is carried out at two levels.

Level 1: Checking Against Coded Extracts

• Reread all data fragments assigned to each theme
• Make sure fragments consistently support the theme
• If fragments don’t align — reconsider the theme: split it, merge with another, or move codes

Level 2: Checking Against the Entire Data Set

• Reread all transcripts completely with the themes in mind
• Check whether important aspects of the data are missing
• Ensure that themes adequately reflect the overall picture

Typical Operations at This Phase

• Splitting a theme — if a theme is too broad and contains heterogeneous ideas
• Merging themes — if two themes significantly overlap
• Deleting a theme — if a theme is not substantiated by enough data
• Creating a new theme — if previously unnoticed patterns are found

Phase 5: Defining and Naming Themes

At this phase the researcher clarifies the content of each theme and formulates its name.

Writing Theme Descriptions

For each theme write a detailed description (1–2 paragraphs) answering the questions: what does the theme cover? What is its “story”? How is it connected to other themes? What aspect of the data does it reflect?

Criteria for a Good Theme Name

• Informative — the name conveys the essence of the theme, not just identifies the topic
• Concise — optimally 3–7 words
• Conceptual — the name reflects analytical work, not mere description

Poor names: “Work,” “Feelings,” “Problems” (too general, not informative)

Good names: “Invisible labor: unpaid emotional work,” “Balancing on the edge: coping strategies for combining roles,” “The ‘glass ceiling’ as everyday experience”

Phase 6: Writing Up

The final phase involves creating a coherent analytical narrative in which data and interpretation intertwine.

Writing Principles

• Each theme — a separate section with a subheading
• Alternating participant quotes and analytical researcher comments
• Quotes should illustrate analytical statements, not replace them
• Analysis should go beyond description — include interpretation, links with literature, and the research question

Typical Errors in Thematic Analysis

Error 1: Themes as Topic Descriptions

Incorrect: creating themes that merely indicate discussed areas (e.g., “Work,” “Family,” “Health”). This describes interview topics, not analytical themes.

Correct: themes must reflect analytical comprehension of the data and answer the research question.

Error 2: Too Many Themes

Excessive number of themes (more than 6–8 for one study) makes building a coherent narrative difficult. It is better to have 3–5 well-developed themes with subthemes than 10–15 superficial ones.

Error 3: Detachment from Data

Themes must be firmly rooted in the data. Every analytical statement should be backed by specific fragments from interviews or other sources.

Error 4: Paraphrasing Instead of Analysis

Simply retelling what participants said is not analysis. The researcher must interpret the data, identify latent meanings, connect results to theoretical frameworks.

Practical Assignments

Task 1. Read the following interview fragment and carry out line-by-line coding: “I always dreamed of becoming a doctor, but my parents insisted on a legal education. I enrolled in law school, but felt out of place. Only ten years later did I decide to change professions, and it was the best decision of my life.” Solution: possible codes — childhood dream of medicine, parental pressure on career choice, compliance with family expectations, feeling professional mismatch, prolonged period of dissatisfaction, career change as act of self-determination, positive evaluation of career transition.

Task 2. The following codes come from a study of students’ online learning experience: isolation, lack of live interaction, flexible schedule, saving commuting time, difficulties with self-discipline, distractions at home, convenience of lecture recording, technical problems. Group the codes into 2–3 potential themes. Solution: Theme 1 “Loss of the social dimension of learning” (isolation, lack of live interaction). Theme 2 “Freedom and its price: challenges of autonomous learning” (flexible schedule, saving time, difficulties with self-discipline, distractions). Theme 3 “Technology as tool and barrier” (convenience of recording lectures, technical problems).

Task 3. Evaluate the following theme names and suggest improvements: (a) “Stress” — too general, does not reflect data specificity. Better: “Chronic stress as the norm: normalization of overload in the academic environment.” (b) “What participants said about motivation” — descriptive, not analytic. Better: “Intrinsic motivation as a protective factor against professional burnout.”