Qualitative data analysis software such as NVivo, ATLAS.ti and MAXQDA is widely used, but it is not essential. Many excellent qualitative studies have been analysed by hand, using paper, highlighters, word processors and spreadsheets. Some researchers prefer manual coding because it keeps them close to the data. Others simply do not have access to a software licence, or have a small dataset that does not justify learning a new program.
Coding manually is a legitimate and rigorous approach, as long as it is systematic and transparent. This post explains what coding is, how to prepare for manual coding, several practical methods (paper, Word and Excel), how to move from codes to themes, and how to keep an audit trail that will satisfy your examiners.
What is coding?
Coding is the process of labelling segments of qualitative data, such as sentences or paragraphs of an interview transcript, with short words or phrases that capture their meaning. A code is like a tag that says, “this part of the data is about X”.
For example, a participant says: “I always feel like I'm the last to know about changes. By the time I hear, the decision has been made.” You might code this as “excluded from decisions” or “late communication”.
Codes help you:
- Organise large amounts of text.
- Find all data related to a topic.
- Compare what different participants say.
- Build towards broader categories and themes.
Software does not code for you. Whether you use software or not, the thinking is the same; software mainly helps with storage, retrieval and organisation. That is why manual coding can be just as rigorous.
Advantages and limitations of manual coding
Advantages
- Closeness to the data. Physically working with transcripts can deepen your engagement.
- No cost or learning curve. You use tools you already know.
- Flexibility. You can spread transcripts on a table, move sticky notes and see patterns visually.
- Suitable for small to moderate datasets.
Limitations
- Time-consuming retrieval. Finding all quotes for one code across many transcripts takes longer.
- Harder to manage large datasets or multiple coders.
- Risk of disorganisation without a clear system.
- Changing codes across many documents requires careful tracking.
If you have a very large dataset, such as more than a few dozen long interviews, consider whether software would save time. For smaller studies, manual coding works well.
Before you start: preparation
1. Prepare your transcripts
Use anonymised transcripts with participant codes. Add line or paragraph numbers so you can reference quotes precisely, for example “P04, lines 112–118”. Leave wide margins if working on paper.
2. Choose your analytic approach
Manual coding can support many approaches: thematic analysis, framework analysis, content analysis, grounded theory and others. Your approach determines how you code. For example:
- Inductive coding: codes come from the data, without a predefined list.
- Deductive coding: you start with codes based on theory, research questions or an existing framework.
- Hybrid coding: you start with some deductive codes and add inductive ones as they emerge.
3. Familiarise yourself with the data
Read all transcripts at least once before formal coding. Write short memos about first impressions, surprising points and possible patterns.
Method 1: Coding on paper
This is the most traditional method.
- Print transcripts with line numbers and wide margins.
- Read and highlight meaningful segments. Use different colours for broad areas if helpful, but do not rely on colour alone.
- Write codes in the margin next to each highlighted segment.
- Create a code list on a separate sheet, with a short definition of each code.
- Cut and sort (optional). Some researchers photocopy transcripts and physically cut out coded segments, labelling each with the participant code and line numbers, then group them into piles by code.
- Use sticky notes or index cards to write codes and arrange them on a wall or table to explore relationships and potential themes.
Take photographs of your arrangements at different stages. These become part of your audit trail.
Method 2: Coding in Microsoft Word (or another word processor)
- Use the comments feature. Highlight a segment and insert a comment containing the code name. Comments appear in the margin, similar to paper coding.
- Use highlighting colours for broad categories.
- Create a separate codebook document listing codes, definitions and example quotes.
- Collect coded segments. Copy segments into a separate document organised by code, adding the participant code and line numbers after each quote. Alternatively, some researchers use tables with columns for code, quote, participant and location.
Word's search function makes it easy to find all instances of a keyword, though remember that coding is about meaning, not only keywords.
Method 3: Coding in Excel or Google Sheets
Spreadsheets are very effective for manual coding, especially for sorting and filtering.
- Break transcripts into segments. Paste each meaningful segment (for example, each answer or paragraph) into a separate row.
- Create columns such as: Participant code, Question or topic, Line numbers, Data segment, Code 1, Code 2, Memo.
- Assign codes in the code columns. Use consistent spelling; a drop-down list created with data validation helps.
- Filter and sort by code to see all segments with the same code together.
- Add columns for participant characteristics, such as role or experience, to compare patterns across groups.
- Use a separate sheet as your codebook, with code names, definitions, inclusion and exclusion notes, and examples.
This method works particularly well for framework analysis, where data are summarised in a matrix of participants (rows) by themes (columns).
Developing a codebook
A codebook is a key tool for consistency, whatever method you use. For each code, record:
- Code name: short and clear.
- Definition: what the code means.
- When to use: the kind of data it applies to.
- When not to use: to distinguish it from similar codes.
- Example quote: a typical data segment.
Your codebook will change as you code more data. Keep dated versions to show how your coding developed.
From codes to themes
Coding is not the end of analysis. The next step is to look for patterns across codes and build higher-level categories or themes.
- List all codes and review them. Merge codes that mean the same thing, and split codes that are too broad.
- Group related codes into categories. Sticky notes, index cards or a spreadsheet column for “category” help here.
- Develop candidate themes. A theme is a pattern of shared meaning that answers your research question, not just a topic label.
- Create a thematic map by drawing themes, sub-themes and their relationships on paper or a whiteboard.
- Review themes against the data. Go back to the coded segments and full transcripts to check that each theme is supported and that themes are distinct.
- Define and name themes, writing a short description of each.
A short worked example
Here is a brief illustration of manual coding using a fictional interview excerpt from a study on remote working among university administrators.
P06 (lines 45–52): “At first I loved working from home. No commute, I could take my child to school. But after a few months I noticed I was answering emails at ten at night. There's no clear point where the day ends. And honestly, I miss just turning to a colleague and asking a quick question. Now everything needs a meeting.”
Possible initial codes:
- Lines 45–46: early enthusiasm; time saved on commuting; flexibility for family responsibilities.
- Lines 47–49: working beyond hours; blurred work–life boundaries.
- Lines 50–52: loss of informal interaction; increase in formal meetings.
After coding several transcripts, you might notice that “working beyond hours” and “blurred boundaries” appear often together, and that participants with caring responsibilities describe both the benefits and the costs of flexibility. You could then group these codes into a candidate theme such as “Flexibility as a double-edged sword”, with sub-themes about family benefits and boundary erosion. Similarly, “loss of informal interaction” and “increase in formal meetings” might form a theme about changing patterns of collaboration.
In a spreadsheet, this excerpt would appear as three rows, one for each segment, each with the participant code, line numbers, text and codes. Filtering by “blurred work–life boundaries” would then show you every similar segment across all participants, making comparison easy.
Keeping an audit trail
Examiners want to see how you moved from raw data to findings. With manual coding, you need to create this trail deliberately. Keep:
- Coded transcripts (scanned if on paper).
- Dated versions of your codebook.
- Analytic memos explaining decisions.
- Photographs of sticky note arrangements or thematic maps.
- A table showing themes, sub-themes, codes and example quotes.
An appendix with a sample coded transcript and a coding framework table is often very persuasive.
Improving rigour
- Code consistently using your codebook definitions.
- Double-code a sample with a supervisor or colleague, then discuss differences. Depending on your methodology, this may be about developing shared understanding rather than calculating agreement statistics.
- Reflect on your position as a researcher and how it may influence your interpretation. Keep a reflexive journal.
- Look for negative cases that do not fit your themes and explain them.
Common mistakes to avoid
- Coding without reading all the data first.
- Creating too many overlapping codes without definitions.
- Treating codes as themes without further analysis.
- Losing track of where quotes came from.
- Not saving versions of the codebook.
- Assuming manual coding is less rigorous than software coding.
Final thoughts
You do not need expensive software to analyse qualitative data well. Paper, Word and Excel can all support rigorous coding if you work systematically. Prepare your transcripts, choose your approach, build a clear codebook, move carefully from codes to themes, and keep a detailed audit trail. What matters most is the quality of your thinking and the transparency of your process, not the tool you use.
Do you code by hand or with software? Share your favourite manual coding tips in the comments.