How to use NVivo for coding and thematic analysis

To analyse interviews in NVivo, you import the transcripts, highlight passages and file them under codes, then group those codes into themes and use queries to check the pattern holds across your participants. The software organises the analysis. It does not do the thinking, and examiners can usually tell which of the two a student relied on.

This guide explains how to use NVivo in the order a real project runs: setting up, coding, building themes, and getting the results into your write-up. The steps match NVivo 14 and NVivo 15 on Windows and Mac. Versions before NVivo 14 say nodes where newer ones say codes, but the logic is identical.

Decide what you are looking for first

NVivo stores decisions. If you have not decided what the analysis is meant to answer, it will happily let you create two hundred codes that answer nothing.

Keep your research questions visible while you code. Then choose an approach. Inductive coding lets codes emerge from what participants say. Deductive coding starts from a framework, often a theory from your literature review, and tests the data against it. Many dissertations combine the two and say so in the methodology. Our guide toqualitative research approaches shows where thematic analysis sits among the alternatives.

Check the transcripts before import as well. Remove names, places and anything else your ethics approval covers, because NVivo copies the text into the project file and the copy travels wherever the file goes.

How to use NVivo, step by step

1. Create a project and import your files

Create a new project, include the date in its name, and save it somewhere your university backs up. Under Files, choose Import and bring in the transcripts. Word documents and PDFs both work. Keep one file per participant with consistent names, such as P01 Interview and P02 Interview, so sorting and comparison stay simple later.

If you recorded details about participants, such as role, age band or site, add them as case classifications now. It is what makes comparisons possible later.

2. Read everything once before coding anything

Braun and Clarke's approach to thematic analysis begins with familiarisation, and skipping it is the most common reason coding goes wrong. Read each transcript right through. Record first impressions in a memo rather than a code, because early codes tend to fix ideas you have not yet tested.

3. Code the first few transcripts

This is NVivo coding in practice. Open a transcript, highlight a passage that says something relevant to your questions, and drag it onto the Codes area in the pane on the left. NVivo asks for a name. Use a short phrase that describes the content, such as "workload blamed on staffing", rather than a single word like "workload" that could hold anything.

Code at the level of meaning. One sentence can sit under two codes, and a long answer can hold several. Coding every line is not a virtue. Coding what matters to the question is.

4. Tidy the codebook, then carry on

After three or four transcripts you will have duplicates and near duplicates. Merge them, give every code a one line description, and delete anything used once and never again. Code the remaining transcripts against that cleaner list, adding a new code only when the data genuinely needs one.

5. Group codes into candidate themes

A theme is a pattern of meaning, not a topic. Drag related codes under a parent code to build a hierarchy, then look at what each group is actually claiming. "Communication" is a topic. "Staff trust decisions they were consulted on" is a theme, and it is the kind of statement a findings chapter can argue for.

Check the themes hold up

NVivo thematic analysis is only as credible as the checks behind it. A coding query gathers every passage under a theme so you can reread them side by side and confirm they belong together. A matrix coding query crosses themes against your case classifications, which is how you notice that a theme appears in four of five managers but only one of eight frontline staff.

Word frequency queries are useful during familiarisation and weak as evidence. Treat a word cloud as a prompt to look closer, never as a finding.

Keep the judgement with you

NVivo 15 includes an AI assistant that can suggest codes and summarise documents. Check your university's policy before you touch it. Many treat AI generated analysis as undeclared help, and even where it is allowed, every code still needs a reason you can give in your own words. An examiner who asks why a passage sits under a theme wants your reasoning, not the software's.

The software cannot tell you whether you have enough data, whether your themes answer the question, or whether an interpretation goes further than participants did. Those judgements belong in your discussion chapter, and they are what earns the marks.

Get the analysis into your dissertation

Export the codebook with its descriptions and include it as an appendix. In the findings chapter, organise by theme rather than by participant, and support each theme with two or three short quotations chosen to show range rather than repetition.

In the methodology, name the software version, explain how you coded, and describe how codes became themes. Our research methods guide covers what the rest of that chapter needs, and data collection methods covers the interviews that feed the analysis.

Fix the usual problems quickly

If you are learning how to use NVivo with a deadline close, a supervised session inside your own project usually saves days of trial and error. Our data analysis coaching works that way, so the codes, the themes and the decisions stay yours.

Last updated: 14 September 2026 · Written by the AssignWise editorial team ·Spotted something wrong?