Guides  /  how to use nvivo

How to use NVivo

NVivo is filing software with a query engine attached. It will not find your themes, and it will not do your analysis — but it will let you code several hundred pages systematically and get back to any extract in seconds. This is how the pieces fit together.

Written and reviewed by Elaine Halliburton, Professor of Applied Statistics
Updated 2 September 20268 min read
How do you use NVivo?

Create a project file, import your sources (transcripts, documents, PDFs), then read through them and code passages to nodes, which are NVivo's term for codes. Once material is coded you can retrieve everything at a node, compare coding across cases using attributes, and run queries. NVivo organises and retrieves your coding; the analytic judgement remains yours.

Orientation

What NVivo is and is not

NVivo does not analyse qualitative data. This is the single most useful thing to understand before you start, and it saves a great deal of disappointment.

What NVivo actually does, in order 1. Projectone file 2. Sourcesimport data 3. Nodesyour codes 4. Codingtext to node 5. Queryinterrogate The vocabulary, translated Source = a document you imported  ·  Node = a code  ·  Reference = one coded extract NVivo does not analyse anything. It organises what you code, so you can find it again. The thinking stays yours. The software is filing, retrieval and counting.
The workflow, and the three words that cause most of the confusion.

What it does is hold your material in one place, let you tag passages consistently, and retrieve every passage tagged the same way instantly. On a project with forty transcripts that is genuinely transformative. On a project with four, a table in a word processor may serve you just as well.

NVivo doesNVivo does not
Store all sources in one project fileIdentify themes for you
Let you tag passages to codes consistentlyDecide what is analytically interesting
Retrieve every extract at a code instantlyWrite your findings
Compare coding across participants or groupsMake your analysis rigorous by itself
Count coding frequency and co-occurrenceReplace reading the data closely
The auto-coding question

NVivo offers automated and AI-assisted coding. It is genuinely useful for a first pass on structured material — splitting an interview by question, for example. It is not a substitute for analytic coding, and an examiner who sees themes that were machine-generated without researcher judgement will ask about it. If you use it, say what it did and what you did.

Terminology

The vocabulary

Three words account for most early confusion, because NVivo does not use the terms you learned in your methods course.

NVivo calls itWhich meansIn methods language
SourceA document you importedA transcript, field note, PDF, image
NodeA container for coded materialA code
ReferenceOne passage coded at a nodeA coded extract
CaseA unit you compare acrossA participant, site or organisation
AttributeA characteristic of a caseA demographic or grouping variable
AnnotationA note attached to a passageA margin comment
MemoA standalone analytic documentAn analytic memo
Node is just the word for code

Nothing conceptual hangs on it. Once you read 'node' as 'code' and 'reference' as 'coded extract', the interface stops being mysterious.

Getting started

Setting up a project

Create a project, not a folder

NVivo keeps everything — sources, nodes, coding, memos — inside a single project file. There is no folder of separate documents to manage.

Save it somewhere local, then back it up

Working from a synced cloud folder while NVivo has the file open is a known cause of corruption. Work locally and copy the file to backup after each session.

Decide your unit of analysis before importing

Usually the participant. This determines how you set up cases later, and changing it afterwards means reworking your case structure.

Back up more often than feels necessary

An NVivo project is one file containing every hour of coding you have done. There is no version history and no autosave to a separate copy. A dated copy after each working session costs seconds and has saved a great many dissertations.

Sources

Importing your data

Import on the ribbon handles documents, PDFs, spreadsheets, audio, video and images. A few things are worth doing before you import rather than after.

Tidy the transcripts first — consistent speaker labels, one file per participant, sensible filenames
Use heading styles in Word documents if your interviews follow a schedule — NVivo can then auto-code by question
Name files as you want them to appear, because the filename becomes the source name
Anonymise before importing, not after — find-and-replace across a project is far more awkward

If your transcripts follow a consistent interview schedule with headings, Auto Code → Use paragraph styles will split every transcript by question in one step. On a structured project that alone can save several hours, and it is entirely legitimate — it is organisation, not interpretation.

The core task

Coding to nodes

Coding is the work. Everything else is preparation for it.

Open a source and read it

Read a whole transcript before coding anything in it. Coding line by line on first contact produces codes that describe sentences rather than meaning.

Select a passage, then right-click → Code → at New Node

Or drag the selection onto an existing node in the Navigation View. The keyboard shortcut Ctrl+F2 (Windows) codes to a new node directly.

Name codes for what is happening, not the topic

“Support” is a topic. “Seeking support without appearing to need it” is a process, and it carries the tension that makes it analytically useful.

Let the list get long before you organise it

Fifty or eighty codes after the first few transcripts is normal. Grouping them into a hierarchy too early forces the analysis into a structure you have not earned yet.

A passage can be coded at as many nodes as it warrants, and overlapping coding is normal rather than a mistake. NVivo tracks each independently.

Write memos as you go

Create → Memo, linked to a node or a source. This is where the analysis actually develops — the codes are only its raw material. If someone later asks how you moved from codes to themes, a dated sequence of memos is the only convincing answer, and it cannot be reconstructed afterwards.

Want the coding framework checked before you commit to it?

Send a few transcripts and your coding so far. A named researcher reviews the codebook, tests it against your extracts, and advises on developing themes.

See qualitative coding

Comparison

Attributes and cases

This is the step that unlocks the comparison NVivo is genuinely good at, and it is the step most people skip.

Set each participant up as a case, then give cases attributes — role, site, years of experience, whatever distinguishes them. Once that is in place you can ask NVivo to show you everything coded at a node by any attribute.

QuestionHow NVivo answers it
What did senior staff say about workload?Matrix coding query: node × role attribute
Do the two sites differ on this theme?Matrix coding query: node × site attribute
Which participants never mentioned this?Coding query filtered by case
Set attributes up early

Adding them retrospectively across thirty cases is tedious. Build the classification sheet when you import, using Import → Classification Sheets from a spreadsheet if you already hold the demographics.

Queries

Queries worth knowing

QueryWhat it answersUse it for
Coding queryEverything coded at a node, filteredPulling extracts for writing up
Matrix codingNode against attributeComparing groups
Text searchEvery occurrence of a word or phraseChecking you have not missed a topic
Word frequencyMost common words across sourcesEarly orientation only
Coding comparisonAgreement between two codersInter-rater reliability, kappa
Word clouds are not findings

Word frequency queries produce attractive visualisations that say very little — the most common word in almost any interview corpus is a filler. Use them to orient yourself early, not as evidence in a results chapter.

The coding comparison query is worth knowing about if your method calls for inter-rater reliability. It computes percentage agreement and Cohen's kappa between two coders on the same sources. Note that this suits content analysis and codebook approaches; in reflexive thematic analysis, reporting a kappa misrepresents what the method claims.

Common problems

Where people get stuck

Too many codes, no structure

Normal at first, and only a problem if it persists. Once the list stops growing, group codes into parent nodes by what they have in common analytically — not alphabetically.

Coding everything

If 90% of every transcript is coded, the codes are not discriminating. Code what bears on your research question.

The project file will not open on another machine

NVivo for Windows and NVivo for Mac have used different file formats. Check the version before assuming a file is corrupt, and convert deliberately rather than by opening and hoping.

Lost work after a crash

NVivo does not autosave to a separate file. Back up the project after every session.

Not knowing what to do after coding

This is a methods problem rather than a software one. NVivo has organised your material; deciding what it means is the analysis, and no query will do it for you.

Answers

Frequently asked questions

What is NVivo used for?

Organising and retrieving qualitative data. It stores transcripts, documents and media in one project, lets you tag passages to codes consistently, and retrieves everything tagged the same way instantly. It does not identify themes or perform the analysis — that judgement remains the researcher's.

What is a node in NVivo?

A node is simply NVivo's word for a code — a container holding every passage you have tagged with a particular idea. A passage coded at a node is called a reference. Reading 'node' as 'code' removes most of the confusion with the interface.

Is NVivo hard to learn?

The mechanics take a few hours. The difficulty is not the software but the qualitative analysis itself, which NVivo neither teaches nor performs. Most people who feel stuck in NVivo are actually stuck on a methods question about what to code and why.

Do I need NVivo for thematic analysis?

No. Thematic analysis can be done well in a word processor or spreadsheet, and Braun and Clarke are explicit that software is not required. NVivo earns its place when you have a large volume of material — roughly twenty transcripts or more — or need to compare systematically across participant groups.

Can NVivo code my data automatically?

It offers auto-coding and AI-assisted coding. These work well for structured splitting, such as dividing transcripts by interview question. They are not a substitute for analytic coding, and if you use them you should state clearly what the software did and what you did.

How do I back up an NVivo project?

Copy the project file to a separate location after every working session. NVivo keeps everything in a single file with no version history and no separate autosave, so a corrupted or lost file takes all your coding with it.

What is the difference between a case and a node in NVivo?

A node is a code — a container for material about an idea. A case is a unit you compare across, usually a participant, and it carries attributes such as role or site. Setting cases up lets you run matrix queries asking how coding differs between groups.