Skip to main content

Complete guide

Grounded Theory: A Complete Guide to the Approach and Analysis

Grounded theory is a qualitative methodology for building theory from data rather than testing a pre-existing one. Developed by Glaser and Strauss (1967), it uses systematic, iterative coding and constant comparison to develop an explanatory account grounded in participants' accounts. It is widely used in sociology, nursing, education, and management research.

This guide covers what grounded theory is, when a grounded theory approach fits your question, how grounded theory analysis works in practice (open, axial and selective coding, constant comparison, theoretical sampling and memo-writing), the main schools, a short worked example, how many interviews you need, and the mistakes examiners flag most often.

What is grounded theory?

Grounded theory is both a method and a product: a systematic approach to analysing qualitative data, and the explanatory theory that results. Unlike methods that describe or interpret patterns, grounded theory aims to generate a theory — an account of how and why something happens — that is grounded in the data.

It is distinctive in that data collection and analysis happen together and iteratively: you analyse early data, let it shape what you collect next, and keep comparing until your categories are well-developed and no new properties emerge (theoretical saturation).

The coding stages: open, axial, selective

Grounded theory coding typically moves through three overlapping stages (terminology varies by school):

  1. 1Open coding — break the data into discrete incidents and label them, generating many initial concepts close to the data.
  2. 2Axial coding — relate concepts to each other — grouping them into categories and identifying connections, conditions, and consequences.
  3. 3Selective coding — integrate categories around a central (core) category to form a coherent theory, and refine until saturated.

When to use a grounded theory approach

A grounded theory approach fits when your question is about a process: how something happens, how people deal with a situation, or how a change unfolds over time, and when existing theory does not explain it well. Questions that start with "How do..." or "What happens when..." are often a good fit.

It is a poor fit when you mainly want to describe experiences or identify patterns across a dataset (thematic analysis is usually better), when you want to understand the lived meaning of an experience (consider interpretative phenomenological analysis), or when your sample and data collection are fixed in advance. Grounded theory depends on theoretical sampling, so it needs the freedom to decide who to interview next as the analysis develops.

The techniques that define grounded theory

  • Constant comparison — continually compare new data and incidents against existing codes and categories, refining them as you go.
  • Theoretical sampling — let your emerging analysis guide what data you collect next, rather than fixing the sample in advance.
  • Memo-writing — write analytic memos throughout to develop ideas about categories and their relationships; memos become the building blocks of your theory.
  • Theoretical saturation — keep sampling and comparing until new data adds no new properties to your categories.

The main schools of grounded theory

Grounded theory split into distinct traditions. Classic (Glaserian) grounded theory emphasises emergence and discourages forcing data into preconceived frames. Straussian grounded theory (Corbin & Strauss) offers a more structured coding paradigm. Constructivist grounded theory (Charmaz) treats the theory as co-constructed between researcher and participants and foregrounds reflexivity.

The schools also name their coding stages differently. Strauss and Corbin use open, axial and selective coding; Glaser distinguishes substantive coding (open, then selective) from theoretical coding; Charmaz describes initial, focused and theoretical coding. They also disagree about when to read the literature: Glaser advised delaying the literature review to avoid forcing ideas onto the data, while Charmaz and Corbin and Strauss accept earlier engagement, provided you stay open to what the data shows.

Examiners expect you to name which version you are following and to apply it consistently — mixing schools without acknowledgement is a common critique.

Grounded theory analysis: a short worked example

This example is invented for illustration. Imagine a study asking how newly qualified nurses learn to cope with their first night shifts.

Open coding of the first six interviews produces many small codes close to the data: "watching the clock", "asking without looking weak", "checking twice", "borrowing a senior's confidence". Constant comparison shows several of these are about the same thing: managing how competent you appear to others. That becomes a tentative category, "managing visibility", and a memo records what it seems to involve and when it does not apply.

Theoretical sampling follows the category rather than a fixed plan. To test it, the researcher interviews senior nurses who mentor new staff, and nurses on wards with different cultures. The new data adds properties (visibility matters more where mistakes are discussed publicly) and conditions (it fades once a nurse has handled a first emergency). Eventually one category explains most of the others and becomes the core category, for example "earning quiet competence", and the theory explains how nurses move through it. Sampling stops when new interviews no longer add properties to the categories.

How many interviews do you need for grounded theory?

The criterion is theoretical saturation, not a number: you stop when new data no longer develops your categories. For planning and ethics applications, methods texts commonly suggest 20 to 30 participants (Creswell, 2013), and a review of 100 published grounded theory studies found an average sample of about 25. Plan for a range, explain that the final number depends on saturation, and record in your memos why you stopped.

Writing up a grounded theory study

Examiners look for evidence that you actually did grounded theory, not just that you named it. A strong write-up usually includes:

  • Which version of grounded theory you followed, and why it suits your question and epistemology.
  • How theoretical sampling decisions were made, with examples of how early analysis shaped later data collection.
  • Extracts from memos showing how key categories developed.
  • How the core category was chosen and how it integrates the other categories.
  • A diagram or model of the theory, with each part explained and supported by data.
  • How you judged saturation, and how the theory relates to existing literature.

Common mistakes in grounded theory studies

  • Calling a study grounded theory when it produces descriptive themes rather than a theory. If there is no explanation of a process, it is usually thematic analysis.
  • Fixing the sample in advance, so there is no theoretical sampling.
  • Claiming saturation without showing how it was judged.
  • Mixing Glaserian, Straussian and constructivist procedures without acknowledging it.
  • Treating open, axial and selective coding as strict one-way steps rather than overlapping, iterative ones.
  • No memos, or memos that are never used in the write-up.

Keeping AI-assisted grounded theory defensible

Grounded theory is deeply interpretive and iterative, so AI can only ever assist — surfacing candidate incidents for a category, helping organise concepts — never do the theorising. The defensible pattern is the same as for any AI-assisted analysis: the AI surfaces candidates, you confirm or reject each, and the audit trail records how categories developed.

QualIntel OS supports the iterative, comparative coding grounded theory relies on, keeps an audit trail of how each category developed, and generates an AI disclosure statement — so the analytic process central to grounded theory rigour is documented.

Frequently asked questions

What is grounded theory in simple terms?

Grounded theory is a way of building a theory from qualitative data rather than testing an existing one. You code data systematically, constantly compare incidents, and keep collecting and analysing together until you have developed an explanatory account grounded in what participants said. It originated with Glaser and Strauss in 1967.

What are the three types of coding in grounded theory?

Open coding (labelling discrete incidents to generate initial concepts), axial coding (relating concepts into categories and identifying connections), and selective coding (integrating categories around a core category to form a coherent theory). The stages overlap and are iterative rather than strictly sequential.

What is the difference between grounded theory and thematic analysis?

Grounded theory aims to build an explanatory theory through iterative coding, constant comparison, and theoretical sampling, with data collection and analysis happening together. Thematic analysis identifies and interprets patterns of meaning (themes) across a dataset and does not require theory-building or iterative sampling. Grounded theory is more demanding methodologically; thematic analysis is more flexible.

What is a grounded theory approach?

A grounded theory approach means building theory from your data rather than testing an existing theory. Data collection and analysis happen together: you code early interviews, compare incidents constantly, decide who to sample next based on the emerging categories, and keep going until the categories are saturated. It suits questions about processes, such as how people deal with a situation over time.

How many interviews do you need for grounded theory?

There is no fixed number; you stop at theoretical saturation, when new data no longer develops your categories. For planning, methods texts commonly suggest 20 to 30 participants, and published grounded theory studies average about 25. Explain in your write-up how you judged saturation.

What is theoretical saturation in grounded theory?

Theoretical saturation is the point at which collecting and analysing more data adds no new properties or insights to your categories. It is the signal that your categories are well-developed and you can stop sampling. Reaching and demonstrating saturation is part of grounded theory rigour.

Can I use AI for grounded theory analysis?

Yes, with disclosure and on the condition that you remain the analyst. Because grounded theory is interpretive and iterative, AI can only assist — surfacing candidate incidents and helping organise concepts — not do the theorising. Keep an audit trail of how categories developed. QualIntel OS supports this and generates a disclosure statement. Confirm your institution's AI policy.

Related guides

Build grounded theory in QualIntel OS

Iterative, comparative coding with an audit trail of how every category developed — and AI disclosure generated as you go. One project free, no card.

Start for free