Short answer: The constant comparative method is a qualitative analysis technique in which every new piece of data is compared with data already coded, codes are compared with codes, and categories are compared with categories until their properties and relationships are clear. Introduced by Barney Glaser in 1965, it is the analytic engine of grounded theory and is now widely used outside it.
Last updated: October 22, 2026
\n\nThe constant comparative method is one of the most cited and least clearly explained procedures in qualitative research. Methods chapters often say \"data were analysed using the constant comparative method\" and stop there, which tells a reviewer almost nothing. This guide explains what the method is, where it comes from, how to apply it step by step to interview data, and how AI-assisted coding can support it without replacing the researcher's comparisons.
\n\nThe constant comparative method is an iterative way of analysing qualitative data in which coding and comparison happen at the same time. Instead of coding everything first and interpreting later, the analyst asks of each new segment: is this the same as, different from, or a variation of what I have already coded? The answer refines the code, splits it, or creates a new one.
\n\nBarney Glaser first set out the procedure in \"The Constant Comparative Method of Qualitative Analysis\" in Social Problems (1965), and it became a core chapter of Glaser and Strauss's The Discovery of Grounded Theory (1967). Glaser positioned it between two older approaches: coding data first to test hypotheses, and reading data to generate ideas without systematic coding. Constant comparison does both: it codes systematically, but in order to generate theory.
\n\nMost textbook accounts, and the 2025 human-in-the-loop computational grounded theory framework by Alqazlan, Fang, Castelle and Procter, describe comparison at three levels:
\nGlaser described four stages that overlap rather than follow in strict order. Each stage keeps operating as the analysis grows.
\n\n| Stage | What you do | Typical output |
|---|---|---|
| 1. Comparing incidents applicable to each category | Code each incident into as many categories as fit, comparing it with previous incidents in the same category | Emerging codes and first properties |
| 2. Integrating categories and their properties | Compare incidents with the properties of the category, not just with other incidents | Categories with defined properties and conditions |
| 3. Delimiting the theory | Reduce the set of categories, drop irrelevant properties, stop coding what no longer adds anything | A smaller, more abstract set of core categories |
| 4. Writing the theory | Use the coded data and memos to write the theory, category by category | Theory or conceptual model, supported by memos |
One rule Glaser gives is worth keeping: while coding an incident, stop and write a memo when a theoretical idea comes up. The memos, not the codes, carry the theory forward. Our guide to analytic memos covers how to write them.
\n\nGlaser's stages describe the logic; they do not tell a PhD student what to do on Monday morning. Hennie Boeije's paper \"A Purposeful Approach to the Constant Comparative Method in the Analysis of Qualitative Interviews\" (Quality & Quantity, 2002) fills that gap. Boeije defines each step by its data, aim, result and guiding questions, using a study of couples living with multiple sclerosis. The general sequence for interview data is:
\n\nFor research teams, Olson, McAllister, Grinnell, Gehrke Walters and Appunn propose a 10-step approach for multiple investigators (The Qualitative Report, 2016) that adds explicit checks for agreement between coders. It is a useful reference if a reviewer asks how three people applied the method consistently.
\n\nImagine a study of how early-career researchers decide whether to use AI tools for transcript coding.
\nMany studies use constant comparison inside thematic analysis, case studies or naturalistic inquiry. Sheila Fram's commentary in The Qualitative Report (2013) argues this is legitimate and that the method can keep an insider (emic) perspective while a theoretical framework provides an outsider (etic) lens. The condition is transparency: say that you used constant comparison as a technique, not that you produced a grounded theory. If you are deciding between the two traditions, see our thematic analysis vs grounded theory comparison.
\n\nThis is an original Qualitati checklist for methods chapters and peer review. If you cannot tick an item, your write-up probably needs one more sentence.
\n\n| Check | What a reviewer wants to see | Done? |
|---|---|---|
| Analysis started during data collection | Dates or interview numbers showing coding began before fieldwork ended | ☐ |
| Levels of comparison named | Within-interview, between-interview, across-group comparisons described | ☐ |
| Code definitions versioned | Codebook versions showing how definitions changed after comparisons | ☐ |
| Memos kept | Examples of memos that led to new or merged categories | ☐ |
| Negative cases sought | Passages that did not fit, and what you did with them (see negative case analysis) | ☐ |
| Sampling linked to comparisons | How gaps in categories shaped who was interviewed next | ☐ |
| Stopping rule stated | Why comparison stopped adding new properties (see data saturation) | ☐ |
| AI role disclosed | Which comparisons a tool suggested, and which the researcher made and checked | ☐ |
Yes, for parts of it. Large language models are good at retrieval tasks that make comparison faster: finding passages similar to a selected excerpt, listing every segment under a code so you can reread them side by side, or suggesting that two codes overlap. Systems such as LOGOS go further and attempt end-to-end grounded theory, reporting partial alignment with expert schemas.
\n\nThe judgement at the heart of the method is harder to hand over. Deciding whether a new passage is a variation of a code or evidence for a new one depends on your research question, your sensitising concepts and what you learned in earlier interviews. Theoretical sampling depends on fieldwork access no model has. Treat AI suggestions as prompts for your own comparison, and record them in your memos.
\n\nQualitati is a European, budget-friendly qualitative research platform for universities and research companies, with GDPR and data privacy as central priorities. For constant comparison, three parts are relevant:
\nQualitati does not do theoretical sampling or write your memos for you, and it should not. You can start free with 30 monthly credits, no credit card required, or view transparent pricing.
\n\nHuman-review note: methodological guidance here summarises the cited sources; check your discipline's and supervisor's expectations before finalising a methods chapter.
\n\nPhD students and research labs doing interview-based grounded theory or theory-building qualitative work, and research firms that need a defensible, documented analysis procedure.
\n\nIf your study tests a fixed framework deductively, uses all data collected before analysis with no chance to sample further, or aims only at descriptive summaries, a codebook or framework approach may fit better. See deductive vs inductive coding.
\n\nBarney Glaser published it in Social Problems in 1965. It was then incorporated into The Discovery of Grounded Theory (Glaser and Strauss, 1967).
\n\nNo. Constant comparison is the core analytic technique of grounded theory, but grounded theory also includes theoretical sampling, memoing and theory generation. You can use constant comparison in other designs if you report it honestly.
\n\nGlaser lists four stages: comparing incidents, integrating categories and properties, delimiting the theory and writing it. For interviews, Boeije (2002) gives practical steps: compare within an interview, between interviews in one group, and across groups.
\n\nThematic analysis usually codes the full dataset and then builds themes. Constant comparison codes while collecting data and uses each comparison to refine categories and decide what data to collect next.
\n\nWhen new data no longer add properties to your core categories, often called theoretical saturation. State the criterion you used and when you reached it.
\n\nAI can retrieve similar passages and suggest code merges, which speeds up comparison. The decision about what counts as the same, different or new should stay with the researcher and be documented.
\n\nThe constant comparative method is less a single technique than a habit: compare each new piece of data with what you already have, write down what the comparison teaches you, and let it guide what you collect next. Reported clearly, with memos, codebook versions and a stopping rule, it gives reviewers what \"we used constant comparison\" never does. If you want to run interviews in waves and code them as you go, start free with Qualitati or compare it with NVivo and ATLAS.ti.
\n\nThis article is an independent editorial summary of the cited publications. It is not affiliated with their authors.
\n","related":[{"slug":"gioia-methodology-qualitative-research-2026","title":"Gioia Methodology: Data Structure Guide with Example","description":"The Gioia methodology explained: 1st-order concepts, 2nd-order themes, aggregate dimensions, a worked example, key critiques, and what AI can and cannot do.","category":"Qualitative Data Analysis","date":"2026-10-18","readTime":"11 min read","author":"Qualitati Research Team"},{"slug":"what-is-abductive-analysis-qualitative-research-2026","title":"What Is Abductive Analysis in Qualitative Research?","description":"Abductive analysis builds theory from surprising data. Learn Timmermans and Tavory's steps, abductive vs inductive vs deductive, and where AI can help.","category":"Qualitative Data Analysis","date":"2026-10-17","readTime":"10 min read","author":"Qualitati Research Team"},{"slug":"discourse-analysis-qualitative-research-2026","title":"Discourse Analysis in Qualitative Research: Types & Steps","description":"Discourse analysis in qualitative research explained: 4 approaches compared, 7 steps, a worked example, and what a 2026 study says about using LLMs for it.","category":"Qualitative Data Analysis","date":"2026-10-15","readTime":"10 min read","author":"Qualitati Research Team"}]}