Template Analysis: A Priori Themes and Iterative Template Revision

Template analysis starts from a coding structure, not a blank page. Before touching a single transcript, the researcher builds an initial set of themes — drawn from the interview guide, prior research, or a theoretical framework guiding the study — and arranges them into a hierarchical template. That template is then applied to the data and revised, over and over, until it fits what the transcripts actually say. This puts template analysis in a specific middle position among qualitative coding approaches: less rigid than a matrix that has to be filled in for every case from the start, and less committed to starting with nothing than an approach that treats any prior theme as contamination. This guide covers how to build a starting template, how the revision cycle actually works, and how to defend the middle-ground choice in a methods section.

What Makes Template Analysis Different: Starting From a Template

Template analysis was set out by Nigel King, initially in a chapter on qualitative methods in organizational research, and developed further through King and Christine Horrocks’ Interview Methods in Qualitative Research and later King and Joanna Brooks’ Template Analysis for Business and Management Students. Its defining move is permitting — and expecting — the researcher to define some themes in advance of coding, then treating that initial set as a working draft rather than a fixed grid. The initial themes come from wherever a study’s design already points: the research questions, the structure of a semi-structured interview guide, existing theory the study is testing or extending, or the researcher’s own prior fieldwork. Those become the a priori themes that start the template.

The a priori themes are not the finished coding scheme. Once real transcripts are coded against them, some will not fit, some data will not fit any of them, and new distinctions will appear that the interview guide never anticipated. Template analysis builds that mismatch into the method itself, through the revision cycle below, rather than treating it as a failure of the initial design.

Building the Initial Template

A workable starting template is deliberately provisional. King’s guidance is to keep it short — a handful of broad themes, not an exhaustive scheme — because an over-specified initial template makes the researcher more likely to force-fit data rather than notice what does not belong. A typical way to build one:

• List the a priori themes. Pull them from the interview guide’s main topic areas, the study’s research questions, and any theoretical framework the study is explicitly working from.
• Code a small subset of the data first — often four or five transcripts, chosen to be reasonably representative — using the a priori list as a starting point, but allowing new codes to be added freely as they come up.
• Cluster the resulting codes into a first draft template, organizing related codes under broader headings and deciding, provisionally, how many levels of hierarchy the template needs.

This first draft is what gets applied — and revised — against the rest of the dataset.

The Revision Cycle: Applying, Revising, Re-Applying

The template is not written once. It is applied to more data, checked against what does and does not fit, changed, and applied again. The changes a template goes through in this cycle generally fall into a few recurring types:

• Adding a theme the a priori list missed entirely.
• Deleting a theme that turns out to have little or no data behind it once more transcripts are coded.
• Changing the scope of a theme — narrowing one that was too broad to be useful, or broadening one that was artificially split.
• Changing a theme’s level in the hierarchy — promoting a sub-theme to a higher-order theme in its own right, or nesting a theme that turns out to be a facet of a broader one.
• Merging or splitting themes once enough coded data makes clear that two draft themes are really one, or one draft theme is really two.

Each round of revision is checked against data the template has not yet seen, not just against the transcripts that produced the change. The cycle stops when applying the template to additional transcripts stops producing meaningful changes — the working definition of a final template for that study, not a fixed number of rounds decided in advance.

Hierarchical Coding: Higher- and Lower-Order Themes

A template is a tree, not a list. Broad, higher-order themes sit at the top; more specific lower-order themes nest underneath them, sometimes several levels deep. The number of levels is not fixed by the method — a straightforward study might need two levels, a richer one might run to four or five — but King’s guidance is to keep the hierarchy only as deep as the data actually supports distinguishing, rather than adding levels for the sake of a tidier-looking template. A common failure mode is over-nesting: creating sub-themes that do not actually carry distinct meaning from their parent theme, which makes the template harder to code against consistently without adding real analytic value.

Integrative Themes

Some patterns in a dataset do not sit neatly inside one branch of the hierarchy — they cut across several branches at once. King’s template analysis literature uses integrative themes for this: a theme that captures a relationship or pattern running through parts of the template that are not otherwise connected in the tree structure. An integrative theme is not another node added to the hierarchy in the ordinary way; it is a separate, explicit statement of a cross-cutting pattern the hierarchy on its own would not surface, and it is written up as such rather than forced into a single branch it does not actually belong to.

Keeping a Template Version Log for an Audit Trail

Because the template changes repeatedly, the version history is itself part of the study’s evidence trail — the same audit-trail logic that underpins documenting a qualitative audit trail more generally. A workable version log records, for each revision:

• A version number or date, and how many/which transcripts had been coded when that version was current.
• What changed — theme added, deleted, rescoped, moved, merged, or split — named specifically, not just “template updated.”
• The reason for the change: which data prompted it, in enough detail that a supervisor or reviewer could see why the change was made.

This is what turns “the template was revised iteratively” from an unverifiable claim in a methods section into something a reader can actually follow. It is also what makes it possible to show, concretely, which parts of the final template were a priori and which emerged — a distinction reviewers of template analysis work routinely ask to see made explicit.

Template Analysis vs. Framework Analysis vs. Grounded Theory

These three are often confused because all three organize qualitative data into named, structured categories rather than leaving analysis as unstructured close reading. What actually separates them is how much structure is fixed before coding starts, and what form that structure takes.

The practical way to tell them apart in a draft methods section: if the write-up describes building an a priori set of themes and then revising a hierarchy against the data, that is template analysis. If it describes charting summarized data into a matrix of cases by themes, that is framework analysis. If it describes deliberately starting with no predefined categories and coding toward a theory through open, axial, and selective coding, that is grounded theory.

When Template Analysis Fits Well (and When It Doesn’t)

Template analysis suits studies that already have a reasonably clear theoretical or applied starting point — an existing framework being tested or extended, or a well-developed interview guide — but where the researcher still expects, and wants room for, the data to push back against that starting point. It is also, by design, flexible about epistemological position: King explicitly built it to work across a range of positions from more realist to more constructionist, rather than being tied to one, which is part of why it shows up widely in applied fields such as organizational, occupational, and health psychology research.

It fits less well where a study genuinely has no starting theoretical position and building one in advance, even provisionally, would misrepresent the design — that is grounded theory‘s territory. It also fits less well where the priority is a highly systematic, auditable matrix that several researchers can code into consistently across a large number of cases for an applied or policy audience — that is what framework analysis is built for. Template analysis sits in the middle of that range by design, not by default; the choice needs to be justified against the study’s actual starting point, not just picked as the option “in between” the other two.

A Worked Example: From Initial to Revised Template

The template excerpt below is an illustrative composite built to show the mechanics of the revision cycle — it is not drawn from a real study, and no specific numbers in it should be read as findings.

Say a study is examining how new academic staff experience onboarding at a university. The interview guide covers workload, mentoring, and administrative systems, so the a priori template starts simply:

• 1. Workload adjustment
• 2. Mentoring and support
• 3. Administrative systems

After coding the first five transcripts against this list, two things happen. First, “Administrative systems” turns out to cover two things participants clearly experience differently — formal IT/HR onboarding tasks, and the informal, undocumented local knowledge nobody had told them about — so it splits into two lower-order themes under a renamed higher-order theme. Second, a pattern shows up that the a priori list never anticipated: several participants describe a specific tension between wanting mentoring and not wanting to look like they need it, which recurs inside both the mentoring theme and the informal-knowledge theme rather than sitting inside either one cleanly. That becomes an integrative theme — “visible competence vs. asking for help” — noted separately rather than forced under one branch.

The revised template, after that round:

• 1. Workload adjustment
• 2. Mentoring and support
• 3. Institutional knowledge
  • 3.1 Formal onboarding tasks (IT, HR, compliance)
  • 3.2 Informal/undocumented local knowledge
• Integrative theme: Visible competence vs. asking for help

The version log entry for this round would record: version 2, transcripts 1–5 coded, theme 3 renamed and split into 3.1/3.2 because IT/HR tasks and informal knowledge produced clearly distinct participant accounts, and one integrative theme added because the competence/help-seeking tension recurred across themes 2 and 3.2 rather than belonging to either alone. This is the level of specificity a version log needs to be useful — not “template revised,” but which change, where, and why.

Common Mistakes

• Treating the a priori template as final. If a template survives contact with the full dataset completely unchanged, that is much more often a sign the data was force-fit to the template than that the initial guess happened to be exactly right.
• Over-nesting the hierarchy. Adding levels because the template looks more thorough, not because the data actually supports distinguishing that finely.
• Skipping the version log. A methods section that asserts “the template was iteratively revised” with no record of what changed and why is not verifiable, the same problem that affects an undocumented bracketing claim.
• Confusing an a priori theme with a hypothesis. A priori themes are organizing categories the researcher expects to be relevant, not predictions about what the data will show — the coding still has to be checked against the data, not used to confirm an expected conclusion.
• Calling any coded-then-revised hierarchy “template analysis.” The method has a specific lineage and specific vocabulary (a priori themes, hierarchical coding, integrative themes); a generic iterative coding process that never explicitly names or applies these is closer to reflexive thematic analysis than to template analysis specifically, and should be labelled as such.

Frequently Asked Questions

What is template analysis in qualitative research?

Template analysis is a qualitative coding technique in which the researcher builds an initial, a priori set of themes from theory, research questions, or an interview guide, arranges them into a hierarchical template, and then revises that template repeatedly against the data until it stabilizes.

What are a priori themes?

A priori themes are the codes a researcher defines before coding any data, drawn from sources like the interview guide’s topic areas, existing theory the study builds on, or prior fieldwork. They are a provisional starting point, not a fixed final coding scheme.

What is an integrative theme in template analysis?

An integrative theme captures a pattern that runs across multiple branches of the coding hierarchy rather than fitting inside a single branch — a relationship or tension the tree structure on its own would not otherwise represent.

How is template analysis different from framework analysis?

Template analysis produces a hierarchical coding tree that is revised iteratively; framework analysis produces a matrix, with cases as rows and themes as columns, that summarized data is charted into. Framework analysis is more procedurally fixed and is commonly used where several researchers need to code consistently against a shared structure, often in applied or policy research.

How is template analysis different from grounded theory?

Grounded theory deliberately minimizes prior theoretical input and aims to build a theory from codes and categories that emerge from the data. Template analysis explicitly permits and expects prior theory or research questions to seed the initial template, and its usual end product is a well-organized thematic account rather than a new theory.

How many levels should a coding template have?

There is no fixed number. The guidance is to add hierarchy levels only as far as the data actually supports distinguishing between them — a simple study might need two levels, a richer one several more — rather than adding depth for its own sake.

Who developed template analysis?

Nigel King, whose work on the approach was developed further with Christine Horrocks and, later, Joanna Brooks, in the organizational and health psychology qualitative-methods literature.

Related CASRAI Guides

• Grounded Theory: Building Theory From Data, Step by Step — the fully emergent alternative to template analysis’s a priori starting point.
• Thematic Analysis: A Step-by-Step Guide to Braun and Clarke’s Six Phases — the more generic six-phase coding process template analysis is sometimes confused with.
• Interpretative Phenomenological Analysis (IPA): From Transcript to Group Experiential Themes — a more procedurally structured, idiographic alternative for phenomenological questions.
• Qualitative Research Methods: The Major Methodologies and How to Choose — where template analysis sits among the broader set of approaches.
• Audit Trail in Qualitative Research: What to Keep and How to Present It — the general audit-trail practice a template version log is one instance of.
• Research Methods — the cluster hub for study design, sampling, and quantitative/qualitative analysis.