Template Analysis: a coding template you are meant to revise

Three features distinguish template analysis from its neighbours.

A priori themes are permitted and declared. A small number of themes may be defined before analysis begins, from theory or from the research question, provided they are stated openly and — this is the condition that matters — are open to being modified or discarded. A priori themes that survive unchanged in every study that uses them are functioning as assumptions rather than as analysis.

The template is built on a subset, then applied and revised. The initial template usually comes from a quarter to a third of the transcripts, and is then applied to the rest. Each transcript that does not fit prompts a change to the template. The final template is a product of the whole dataset even though it started from part of it.

Hierarchy is used for depth where it is warranted. Template analysis permits several levels of nesting, and does so unevenly on purpose: the themes central to the research question get elaborated to three or four levels, and peripheral ones stay at one. That asymmetry is a feature and is what makes a template readable as a statement of what the study found important.

Template analysis was developed by Nigel King, primarily in organisational and occupational psychology, where research questions are usually well specified in advance and the pretence of a blank slate is unhelpful. His practical account is Doing Template Analysis in Qualitative Organizational Research.

Brooks, McCluskey, Turley and King set out the case for the method and its relationship to other thematic approaches in The Utility of Template Analysis in Qualitative Psychology Research, which is the best single starting point.

A worked application is available in Brooks and King's Doing Template Analysis: Evaluating an End-of-Life Care Service, and King and Brooks extend the treatment to management research in Template Analysis in Business and Management Research.

A study of a new performance-appraisal system in a professional services firm interviewed 24 managers. Three a priori themes came from the literature and the commissioning question: perceived fairness, time cost, and relationship with existing informal feedback.

The initial template, built on six transcripts, elaborated fairness to three levels (procedural, outcome, comparative) and left time cost at one. It also added an inductive theme the a priori set had missed: what managers thought the system was for, which varied wildly and turned out to be the study's central finding.

By transcript fourteen the template had changed three times. "Relationship with informal feedback" was split, because two distinct things were being described — managers who used the system to formalise conversations they were having anyway, and managers who maintained two parallel systems and resented it. Merging them had been concealing the distinction.

By transcript twenty, "time cost" had been deleted as a top-level theme and re-sited as a sub-theme under purpose, since almost every complaint about time turned out on close reading to be a complaint about spending time on something whose point was unclear. That move is what template analysis is for, and it would not have happened had the a priori themes been treated as fixed.

The final template had four top-level themes with an integrative theme — "performing compliance" — running across three of them. The report was organised around the argument rather than around the template.

• A priori themes that never change. If the final template contains the initial themes untouched, the data has been used to illustrate a position rather than to test one.
• Too many a priori themes. Beyond about five, the template constrains rather than orients.
• Freezing the template early. Fixing it after the subset and applying it mechanically converts the method into deductive coding.
• Uniform hierarchy. Elaborating every branch to the same depth throws away the signal that uneven depth carries.
• No version history. Without dated templates there is no evidence that revision happened, and revision is the method.
• Reporting the template as the findings. The template organises the analysis; it is not the analysis.
• Choosing the subset for convenience. A template built on the five easiest transcripts will need rebuilding.

The specific thing to look for in a template analysis is evidence of revision: successive dated templates, and an account in the write-up of what changed and why. A study that shows only its final template has not demonstrated that the method was followed.

Beyond that: are a priori themes declared with their source? Is the hierarchy uneven in a way that reflects the research question? Are the subset transcripts chosen for variety and the choice justified? And is the results section built around an argument rather than around the template's structure?

The a priori themes are a real risk however carefully they are handled. Starting with a position makes it easier to find, and the method's safeguards are procedural rather than structural — they depend on the analyst genuinely being willing to discard a theme they arrived with.

Template analysis is also weaker than framework analysis for case-level comparison: the hierarchy organises themes, not cases, and reconstructing a single participant's whole account from a coded template takes work the matrix would have done for you.

And like framework analysis it is descriptively oriented. It organises and structures well; it does not generate theory, and studies that need to should look to grounded theory instead.

Framework analysis shares the structured, revisable-framework logic and adds the case-by-theme matrix.

Directed content analysis starts from theory in a similar way but is usually more rigidly deductive and often quantifies.

Reflexive thematic analysis is the alternative where the study wants interpretive depth and treats the researcher's position as a resource rather than a risk to manage.

Constructivist grounded theory is the right choice when the aim is to build theory rather than to organise data against an existing one.

Template analysis is hierarchical coding with a template that keeps changing, which is precisely the operation that is expensive by hand and cheap in a tool. Splitting a theme at transcript fourteen means every passage already coded to it has to be revisited; moving a theme to a different level means the whole tree updates.

Version history matters too: the sequence of templates is the audit trail, and reconstructing it afterwards from memory is not the same thing. Evidano supports template analysis as a named methodology, holding the hierarchy and re-coding cleanly when it changes. Deciding that a theme is doing too much work and must be split is analytic judgement, and it is the decision the method exists to make you take seriously.