Thematic Analysis for Action Research: Steps and Sample

Interviews, focus groups, journals and open-ended survey answers give an action researcher pages of words. Thematic analysis turns those words into a few themes that answer the research question. The Department of Education’s (DepEd) Action Research Toolkit sets out five steps for it in section 5.4, with a worked example from ten Grade 1 learners and a practice exercise for the next learning action cell (LAC) session. Below are the steps with the toolkit’s example, how to count and rank themes in Excel, the practice exercise solved, and how the steps compare with Braun and Clarke’s widely cited six phases of thematic analysis.

It is written for teachers analyzing qualitative data for action research, and for school heads and research committees who review the results.

Key points

• Qualitative analysis is not the easy option. The toolkit warns that it is “more arduous and demanding of one’s critical thinking and time” than statistical analysis.
• Five steps: prepare the data, get to know it, develop themes, classify the responses by theme, and summarize the results in a table.
• Three to five themes, each tied to the research question. The toolkit says more than five “tend to be unwieldy”; broaden them instead.
• A response can carry more than one theme, so frequencies can add up to more than the number of respondents. One that fits no theme goes into a revised theme or into “Others”.
• Check your themes with a peer. The toolkit recommends peer debriefing: a colleague codes the same data, and you compare.
• Counts help, but they are not the finding. Report each theme with what it means and the participants’ own words.

Codes and themes

A code is a short label for one idea in a piece of text: “learning to read”, “friends”, “baon”. A theme is a pattern across many codes and participants that says something about the research question: “School as a place to learn”. The toolkit calls the first stage of this open coding, “probing the data and extracting main thoughts”, and quotes Sutton and Austin (2015) on theming: “drawing together codes or key terms from one or more transcripts to present the findings of qualitative research in a coherent and meaningful way”.

The five steps, with the toolkit’s example

The toolkit’s example is hypothetical data from ten Grade 1 learners asked one question: Ano ang nagugustuhan mo sa pagpasok sa paaralan? (What do you like about going to school?)

Step 1. Prepare the data

Give each participant or document an ID, note their background, transcribe interviews and observations word for word, and do it the same day, while unclear notes can still be clarified. The toolkit suggests typing the data into a word processor:

Step 2. Get to know the data

Read everything, more than once, before coding. The toolkit stresses the researcher’s “tacit knowledge” of the setting and the participants, and asks you to “unearth” ideas that some responses state outright and others only imply. Break longer answers into chunks, each carrying one idea.

One-line answers are already chunks. A longer interview answer is not. Here is an invented excerpt from a teacher interview about a daily letter-sound routine, with each chunk coded:

“At first my Low Emerging readers would not read aloud; they just copied their seatmates. [reluctance to read aloud] After the second week some started sounding out words on their own. [decoding independently] What helped most was doing it in the first five minutes of every period, because they knew it was coming. [predictable routine] But the take-home cards rarely came back. [home follow-through weak] Many parents leave for work very early. [parents’ work hours]”

Step 3. Develop themes

From a first reading, or from a sample of the responses, draw out the main patterns and give each a short label. For the Grade 1 data, the toolkit’s themes are:

Then go back to the research question. The toolkit is firm on this: “Themes that are not related to the research question should be either eliminated or revised.” Aim for three to five themes; three already give “substantial material to discuss as findings”.

Step 4. Classify the responses

Return to each response and mark it with the label of its theme, by underlining key words and writing the label beside them, or with highlighter pens, one color per theme. The toolkit’s classification:

Two rules from the toolkit decide the hard cases:

• A response that fits no theme either leads you to revise the themes or goes into “Others”. S4, about baon, went to “Others”.
• A response with two themes is split, and each part is counted under its theme. S6, “My classmates and I have fun playing”, is both an activity and a social one. The toolkit counts it once, under Activities; coded under both, it would make the total 11 for 10 learners.

Responses that do not answer the question at all, such as a learner who describes the trip to school, are set aside: the toolkit calls this filtering. Mention them in the findings or the reflection, though, because what learners say unprompted can matter.

Step 5. Summarize in a table

The toolkit’s summary table (Table 5.4.1) gives each theme, a sample response, the frequency, the percentage and the rank:

Activities and Social aspect tie for second and third place, so each gets the average rank, 2.5.

Counting and ranking themes in Excel

The toolkit uses a word processor; a spreadsheet does the counting for you.

1. Put the IDs in column A, the responses in column B, the first theme label in column C and, for responses with a second theme, the second label in column D.
2. List the labels in F2:F5 (Learn, Act, Soc, Others). In G2, count each label across both columns: =COUNTIF($C$2:$D$11, F2).
3. In H2, the percentage of learners who mentioned the theme: =G2/COUNTA($A$2:$A$11), formatted as a percentage.
4. In I2, the rank, with ties averaged as in the toolkit: =RANK.AVG(G2, $G$2:$G$5).

Copy the formulas down. With S6 coded under Activities only, the counts are 5, 2, 2 and 1 and the ranks 1, 2.5, 2.5 and 4, as in the toolkit. With S6 under both Activities and Social, Social rises to 3 learners (30%) and second place, and the percentages add up to 110%. That is correct: they are percentages of learners, and a learner can mention two things. Say so under the table.

The toolkit’s practice exercise, solved

Section 5.4 ends with an exercise: fifteen Grade 9 students answered, after a Leadership Training Program, “What changes did you see in yourself as a result of attending the Leadership Training Program?” The toolkit gives no answer. Here is one reasonable analysis, with three themes:

Sample responses:

• Mean: “Akala ko ang leadership ay may kinalaman lang sa pagbibigay ng utos; ngayon alam ko na hindi pala.” (I thought leadership was only about giving orders; now I know it is not.) (15)
• Self: “I realized that I still have a long way to go to be a leader.” (03)
• Grow: “Kahit ’di pa naman ako leader sa ngayon, feeling ko gusto kong maging isang leader balang araw.” (Even if I am not a leader yet, I feel I want to be one someday.) (05)

Judgment calls to discuss in the LAC session:

• Responses 12 (“I have to lead by example”) and 13 (“I must be willing to sacrifice myself for the sake of others”) state what leadership requires, so they are under Mean; read as personal commitments, they could go under Self.
• Response 14 (“Jesus is my example of how to be a good leader”) names a role model rather than a change. It fits Mean, alongside 12 and 13, as an idea of servant leadership; a stricter reading puts it in Others.
• Response 10 (“It was good for us to learn how to be a good leader so that we can be more effective”) is close to an evaluation of the program. Here it counts as a skill learned.

A colleague may well build different themes from the same fifteen answers. That is the point of the toolkit’s peer debriefing: compare, discuss the differences, and keep or adjust your themes.

The six phases of Braun and Clarke

Outside DepEd, the most widely cited guide to thematic analysis is Braun and Clarke (2006), with six phases. They match the toolkit’s steps closely:

Two of their points are worth adding to the toolkit’s steps. First, the analysis moves back and forth between the phases; it is not a straight line. Second, a theme’s importance does not depend only on how many people mention it, but on whether it captures something important for the research question. In the Grade 1 data, a single learner’s answer about baon may say more about attendance than the five answers about learning.

Inductive or deductive? Themes built from the data, as in the toolkit’s example, are inductive. Themes set in advance from a theory or framework, such as coding teachers’ answers by the components of a reading program, are deductive. Many action research studies combine the two.

Making the analysis trustworthy

Qualitative research is judged by its trustworthiness, a term from Lincoln and Guba (1985), rather than by statistical tests. Practical ways to build it in an action research:

• Peer debriefing, as the toolkit recommends: a “disinterested but equally competent” colleague examines your themes, or codes the data without seeing yours.
• Member checking: read your summary of an interview back to the participant and let them correct it. The toolkit’s interview guide asks for this at the end of each interview.
• Triangulation: compare what interviews say with observations, documents, and test scores.
• An audit trail: keep the transcripts, the coded copies and notes on why you merged or renamed themes.
• Thick description: describe the school, the learners and the intervention well enough for readers to judge whether the findings fit their own setting.
• Reflexive notes: write down what you expected to find. Teachers studying their own classes have a stake in the result.

How to report the findings

Present each theme in the order of the summary table: name it, define it in one or two sentences, give how many participants mentioned it, and quote two or three responses with their IDs, translated, with the original in the text or an appendix. For the Grade 1 example:

Half of the ten Grade 1 learners (5 of 10) liked school because they learn there, from learning to read (S8) to having lessons explained (S5, S10). Two mentioned fun activities (S6, S9) and two their friends and their kind teacher (S2, S7). One said that on school days there is baon (S4), a reminder that for some learners, coming to school also means a meal.

In a mixed-methods study, place the themes next to the numbers they help explain: for example, why some learners’ scores rose and others’ did not.

Common mistakes

• Too many themes, each with one or two responses. Merge them into three to five broader themes.
• Themes that only repeat the interview questions, such as “Question 1 answers”. A theme is a pattern in the answers.
• Dropping responses that do not fit. Revise the themes or use “Others”, and report it.
• Counting only. A table of frequencies without definitions and quotes is not a qualitative finding.
• Quoting learners by name. Use IDs.
• Pasting transcripts with personal data into AI tools. DO 003, s. 2026 prohibits entering personal information into AI systems, and any AI-suggested codes still need your own checking against the data.

Frequently asked questions

What is thematic analysis in action research?

A way of analyzing words, such as interview answers, focus group transcripts, journals and open-ended survey responses, by coding them and grouping the codes into a few themes that answer the research question. DepEd’s Action Research Toolkit describes it in five steps in section 5.4.

How many themes should a qualitative study have?

DepEd’s Action Research Toolkit suggests three to five. Three give enough to discuss; more than five become unwieldy, and the toolkit recommends broadening the themes if there are more.

What is the difference between a code and a theme?

A code labels one idea in one piece of text, such as “learning to read”. A theme is a broader pattern across many codes and participants, such as “School as a place to learn”, that helps answer the research question.

Can I use percentages in qualitative analysis?

Yes. The toolkit’s summary table shows frequencies, percentages and ranks. State what the percentages are of, usually the participants, and note that they can add up to more than 100% when a response has more than one theme.

What is peer debriefing?

Asking a colleague who has no stake in the study, but who knows research, to review your themes, or to code the same data independently so that you can compare. DepEd’s toolkit recommends it to validate a qualitative analysis.

More on DepEd research

• Start here: DepEd Action Research Guide: From Topic to Report
• Interview and FGD Guide Questions for Action Research
• Action Research Questions and Hypotheses, With Examples
• DepEd Action Research Report Format, With a Sample
• DepEd Action Research Ethics: Consent, Assent and Privacy
• Weighted Mean and Likert Scale Interpretation in Excel

Sources

• Department of Education and the Basic Education Sector Transformation (BEST) Program. (2019). Action Research Toolkit: Teacher’s Guide, sections 4.3 (data collection through interviews) and 5.4 (qualitative data analysis), with support from the Australian Government.
• Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101. https://doi.org/10.1191/1478088706qp063oa
• Sutton, J., & Austin, Z. (2015). Qualitative research: Data collection, analysis, and management. The Canadian Journal of Hospital Pharmacy, 68(3), 226–231. https://doi.org/10.4212/cjhp.v68i3.1456
• Lincoln, Y. S., & Guba, E. G. (1985). Naturalistic inquiry. Sage.
• DepEd Order No. 003, s. 2026, Foundational Guidelines on Artificial Intelligence in Basic Education

Written by Mark Anthony Llego for NASH Reviewer. NASH Reviewer is not affiliated with DepEd.