Action Research: Cycle, Steps & Examples

What is action research?

Action research is a form of practitioner inquiry that deliberately blurs the line between researcher and participant. The defining feature is that the person studying the situation is also the person trying to change it. Kurt Lewin, usually credited as the founder, captured the spirit in a much-quoted line about practice and theory feeding one another, and he described research that proceeded through a “spiral of steps, each of which is composed of a circle of planning, action and fact-finding about the result of the action”. That image of a spiral — not a single loop but repeated, improving cycles — is the single most important idea to grasp.

Because it is rooted in real settings and real problems, action research is unapologetically practical. Its aim is not primarily to produce generalisable laws (as experimental research tries to) but to produce actionable local knowledge: a change that demonstrably improves practice in this classroom, on this ward, in this team, together with a transparent account of how and why it worked. That said, a well-documented action-research study still contributes to wider knowledge, because other practitioners can judge whether the lessons transfer to their own context.

“Research that produces nothing but books will not suffice.” (Source: Lewin, 1946)

It is worth situating action research against other approaches early. Where quantitative and qualitative research are often discussed as fixed paradigms, action research is better thought of as a stance or methodology that can draw on either or both — it is the cyclical, change-oriented, insider logic that makes it action research, not the data type.

The action research cycle: Plan → Act → Observe → Reflect

The engine of every action-research project is a four-phase cycle, repeated until the problem is adequately addressed or the dissertation timeline runs out. The classic phrasing comes from Stephen Kemmis and Robin McTaggart, who refined Lewin’s spiral into Plan → Act → Observe → Reflect.

1. Plan. Diagnose the problem from evidence, review what is already known, and design a specific, deliberate change (the “intervention”) together with a plan for how you will gather evidence about it.
2. Act. Implement the change in the real setting — teach the new way, run the new meeting format, introduce the new checklist.
3. Observe. Systematically collect data on what happens: outcomes, behaviours, reactions, unintended effects. Observation must be planned, not anecdotal.
4. Reflect. Make sense of the evidence. What changed? Why? What did not work, and what does that tell you? Reflection is where learning is crystallised and where the next cycle’s plan is born.

The reflect phase loops back into a revised plan, and the cycle turns again. Each turn is informed by the last, so the intervention becomes sharper and the researcher’s understanding deeper — hence a spiral rather than a closed circle.

Types of action research

Action research is an umbrella term, and your dissertation should name the variant you are using. The three most commonly distinguished — following a typology popularised by Wilfred Carr and Stephen Kemmis — differ mainly in who controls the inquiry and what kind of change is sought.

For most taught-masters and undergraduate dissertations, practical action research is the realistic choice: you study and improve your own practice within a manageable scope. Participatory action research (PAR) is powerful but demanding, because genuine co-ownership with participants takes time and raises extra ethical and logistical questions. Technical action research shades towards a small implementation study and can look like quasi-experimental work.

When to use action research

Action research earns its place when three conditions hold together: you are an insider with legitimate access to a practice setting; there is a genuine problem you want to improve, not merely describe; and change is something you can realistically introduce and observe within your timeframe. It is especially at home in:

• Education — a teacher improving engagement, feedback, behaviour, or attainment in their own class. This is the classic home of action research.
• Nursing and allied health — a clinician reducing errors, improving handover, or embedding a new care pathway on their ward.
• Organisational change and management — a manager improving team meetings, onboarding, or a workflow, often drawing on Lewin’s organisational-change roots.
• Community and social work — practitioners and service users co-designing a better service (typically PAR).

Conversely, do not choose action research if your question is purely descriptive or explanatory and you have no intention (or no mandate) to intervene. If you simply want to understand a bounded phenomenon in depth, a case study is a better fit; if you want to measure relationships between variables, a survey or correlational design is more appropriate.

The steps of an action research dissertation

The four-phase cycle is the conceptual core; a dissertation wraps practical steps around it. A reliable sequence is:

1. Identify and diagnose the problem. Ground it in evidence from your own setting (e.g. attendance data, marks, observation notes), not just a hunch.
2. Define a focused research question. Phrase it as an improvement question: “How can I improve…?” or “What happens to… when I introduce…?”
3. Review the literature. Find evidence-based interventions others have tried, and a theory to frame your change.
4. Secure ethical approval and access. Insider research needs careful handling of consent, power, and the dual role of practitioner-researcher (more below).
5. Plan Cycle 1. Specify the intervention, the success criteria, and exactly which data you will collect, from whom, and when.
6. Act and observe. Implement the change and gather your planned data throughout.
7. Reflect and analyse. Make sense of the evidence against your success criteria; involve a critical friend if you can.
8. Re-plan and run Cycle 2 (and beyond). Adjust the intervention in light of Cycle 1 and repeat. Two to three cycles is typical and defensible for a dissertation.
9. Draw conclusions and assess validity. State what improved, what the evidence supports, the limits of transferability, and the implications for your practice and others’.

Data collection within action research

Action research is method-agnostic: it can be qualitative, quantitative, or — most commonly — mixed methods, because combining numbers and narrative strengthens your claim that something genuinely improved. The principle is triangulation: collect from multiple sources so no single biased lens drives your conclusion. Common choices include:

• Observation — structured tallies (e.g. on-task behaviour every five minutes) and field notes.
• Reflective journal — the researcher’s own dated log; a signature data source in action research.
• Pupil/participant work and outcome data — marks, attendance, error rates, quiz scores.
• Questionnaires and short surveys — before-and-after attitude or confidence ratings.
• Interviews or focus groups — to understand why the change worked or did not.
• Critical-friend or peer feedback — an external check on your interpretations.

Qualitative data is usually analysed with thematic analysis, while simple quantitative data (counts, means, percentages) is reported descriptively. Whatever you collect, plan it before the cycle starts so the “observe” phase is systematic rather than retrospective.

A full worked example: improving Year 9 engagement over two cycles

The example below shows how the cycle plays out concretely, end to end, in a mixed-methods education dissertation.

Notice what makes this action research rather than a one-off experiment: the second cycle’s plan is built directly from the first cycle’s reflection, the researcher is the practitioner, and the goal is a real improvement in a real classroom, evidenced transparently.

The dos and the don’ts

Because action research is so close to everyday practice, it is easy to do casually and hard to do rigorously. Keep these in view.

Do:

• Anchor your problem in baseline evidence before you intervene.
• Plan more than one cycle — a single loop is barely action research.
• Define success criteria in advance so “it improved” is measurable, not wishful.
• Triangulate across multiple data sources to offset your insider bias.
• Keep a dated reflective journal and an audit trail of decisions.
• Use a critical friend or peer to test your interpretations.
• Be transparent about your dual role as practitioner and researcher.

Don’t:

• Don’t treat it as an excuse to skip rigour — “just trying something” is not research.
• Don’t over-claim generalisability; your findings are context-bound by design.
• Don’t ignore disconfirming evidence because you want the intervention to work.
• Don’t bolt “reflection” on at the end — it must shape the next cycle.
• Don’t neglect the power imbalance when your participants are your own pupils, patients or staff — consent must be genuinely free.
• Don’t run so many cycles or such a broad intervention that you cannot finish in time.

Strengths and limitations

Action research has real strengths but well-known weaknesses you must address head-on in your methodology chapter.

Rigour and validity in action research

The commonest examiner challenge is “how do I know this isn’t just you convincing yourself?” You answer it with explicit rigour. Action researchers often draw on Herbert Altrichter and colleagues, and on validity criteria associated with Bridget Somekh and with John Elliott, but the practical moves are clear and you should also discuss them in the language of reliability and validity:

• Triangulation — converging evidence from several sources and methods.
• An audit trail — dated records of decisions, raw data, and analysis so others can follow your reasoning (dependability).
• Critical-friend and peer scrutiny — inviting external challenge to your interpretations.
• Member checking — testing your readings with participants where appropriate.
• Reflexivity — openly examining how your insider position shapes what you see, rather than pretending to neutrality.
• Thick description — describing the context richly so a reader can judge transferability to their own setting.

Framed this way, action research is not “soft”: it trades the laboratory’s control for situated authenticity and defends its claims through transparency and triangulation rather than randomisation. For broader context on how it sits among other designs, see our overview of the types of research.

Frequently Asked Questions