What Action Research Actually Looks Like in a Math Classroom
Action research is just what it sounds like: you teach something, notice it isn't working the way you expected, tweak your approach, and check again. The cycle repeats until the data tells you something changed. It is not an elaborate academic exercise. It is a practical loop that takes one or two weeks to run properly, not a semester-long commitment. I have done this with remedial algebra, intermediate geometry, and a graduate-level discrete math course, and the structure stays the same even when the math changes. Here is a straightforward example that comes up a lot. You teach a unit on solving linear equations using the standard algorithm—distribute, combine like terms, isolate the variable. Midway through, you notice your formative quiz scores are sitting around 52 percent, and the errors are not random. Students keep distributing across addition instead of subtraction, or they drop negative signs entirely. The same three mistakes repeat across almost every problem set. That is your trigger to pivot. Instead of continuing straight, you design a small intervention. You spend two days having students write out each step in words before they write any symbols. You also introduce a color-coding scheme where terms being combined share a highlight. You run a second formative check after four class periods. If the error rate on those specific mistakes drops from roughly 60 percent to under 20 percent, you have evidence the intervention worked. If it does not, you adjust again—maybe you switch to concrete manipulatives or a different notation system. The point is that you are making decisions based on what the students actually do, not what the textbook assumes they will do.
I ran a nearly identical loop with a group of eleventh graders who were failing quadratic factoring. The diagnostic showed they understood multiplication but reversed the sign logic when extracting factors. I tried mnemonic devices first. They did nothing. I then shifted to a visual area model for three days, and the error rate on sign flips dropped from 41 percent to 12 percent in one week. That outcome was not guaranteed. The area model was slow for proficient students, so I had to let faster learners move on to practice sets while I pulled the struggling group back. Small group rotation during that intervention added about twenty minutes of prep time per lesson, which is real. You need to budget for that.
The Cycle, Broken Down Without the Textbook Language
You start by identifying a problem rooted in your own classroom. The problem should be specific enough that you can measure it. Vague problems like students do not understand algebra will not work. Specific problems like students miss negative sign conventions during polynomial division on weekly quizzes will. Then you collect baseline data. This is usually a quiz, a diagnostic, exit tickets, or observations over three to five sessions. Write down what you see before you change anything. Keep the numbers. When you later compare, you need that anchor point. Next you plan an intervention. This is where you decide what you will change. It could be a new instructional strategy, a different sequence of examples, a peer tutoring structure, or even just changing how you give feedback on assignments. The intervention should be something you can sustain for at least one to two weeks. Anything shorter gives you noise, not signal.
Get the Full Details
After that, you implement the intervention while collecting new data. Exit tickets are efficient here because you can grade them in ten minutes and see the trend immediately. You do not need sophisticated statistical software. A simple tally or percentage breakdown is enough for action research of this scale. Finally, you reflect and decide whether to continue, adjust, or move on. The reflection step is where most people rush. Do not skip it. Note what changed, what stayed the same, and what you would do differently next time. That note becomes your baseline for the next cycle.
What Beginners Miss About the Data Part
One thing people get wrong is treating any data as sufficient. If you only look at final test scores, you lose the detail that tells you why the intervention worked or failed. Track the errors by type. Track which students improved and which did not. Track the time it takes them to complete problems, not just whether they got them right. Speed and accuracy together reveal misconceptions that raw scores hide. Another trap is the sample size illusion. Running an intervention with thirty students feels substantial. It is not, if only eight of them actually engage with the new material. Participation matters more than headcount. In my experience, a focused group of twelve to fifteen students who are actively working through the intervention produces clearer results than a full class where half are disengaged. I once ran a three-week intervention with a full section and barely moved the needle because the data was too diffuse. I then switched to a smaller pull-out group and got usable results in five days. The trade-off is scheduling, which is never easy. But the data quality difference was stark.
A Hard Edge Case I Hit
There is a scenario that does not get discussed enough. When you are working with students who have significant gaps from prior years, the standard action research cycle can stall because the root cause is not the current lesson. It is a missing foundation three grade levels back. I encountered this with a geometry class where the problem appeared to be proof writing. We spent two weeks trying different scaffolds for two-column proofs. Scores barely moved. I finally pulled a small group and gave them a diagnostic on proportional reasoning and basic equation solving. Half of them could not isolate a variable cleanly. The proof issue was a symptom, not the disease. I paused the intervention, ran a two-week targeted review on algebraic manipulation embedded in geometric contexts, and then revisited proof scaffolding. That is when the scores shifted. The lesson was that the action research cycle assumes the problem is local. It is not always. You may need to widen the diagnostic lens before you trust the intervention data. Action research sounds rigorous but it can easily become decorative if you are not careful. Here is how to keep it honest. Set a clear success criterion before you begin. Not a gut feeling. A number. For example, reduce the error rate on a specific skill from 55 percent to below 30 percent within two weeks. If you do not define success upfront, you will interpret any tiny shift as a win and move on without really knowing whether your intervention did anything.
Limit the scope. One skill. One class. One or two weeks. When you try to fix three things at once, you cannot tell which change caused any observed effect. I once tried to improve both fraction understanding and decimal conversion simultaneously. The data became a mess. I ended the study, picked one skill, and got a clean result the second time around. Document everything in a way that lets you revisit it later. A single spreadsheet with dates, intervention type, error counts by category, and notes on student comments is enough. You do not need fancy tools. What you need is a record that survives past the moment you write it. I have lost data twice because I kept notes in a document that got overwritten. After that, I moved to a simple shared sheet that I updated daily. Ten minutes a day prevents a week of reconstruction later.
When This Method Fails Completely
Action research is not a universal fix. It breaks down when the problem is structural rather than instructional. If your students are absent frequently due to transportation issues, housing instability, or chronic health problems, no classroom intervention will move the needle on math scores. The data will look random because the variance is coming from outside the classroom. In those cases, the right move is not more cycles. It is referral to support services, attendance intervention, and administrative coordination. You can still collect data, but you should not pretend a teaching tweak will solve an attendance crisis. It also fails when the intervention is too vague to replicate. If your intervention is simply I tried being more engaging, you have nothing to measure. Engagement is not a method. It is an outcome. You need concrete actions: specific questioning techniques, structured peer review, timed practice sets, or explicit error analysis routines. If your goal is publishable, generalizable research rather than classroom improvement, action research is the wrong tool. You need a controlled study with larger samples, randomization, and peer review. Action research is for teachers who want to improve their own practice, not for building broad theoretical claims.
A Practical Template You Can Adapt
Week one: administer a short diagnostic, record baseline percentages by error type, and write a one-paragraph problem statement. Week two: implement the intervention, collect daily exit tickets, and log error trends in a spreadsheet. Week three: analyze the data against your success criterion, note which students responded and which did not, and write a brief reflection. If the criterion was met, consider whether to sustain the approach or generalize it to other topics. If not, identify the barrier and plan a revised intervention for the next cycle. The whole process usually takes about three hours of active work per cycle once you have it in your routine. The first time you run it, expect four to five hours because you are still building the habit. After that, it settles into something manageable alongside regular grading and lesson prep.
