Design your method
You have a sharp question and you know what's already out there. Now for the plan: how will you actually answer it? Your method is your recipe — what you'll gather, from whom or what, and exactly how you'll do it.
Your method (or research design) is the game plan for answering your question. It spells out what information you'll collect, who or what you'll collect it from, and the exact steps you'll follow. A good method is specific and repeatable: someone else should be able to read it and run the same study. This is the step that turns an idea into something you can actually do.
Let your question pick the method
Don't start with “I want to do an experiment” and force a question onto it. Work the other way: your question already hints at the method that fits it best.
- Cause and effect? (“Does X change Y?”) → an experiment.
- A link or pattern? (“Is X related to Y?”) → an observational or correlational study.
- Attitudes or behavior? (“What do people think or do?”) → a survey.
- Depth and “why”? (“How do people experience X?”) → interviews.
- Already-collected data exists? → analyze an existing dataset.
Five ways to run a study
These are the main method families. Tap each to see what it's good for — and what to watch out for.
You deliberately change one thing and measure what happens, keeping everything else the same. It's the only method that can really show cause and effect.
Fits: “Do plants grow taller under red or blue light?”
Watch out: you have to control the other variables, or you won't know what caused the result.
You measure things as they naturally are — without changing anything — and look for relationships. Best when an experiment isn't possible or ethical.
Fits: “Is screen time before bed linked to shorter sleep?”
Watch out: a link isn't proof of cause — correlation is not causation.
You ask a group the same set of questions to capture their attitudes, habits, or opinions — usually turned into numbers.
Fits: “How do students at my school feel about later start times?”
Watch out: question wording, honesty, and who you ask can all skew the results.
Fewer people, far more depth. Open conversations that get at the why and how behind what people do.
Fits: “How do first-generation students experience applying to college?”
Watch out: it's time-intensive and harder to generalize to everyone.
Use data someone else already collected — public datasets, government records, sports stats. No fieldwork required.
Fits: “Has average city temperature risen over 30 years?”
Watch out: you're limited to whatever the dataset already contains.
The building blocks of any study
Whatever method you pick, three pieces make it real. Nail these and your method is basically written.
Your variables
The thing you change or compare (independent), the thing you measure (dependent), and everything you keep the same (controls).
Light color = what you change · plant height = what you measure · water & soil = kept equal.
Your sample
Who or what you'll actually study — and how you'll choose them fairly, so they represent the group you care about.
30 students picked at random beats “my 5 friends who agreed.”
Your procedure
The exact, numbered steps you'll follow — detailed enough that a stranger could repeat your study and get comparable results.
If a classmate couldn't run it from your notes, add more detail.
Is your method solid?
Run your plan through these six checks — tick each one it passes and watch the meter. The more it passes, the more you can trust your results.
Numbers or words?
Most methods lean one of two ways. Knowing which you're collecting tells you how you'll analyze it later, in Step 5.
Numbers
Counts and measurements. Answers “how much,” “how many,” or “is there a difference?” Analyzed with math and statistics.
Test scores, heights, survey ratings, response times.
Words
Descriptions and themes. Answers “why,” “how,” or “what's it like?” Analyzed by finding patterns in what people say.
Interview answers, open-ended responses, observations.
Plenty of strong projects use both — numbers to show what happened, words to explain why.
If your study involves people
Working with people (or animals) comes with real responsibility. Sort this out before you collect anything.
Ethics first
Ask for consent (people should know what they're agreeing to and be free to say no), keep responses anonymous or private, never pressure anyone, and avoid anything that could cause harm or distress. Studying minors, sensitive topics, or animals raises the bar further. Always run your plan past a teacher or mentor first — some studies need formal approval before you can begin.
Traps beginners fall into
- Changing too many things at once — then you can't tell what caused the result.
- A biased or tiny sample — five friends can't represent your whole school.
- Leading survey questions that nudge people toward the answer you want.
- No comparison or control group — you need something to compare against.
- Not writing the procedure down — if it's not recorded, it can't be trusted or repeated.
- Planning way too big for the time, access, or materials you actually have.
Pro tip · pilot it first
Before the real thing, do a tiny test run — try your survey on three friends, or run one trial of your experiment. You'll catch confusing questions, broken steps, and missing materials before they ruin your actual data. Ten minutes of piloting saves days of redoing.