The core difference in one sentence
An experimental research design manipulates one variable under controlled conditions to measure a cause-effect claim across groups; a case study stays inside one bounded setting — a company, a ward, a school — and combines several evidence sources to explain how and why it works the way it does. If your question needs numbers that travel, run the experiment. If it needs depth in context, build the case study.
That sentence is the whole decision; everything after it is detail. The two designs are not rivals on a leaderboard where one scores higher — they answer different shapes of question. "Does this training raise test scores?" is a comparison, so it wants groups, a controlled procedure and a measured outcome. "How did this hospital cut its readmissions?" is an explanation, so it wants one setting, interviews, documents and observation, and room for context to carry part of the answer.
Students lose marks in two symmetrical ways. They force an experiment onto a how question, and end up with thin interviews plus a spreadsheet nobody asked for. Or they force a case study onto a whether question, and end up defending a single setting where the grader wanted evidence that the pattern repeats. Read your brief for its question shape before you touch the literature — the wording usually announces the design it expects, and section 5 of this page shows you exactly where to look.
Which method fits your question? The 5-question chooser
Answer the five questions below, one at a time. Each answer tips a live tally toward the case-study side or the experiment side, and after the fifth you get a band — case study territory, experiment territory, or a genuine split — plus a justification sentence assembled from your own answers, ready to copy into your methodology draft. It runs entirely in your browser.
Which Method Fits Your Question?
Five questions, one lean per answer — a mirror, not a verdict.
Question 1 of 5
Loading…
One warning before you trust the tally: the chooser mirrors your answers, so answer what your brief says, not what you would enjoy doing. A student who loves interviews can still be staring at a cause-effect question, and the tally will say so.
What counts as experimental research
An experiment is easy to recognize once you know its four moving parts, because every textbook definition is just those parts in different clothes:
- Variables. You change one thing — the independent variable — and measure another — the dependent variable. In plain words: the thing you do, and the thing you count afterwards.
- Groups. At least two: a treatment group that gets the change and a control group that does not, so the comparison has a baseline.
- Manipulation. You decide who gets what, usually by random assignment, instead of letting the world sort people for you.
- Control. Everything else stays steady — same room, same task, same timing — so any gap in outcomes can be pinned on the one change you made.
Here is the shape in a classroom-sized example, and like every example on this page it is illustrative, not a real study. Two seminar groups study the same topic for two weeks. One group revises with spaced flashcards; the other revises however it usually does. Both sit the same quiz on the same morning. If the flashcard group scores higher, you have a measured answer to "does spaced revision change quiz performance?" — because the change was introduced, the groups are comparable and everything else was held steady.
Now the negative test, which catches more students than the definition does: remove the manipulation and the design stops being experimental. Surveying students about their existing sleep habits is not an experiment, because you changed nothing — you measured what was already there. Waiting to see whether the university's new policy improves attendance is not an experiment either, for the same reason. No manipulation, no experiment, whatever the dataset looks like. Where the full write-up of such a design lives, and how to defend every choice in it, is the job of the research paper methodology section.
What counts as a case study
A case study answers a different kind of question by staying inside one bounded system — one company, one ward, one school, one campaign, one class — and explaining how it works from the inside. The boundary matters: "customer service" is not a case, but "the complaints desk at one regional bank between March and May" is. The boundary is what makes the study finishable and the claims defensible.
The second signature is multiple evidence sources, deliberately combined. You interview the two people who ran the change, read the policy documents and meeting notes, watch a shift or two, and pull the numbers the setting already keeps. Where the sources agree, your finding stands on triangulation; where they disagree, the disagreement is usually the most interesting paragraph you will write. A study that interviews one person and stops there is not yet a case study — it is one interview wearing the name.
Third, context carries weight. In an experiment, context is noise to be controlled away; in a case study, context is part of the finding, because "how" and "why" questions only have answers inside a situation. The ward was understaffed, the company had just merged, the school's policy dated from an earlier inspection — remove those facts and the explanation collapses. That is also why case study claims stay modest about travel: they describe this setting convincingly, not every setting in principle.
If this is the side your brief lands on, the writing side is covered by the how to write a case study walkthrough, and the methods toolkit behind it — designs, sampling, coding, triangulation — is the territory of the qualitative case study methodology guide.
Case study vs experimental research: the 7 differences
Here is the anchor table for the whole comparison. Screenshot it, or copy the row you need into your justification paragraph.
| Aspect | Case study | Experimental research |
|---|---|---|
| Question type | How or why does this work in its real context? | Does X change Y — and by how much? |
| Sample | One bounded case, chosen because it can teach something | Many participants, ideally randomly assigned to groups |
| Control | None — the setting is studied as it actually lives | High — conditions held steady, one variable manipulated |
| Data | Mixed by nature: interviews, documents, observation, existing numbers | Structured: tests, scores, counts, timed measures |
| Analysis | Themes and explanations, checked across sources | Statistical comparison between the groups |
| Write-up | Narrative: context first, evidence woven through, claims bounded | Report: design, results tables, size of the effect |
| Generalizability | Limited by design — depth replaces spread | Built for it — that is what the comparison is for |
Four terms the table leans on (flip each card)
Validity
The word graders reach for when a measure looks shaky.
In this comparison
Does your measure capture what the question actually asks? Experiments guard it with control; case studies argue it with converging sources.
Generalizability
Whether findings travel beyond the study.
In this comparison
Experiments aim for it through sampling and repetition; case studies replace it with depth and carefully bounded claims.
Triangulation
The case study's credibility move.
In this comparison
Checking one finding against two or more evidence sources — what stands in for a control group when no control group exists.
Manipulation
The experiment's signature move.
In this comparison
Deliberately changing one variable to watch its effect — the first thing ethics rules or a real-life setting usually rules out.
Keep the table beside you while drafting the whole paper, too: the full research paper walkthrough slots this decision into the design phase before a single paragraph is written.
When your assignment names the method: reading the brief's wording
Most briefs never write "experiment" or "case study" in so many words — the verb does the announcing. Marking teams choose their verbs the way the two designs choose their questions, so a two-line brief already tells you which design it expects:
| The brief says | It is asking for |
|---|---|
| "Investigate the effects of … on …" | An experiment — a change introduced and an outcome measured |
| "Explore how … responded to …" | A case study — explanation inside one setting |
| "Compare group A and group B on the same test" | An experiment, or at minimum a controlled comparison |
| "Examine the change process at [one organization]" | A case study, with documents and interviews |
| "Measure the relationship between …" | Numbers first — often a survey, since nobody is manipulated |
Two reading rules make the signal sharper. First, underline the object of the verb: if it is a named organization, ward or campaign, the case is already fixed and an experiment was never on the table. If it is a pair of variables — sleep and recall, price and demand, screen time and focus — a measurement design is waiting. Second, check the deliverable: "a report on the program" bends toward narrative; "test the hypothesis" bends toward comparison.
When the brief does name the method outright, stop litigating and follow it — the rubric was written for that design. The brief-decoding moves above work just as well on case-study assignments specifically, which is why the case study assignment guide starts from the same verb-spotting habit.
Mixed evidence: each design hides inside the other
Real research programs blur the line more than the table admits, and admitting it makes your justification stronger, not weaker. Case studies sit inside experimental programs all the time: a team runs a trial across twenty schools, then studies two of them up close to explain why the numbers moved — the case study explains what the statistics can only report. And experiments sit inside cases: a researcher studying one marketing team's rebrand can still run a small controlled test of two email subject lines within that setting. The bounded case frames the study; the mini-experiment supplies one kind of evidence inside it.
The fastest way to feel the difference is to take one question and run it both ways. Take this one, purely illustrative: does a ten-minute daily quiz routine lift final-exam scores?
What you would do
Split two matched seminar groups. One adopts the daily quiz for six weeks; the other revises as usual. Everyone sits the same final exam, and you compare the mean scores.
What you could claim
"The routine changed scores in these groups" — a cause-effect statement with numbers behind it, valid inside the sample you tested and cautious beyond it.
What you would do
Follow one course team that adopted the routine: interview the two instructors, collect the weekly quiz sheets, watch three sessions, read the department's policy notes.
What you could claim
"Here is how and why the routine took hold — or stalled — in this course" — an explanation in context, with no travel-ready number attached.
Same question, two deliverables, two kinds of claim. If your brief could happily accept either, that is not indecision — that is the mixed-methods lane, and the honest move is to name which strand answers the main question and which one supports it.
How to justify your method choice in one paragraph
Whichever side the tally lands on, a methodology section expects one paragraph that names the design and defends it against the obvious objection. Fill the brackets, copy it, and let it sit at the top of your methods section — it is also the sentence to echo when your introduction announces the design.
I chose [a case study / an experiment] because the brief asks me to [the brief's task in your own paraphrase]. A [case study / experiment] fits that question because [reason one: the shape of the question] and [reason two: your access, ethics or timeline]. The main cost of this design is [one honest limitation], which I manage by [how you limit the damage].
Replace the brackets with your course's specifics — the paragraph only works when the reasons are yours.
Notice the shape: claim, two reasons, one owned cost. That last part does more work than students expect, because graders read a limitation named early as judgment rather than confession. Where the paragraph sits inside the larger paper — and what surrounds it — is mapped by the research paper outline guide.
Common confusions, cleared
Five muddled pairs come up every semester. Each row is one sentence you can reuse if an examiner asks.
| The confusion | The cleared version |
|---|---|
| "A case study is always qualitative." | No — a case study can count things: usage logs, attendance figures, sales. The bounded case defines it, not the data type. |
| "An experiment is always quantitative." | Mostly, but observers can collect notes alongside scores, and the logic — manipulate, then compare — stays experimental either way. |
| "A study of one person can't be real research." | It can, when the single case is the point. Depth is a legitimate trade for spread, as long as you say that is the trade you made. |
| "A single-case design is just a case study." | Not quite — a single-case experiment still manipulates a treatment on and off with a protocol. The design is experimental logic wearing a small sample. |
| "An A/B test is a case study of the website." | It is a small randomized experiment: random split, one change, one measured outcome. Studying the rollout afterwards in context is what would make it a case. |
Method chosen, deadline still winning?
Deciding is the fast part; drafting is what eats the week. A quoted, deadline-bound writer can take the design decision and the draft off your hands — whichever side of the table your assignment lands on.
Quiz: which design is hiding in this brief?
1. A brief says "investigate the effects of late-night screen use on next-day focus." What is the wording asking for?
2. Which brief line is the strongest signal that a case study is expected?
3. One participant, a treatment switched on and off on a schedule, behavior measured in every phase. What is this?
4. "Every case study is qualitative." True or false?
5. An online shop shows half its visitors a new checkout page and measures completed purchases. What is this?
Frequently asked questions
Can one assignment use both a case study and an experiment?
Yes, if the brief leaves room. Researchers sequence them all the time: a case study can sit inside a trial program to explain its numbers, and a small pilot experiment can run inside a case to test one hunch. The discipline is saying which strand answers the main question, which one supports it, and in what order they ran — your methodology needs that map even when your tutor never asks for it.
My brief says case study but I would rather run an experiment — is swapping allowed?
The named method usually wins, because the deliverable and the marking rubric were written for it. There is often room to keep a small experimental element inside the case — a controlled test of one detail — but a full swap needs your tutor's sign-off before you collect any data. Swapping silently is the version that costs marks.
Is an interview-based study automatically a case study?
No. Interviews are a data type, not a design. A study becomes a case study when it examines one bounded setting in depth and combines several evidence sources — interviews plus documents, observation or existing numbers. A set of interviews with no fixed case is simply an interview study, and it should be labeled that way in your methodology.
How many participants does an experiment need?
There is no universal number a guide can hand you without inventing one. The honest version: the sample must be large enough to make the comparison you promise meaningful, your methods course or rubric sets the floor, and if your sample turns out small, the limitations paragraph says what that costs. Graders reward that sentence far more than a round number.
Where do I announce which method I chose?
Twice, briefly. One sentence at the end of the introduction names the design so the reader knows where the paper is going. The full defense then lives in the methodology — that is exactly where the one-paragraph scaffold earlier on this page belongs, with your two reasons and one owned limitation in place.
Did this settle the decision for you?