How to Draw a Research Framework Diagram: Variable Model, IPO, and Overall Framework (2026)
Three ways to draw a conceptual framework for a thesis — the variable relationship model, the IPO input-process-output framework, and the overall research framework — with worked examples and the mistakes advisors send back.
"Where is your conceptual framework?" is the question that ends more proposal defenses than any methodological flaw. Prose can hedge; a diagram cannot. A research framework diagram compresses your variables, your logic, and your expected results onto one page, and whether that logic holds is visible in about ten seconds.
The trouble is that "research framework diagram" names at least three different figures. Most drafts get sent back not because the student cannot draw, but because they drew the wrong type for their study.
This guide covers the three you will actually be asked for — the variable relationship model, the IPO framework, and the overall research framework — when each one fits, what reviewers reject, and how to get a first draft out of a plain-text description.
What you'll learn:
Which of the three framework types your study needs
The specific errors advisors send back
One worked example for each type
How to generate a first draft from a paragraph of text
Which variables affect which, and through what path?
Management, psychology, education, public health
IPO framework
What goes in, what process runs, what comes out?
Information systems, capstone and system-development projects
Overall research framework
How do my problem, theory, content, methods and results connect?
Almost any thesis, especially at the proposal stage
The test is simple. If your study measures variables and tests hypotheses, you need a variable model. If it builds something, you need IPO. If it needs to show a committee how the whole study hangs together, you need an overall framework. Many theses include two: an overall framework in the proposal, a variable model in the methods chapter.
Davie Chen is a researcher at the Faculty of Animation and Intermedia, University of Arts in Poznan, studying generative AI for scientific figure creation, patent illustration, and manuscript drafting. SciDraw AI is one of the research-to-product tools built from this work.
Miles de investigadores usan SciDraw AI para crear en minutos figuras listas para publicar en artículos, solicitudes de financiación y envíos a revistas, sin experiencia en diseño.
This is the framework most quantitative theses need. It shows independent variables, the dependent variable, and the paths between them — including anything that sits on or alters those paths.
What it must contain:
Independent variables (IV) on the left, in separate boxes
Dependent variable (DV) on the right
Mediators on the path itself, between IV and DV
Moderators above or below, with an arrow pointing at the path they moderate, not at a variable
Control variables in their own box, usually bottom-left, unconnected
Hypothesis labels (H1, H2, H3) on the arrows, if your text numbers them
Worked example — technology adoption:
Independent variables: perceived usefulness, perceived ease of use. Mediator: attitude toward use. Dependent variable: behavioral intention. Moderator: trust, acting on the attitude → intention path. Hypotheses H1 through H4 on the four arrows.
Drawn correctly, a reader can restate your hypotheses from the figure alone. That is the actual standard.
What gets sent back:
A moderator drawn as another independent variable. A moderator changes the strength of a relationship. Its arrow must land on an arrow, not on a box.
Methods listed as variables. "Questionnaire survey" and "regression analysis" are not constructs. They belong in your technical route diagram, not here.
Arrows with no direction or meaning. Every arrow is a hypothesis. If you cannot say which hypothesis an arrow is, delete it.
No theoretical grounding. If your framework does not visibly come from a theory or from prior findings, expect to be asked where it came from.
Input-Process-Output frameworks are standard in information systems, engineering capstones, and system-development theses — and are especially common in Philippine and Southeast Asian programs. The figure is three boxes in a row, sometimes with an evaluation loop.
What goes where:
Input — what you gather before building: knowledge requirements (the domain rules, the standards, the legislation you must satisfy), software requirements, hardware requirements, respondent data
Process — the development methodology and its phases: requirements gathering, design, coding, testing, deployment, named after whichever model you used (agile, waterfall, RAD)
Output — the deliverable itself, stated concretely
Evaluation — an arrow from Output back to Input, if you evaluated the system and fed results back
Worked example — a barangay health record system:
Input: knowledge requirements (health records management, data privacy legislation); software requirements (PHP, MySQL, Bootstrap); hardware requirements (workstation, printer). Process: requirements gathering, system design, coding, testing, deployment, following the agile model. Output: a web-based barangay health record system. An evaluation arrow returns from Output to Input.
What gets sent back:
A process box that just says "development." Name the phases and the methodology.
An output that is a hope, not a deliverable. "Improved efficiency" is an outcome, not an output. The output is the system.
Mixing IPO with a variable model. If you find yourself putting a mediator inside a Process box, you have picked the wrong framework type.
This is the one-page figure that answers "how does your whole study fit together." Committees at the proposal stage often want this before anything else.
Structure, top to bottom:
Research background or problem — one line, stated as a problem, not a topic
Theoretical basis — the theories or models your study rests on, side by side
Research content — two to four blocks, each a sub-question you will actually answer
Methods and data — under each content block, the specific method and data source for that block
Expected results and contributions — what each block produces
The value of this figure is that it forces one-to-one alignment. Every research content block needs a method underneath it and a result below that. A block with no method attached is the gap your committee will find.
Worked example — digital transformation and SME performance:
Background: small manufacturers face digital transformation pressure with unclear returns. Theoretical basis: dynamic capabilities theory, resource-based view. Content: (1) measuring the level of digital transformation, (2) the mechanism from transformation to performance through dynamic capabilities, (3) the moderating role of environmental uncertainty. Methods: a survey of 320 SMEs, structural equation modeling, hierarchical regression. Expected results: a measurement scale, a mediation model, management recommendations.
What gets sent back:
Content blocks that are chapter titles. "Literature review" is not research content. Research content is a question you answer with data.
A theory layer that never reappears. If dynamic capabilities theory sits at the top but never touches a content block, the reviewer will ask what it is doing there.
Four methods for three content blocks, or none for one of them. Alignment is the whole point.
Dragging boxes in PowerPoint and rerouting arrows after every advisor comment is the slow way. The faster path is to write the framework as a paragraph, generate a draft, then refine.
Variable relationship model — list your IVs, DV, mediators, moderators and hypothesis numbers, and it places them on the correct paths
IPO framework — list what goes in each of the three columns, and it keeps them separate
Overall research framework — give background, theory, content blocks, methods and expected results, and it stacks them with the alignment visible
Pick the mode first. The prompt constraints differ per mode, which is why a variable model generated in the overall-framework mode comes out looking like an outline instead of a path diagram.
Two things worth knowing. The generator draws only the variables and relationships you give it — it will not invent a mediator to make your model look better, which is exactly what you want in a thesis figure. And if what you actually need is the method chain rather than the variable structure, that is a different figure: use the technical route mode instead.
Which framework type does my thesis need?
If you test hypotheses about variables, use the variable model. If you build a system or product, use IPO. If you are preparing a proposal defense and need to show the whole study, use the overall framework. A thesis can legitimately contain two of them in different chapters.
Is a conceptual framework the same as a theoretical framework?
They overlap but are not identical. A theoretical framework is the body of theory you adopt; a conceptual framework is your own diagram of how the constructs relate in your study. Most programs want the theory discussed in prose and the framework shown as a figure.
How many variables should a framework diagram have?
Enough to state your hypotheses and no more. Beyond roughly eight boxes, readers stop tracing paths. If your model is genuinely larger, split it: an overall framework figure plus a detailed variable model per sub-study.
Can I reuse the proposal framework in my final thesis?
Yes, and you should — a framework that changes silently between proposal and defense invites questions. Keep an editable version, update it when your model changes, and note the change in your text rather than quietly redrawing it.
