What is each comment actually asking you to change?
Supervisors write in shorthand. These are the comments thesis writers bring to a framework figure, and what each one normally means.
| The comment | What it usually means | What to change |
|---|---|---|
| "Too busy" / "too cluttered" | More boxes than your hypotheses need, or arrows that cross | Drop constructs you do not test; move control variables to a note under the figure; re-order boxes so no arrow crosses another |
| "Where are your hypotheses?" | The paths are unlabelled | Put H1β¦Hn on the arrows β one label per arrow, one arrow per hypothesis |
| "This is a theoretical framework, not a conceptual one" | You drew the published theory instead of your study | Keep your own constructs and the relationships you predict; name the theory in the text, not as the diagram |
| "The arrow is the wrong way round" | A path runs from the outcome back to the predictor | Flip it: arrows run from cause to effect |
| "That is not a mediator" | The mediator sits beside the predictor instead of on the path | Put it between the independent and dependent variable, with an arrow in and an arrow out |
| "How does the moderator work?" | The moderator is drawn as another predictor | Draw it pointing at the middle of the path it moderates, not at the outcome |
| "It doesn't match Chapter 2" | Names in the figure differ from the names in the text | Copy the construct names from your chapter, word for word, and regenerate |
| "It's not clear" | Often not a drawing problem at all | See when the figure is not the problem below |
Do them one at a time. If you send four changes in one instruction, you cannot tell which one caused the fifth thing that moved.
Two ways to revise: mark it, or rewrite the description
Mark it on the figure. Open the figure, draw on the part you want changed with the pen, arrow or text tool, write what should happen β "change this label to Perceived Usefulness", "make this arrow dashed" β and apply. Best for one or two local changes.
Rewrite the description. Edit the text that produced the figure: roles, directions, hypothesis numbers, and generate again. Best when the model itself changed β a construct added, a variable dropped, a mediator promoted.
The two routes are not equivalent, and for a conceptual framework the difference matters. Marking the figure hands it to an image model, which repaints the whole picture: parts you did not mark can shift, and your construct names are painted rather than typeset, so they can come back misspelled. Editing the description keeps the figure on the code renderer, where the names are set in a font exactly as you wrote them. For a wording change, edit the description; keep marking for visual changes such as line weight or box order. Either way, put the new version next to the previous one before you keep it β how Vizcept draws figures sets out which figure types are drawn which way.
Each generation or edit costs 5 credits at 2K and 8 at 4K. Budget for rounds, not for one perfect draft: among paying thesis authors on Vizcept, single framework figures have gone through 13 to 24 versions.
How do you keep the versions straight?
The Studio keeps every version of a figure β A1, A2, A3 β and you can open any of them again, which matters when round four turns out worse than round two.
A few habits that save a re-run:
- Keep the description, not just the picture. The text that produced the accepted figure is your real source file. Paste it into the same document as your framework notes.
- Name what you export for what it is:
Framework_v3_2026-09-20.png. Supervisors read several students' figures in a week. - Use the same figure number as the chapter (Figure 3.1, not "new framework final") so comments can point at it.
- Do not delete the rejected versions until the chapter is signed off. "Actually, go back to the earlier one" is a normal outcome.
What do you check after every round?
A round can introduce a mistake that was not there on the previous one β especially a round you made by marking the figure, which repaints it β so this is a per-round check, not a one-off:
- Every construct name matches your chapter exactly β spelling, capitalisation, no abbreviation the text never uses.
- Every arrow points from cause to effect.
- Each H label sits on exactly one arrow, and every hypothesis in the text has an arrow.
- Mediators sit on the path; moderators point at the path, not at the outcome.
- Nothing you did not ask for has changed β compare against the previous version, not against your memory of it.
If the figure passes all five, export it with the format your department asks for: putting the figure into Word, PowerPoint or LaTeX.
How do you reply when you send the new version?
Make the reply as short as the comments were, and answer every one β including the ones you did not act on. Something like:
Figure 3.1, v3. Changes: (1) removed the two control variables, now a note under the figure; (2) H2 and H3 labelled on the paths; (3) job satisfaction moved between leadership and performance as a mediator. Not changed: the construct names, which follow Section 2.4 β tell me if you would rather use the shorter labels.
This turns the next meeting into a decision about the model instead of a hunt for what moved.
When the figure is not the problem
If "it's still not clear" survives two rounds of redrawing, the model is what is unclear, and another version of the picture will not fix it. Three questions usually find it:
- Can you say each arrow as a sentence? "X increases Y, because β¦" If an arrow has no sentence, it has no hypothesis, and it should not be in the figure.
- Does every box appear in your research questions? A construct that appears in the figure but in no question is either a missing question or a box to delete.
- Are two boxes the same thing under different names? Constructs borrowed from two papers often overlap; your supervisor sees it faster than you do.
Get the sentences right first, then draw. How to develop a conceptual framework walks through the five steps from research question to diagram.
Questions about revising a figure
Will regenerating change parts of the figure I did not mark?
It can. Marking a change on the figure hands it to an image model, which repaints the whole picture every time, so compare the new version with the previous one before you keep it. Editing the description instead keeps a conceptual framework, syntax tree or sentence diagram on the code renderer, where only what you changed changes.
Can I go back to an earlier version?
Yes. Every version of a figure is kept in the Studio and can be reopened, so a round that makes things worse costs you nothing but the credits.
Does every revision cost credits?
Yes β 5 credits for a 2K figure and 8 for 4K, whether it is a new generation or an edit of an existing figure. Credit packs do not renew, so a thesis can be budgeted as a one-off.
Should I tell my supervisor the figure was drawn with AI?
Ask, and check your university's guidance first; many departments now expect a line in the figure note. Our guide to AI figures in journals and theses has a disclosure generator with a thesis option.
Can Vizcept decide my variables or hypotheses for me?
No. It draws the model you describe. Which constructs belong in the framework, and what you predict about them, comes from your literature review and your supervisor.
Revise the framework
Open your figure, mark what changes, and keep every version side by side.