Figure Quality Check Before Submission

This is why figure review belongs before final export. If a figure's role changes, the file work changes too.

In our pre-submission review work, figure problems usually fall into one of five patterns.

Reviewer cannot read it: the panel text is too small, the colors are too similar, or the labels require zooming.

Figure does not prove the claim: the abstract makes a strong claim, but the figure shows a weaker or different result.

Legend cannot stand alone: the figure only makes sense if the reviewer has just read the methods section.

Integrity risk is accidental: duplicated panels, unclear crops, undisclosed splicing, high contrast, or missing raw files create suspicion even when there was no misconduct.

Wrong figure order: the strongest evidence appears too late, while an attractive but secondary figure gets the first slot.

Those are not cosmetic problems. They affect editor trust.

Nature's image-integrity guidance says digital images should be minimally processed and correctly represent original data. Editors may request unprocessed data during manuscript evaluation. Nature also tells authors to check for duplicated figures, lane splicing, controls, and whether unprocessed scans match the figures.

Elsevier's artwork guidance emphasizes file format, resolution, naming, and separate artwork files. TIFF is recommended for pixel-based images, EPS is recommended for vector-based images, and figure captions may need to be listed in the manuscript file when no separate upload item exists.

PLOS figure guidance places responsibility for figure preparation and final quality on the author and provides tools and instructions to check figure presentation. Nature's final-submission guidance also gives standard figure widths, including single-column and double-column sizing.

The pattern is consistent: journals expect figures to be readable, compliant, and trustworthy before publication workflows begin.

For biological, clinical, and medical figures, check:

• whether representative images match the quantified result
• whether controls are visible and described
• whether scale bars are present where needed
• whether color maps are interpretable
• whether microscopy adjustments are disclosed
• whether blots, gels, and cropped images have raw-data support
• whether patient or participant data are de-identified
• whether panel order matches the results narrative

Nature's guidance on image integrity is especially relevant for gels, blots, microscopy, and any figure where processing decisions can change reader trust.

For plots, models, clinical endpoints, and statistical figures, check:

• axes and units
• sample size per group
• error bars and uncertainty
• statistical test labels
• legends for colors and symbols
• outlier handling if visible
• subgroup and panel consistency
• whether the figure implies causality the design cannot support

The figure should not make the result look cleaner than the study design allows.

For AI, computational biology, computer vision, and prediction-model papers, check:

• train, validation, and test split clarity
• external validation status
• confusion matrix or error examples where useful
• calibration or uncertainty where relevant
• failure cases, not only best cases
• dataset and subgroup representation
• code and data availability cues
• whether architecture diagrams explain decisions or only decorate the paper

An AI figure that shows only performance lift is often weaker than one that shows when the model fails.

Send the manuscript, every figure at submission size, figure legends, source-data tables if relevant, supplementary figures, target journal artwork instructions, and any raw-image notes for blots, gels, microscopy, or medical images.

If figures were assembled from multiple sources or software packages, include export formats and source files when possible. If a figure depends on third-party artwork or adapted content, include permissions status.

A useful figure quality check should include:

• figure-by-figure verdict
• one sentence naming each figure's job
• readability risk list
• legend fixes
• integrity or raw-data risk notes
• statistics and annotation issues
• file-format and upload concerns
• final order recommendation

The review should say which figure is doing too much and which figure is not earning its space.

The first figure has to earn its position. It should tell the editor what kind of evidence the paper is built around and why the claim deserves attention. A first figure that looks polished but does not carry the argument wastes the fastest moment of editorial attention.

For many manuscripts, the better first figure is not the most visually attractive panel. It is the figure that makes the central claim harder to dismiss.