The following is the checklist I use when making explanatory figures (including tables), which is a special class of figures. Other kinds of figures include “exploratory” and “schematic”, which have somewhat different guidelines, and will be discussed elsewhere.

  1. Main Point The main point of the figure “pops-out” to your audience (note: addressing this successfully requires knowing precisely what the main point is, and who your audience is). This is by far the most important property of any figure, until you figure this out, there is no reason to do anything else.
  2. Data to Ink Ratio All data presented are critical to the main point (meaning none can be removed or summarized elsewhere). See Tufte’s “data to ink ratio” principle, for example, here. In other words, is there any way to reduce cognitive load for the audience, and still make the same point, for example, by removing unnecessary:
    • borders,
    • gridlines,
    • data markers,
    • legends (by labeling data directly),
    • if the data are 2D, are you displaying it in 2D? (If not, remove that additional 3rd dimension. It is just confusing and obfuscatory.)
  3. Color Use a clear consistent color palette.
    • I recommend one of the following:
    • Color use is consistent (within and across figures), meaning if a given algorithm uses blue as its color in one figure, it should use blue for that algorithm in all other figures, and blue should not be used for anything else.
    • Is the default color probably gray, so that black can be used for emphasis?
    • Make red the color of the line that you want the audience to attend to most. For example, upon introducing a new algorithm, and comparing to others, color the new algorithm red.
  4. Title The title provides context (e.g., sample size, dimensionality, dataset/simulation name, etc.), and possibly conclusion.
  5. Axes
    • All axes are labeled (with units).
    • Show the minimal possible number of numbers on the x- and y-axis. For example, if the units are not interesting/relevant, do not display numbers at all. If they are, try only two or three numbers.
    • If it’s a log scale, the axis label says so.
    • Are the number of significant digits reasonable (eg, the number of significant digits cannot reasonably be larger than the sample size)?
    • Axis labels should be words, not symbols, not abbreviations, so one need not read the caption to understand the figure.
    • Are your axes ‘tight’ (that is, are the bounds of the axes just larger than the max and min of what you want to show)? If not, do you have good reason for the excess?
    • Is the aspect ratio correct? (Hint: if you rescaled both the width and height separately, almost definitely not.)
  6. Text
    • All font sizes are legible for the typical reader. Specifically, all fonts sizes (including axis labels, numbers, titles, etc.), should be the same size as the caption in the text. Note that there is basically no advantage to smaller fonts, but there is a huge advantage to clarity, especially for readers with poor eyesight (which is a larger fraction of the readers).
    • All the letters/numbers are fully visible (i.e., not obscured by part of the figure).
    • All lines are labeled (e.g., in legend) with clearly different colors/styles.
    • Is your method named something other than ‘proposed method’ or ‘our method’? If not, name it and use it throughout.
  7. Caption
    • Begin with a sentence (fragment) stating what the figure is demonstrating (i.e., why it is there). If the figure consists of multiple panels, one sentence explains the collective take home message of all the panels.
    • Define all acronyms used in the figure.
    • When relevant, state the sample size, dimensionality, and statistical test / procedure.
    • End by pointing out particularly interesting aspects of the figure that one should note. - Is below the figure.
  8. Lines & Markers
    • If there are multiple lines/dots, is each a different line style and color?
    • Are all lines sufficiently thick? (If you used matlab, and they are the default thickness, the answer is no.)
    • Are all markers clearly different?
    • If errorbars make sense, are they there? If there, does the caption explain whether they are standard error? If they are not there, is there a good justification provided for that?
  9. Aesthetics
    • Are all graphics that can be vector graphics actually vector graphics?
    • Is anything not aligned that could be aligned?
    • If it’s a bar chart, does the y-axis start at zero (if it is a log axis, the answer is no)?
    • If it’s a pie chart, can you replace it with a stacked barchart (or something else)?
    • If you are comparing multiple approaches across multiple settings, are you grouping by the key comparison (eg, usually group by setting to make comparisons easier)?
    • Is there sufficient whitespace?
  10. Multipanel figures
    • Can certain axes/labels be removed because they are redundant?
    • Does the caption specifically mention each?
    • Are the captions for each collected together in the overall figure caption?
  11. Table
    • Can it be converted to a figure? If so, do it (put table in appendix if you are scared of losing the information)!
    • Are the rows sorted in a reasonable fashion (ie, according to the most important column)?