1. Get feedback from others

We are experimental scientists and testing your graph out on viewers is the single most important thing you can do. If you forget everything else on this list, then at least show your graph to someone else who isn't too close to your project so that they have to really work out the message of the data through the graph alone.

2. Directly label features of the graph wherever possible

Pictures are great, but text is a power tool in your inventory. Directly label interesting points. Directly label data series. If you have to rely on a legend/key, so be it. But it's faster and easier for readers if you can directly label what they need to see.

3. Label the axes

The easiest way to interpret a graph is to understand the axes. Please label them, fully, clearly, and in useful units without lots of zeros.

4. Avoid relying solely on color

About half of people suffer from chromosomal deficiency that corelates with limited color differentiation skills. Even tetrachromats don't like having to distinguish teal from turquoise in order to understand the message. So, wherever possible, try to use multiple features to differentiate data, like direct labels, or darkness/lightness, etc.

5. Don't just do the defaults

Excel and other graph-making programs have default settings that aren't inherently bad, but also aren't always the best choice. Do you really need all those axes numerical intervals? Do you really need to show your units as $1,000,000 instead of maybe $1M? Do you need those grid lines? Think about your choices instead of letting them be made for you.

6. Say what your error bars represent

If your graph has error bars, please say what they represent: Standard deviation, standard error, confidence interval, etc. The interpretation is wildly different depending on which it is, so it's really a shame that we don't have a visual way to tell the reader. Make sure you let them know somehow.

7. Don't ruin the graph by inserting it into your final medium at low resolution

If at all possible, keep your graph as a vector graphic.

Don't screen-shot your graph to get it into Word/PPT.

Going into Word or PPT: use an SVG:

• From Excel: Regular copy/paste or if you don't want it to be easily editable anymore: paste special > Enhanced Metafile.
• From ggplot2 (R): Install svglite.
• From GraphPad Prism: File > Export, SVG.
• From Flourish.studio :Export & publish > Download image > Format > SVG
• From other: print to PDF and use Adobe Illustrator to convert into whatever formats you need.

Going into Google Docs: Don't.