General best practices
Throughout the data visualization process, keep in mind some general best practices.
What is your purpose--exploratory or explanatory?
A main point to consider is if your visualization is meant to be exploratory (something you are doing to get more familiar with your data) or explanatory (you know your result and want to communicate it to your audience). See here for additional examples of explanatory vs exploratory visuals.
For an example that applies her five best practices in a viz showing philanthropic impact, with before and after views, see Cole Nussbaumer Knaflic's blog post Lessons from GMN.
Preattentive attributes
Our brains use certain characteristics like color, shape, and position to understand the relationship between different graphic elements. Do these points represent the same type of data? How are they different from each other? Make the most of these visual characteristics, but don’t overdo it or the user can be overwhelmed. Emphasize attributes in this order of importance:
- Position
- Color
- Size
- Shape
Overcome memory limits
Even the best visualizations can tax the user’s memory limits. There are a number of ways to prevent this:
- Use familiar chart types
There's a reason pollsters and newspapers use bar charts a lot--readers don't have to invest time figuring out the results. - Don't make people remember views
It makes decoding your viz harder and more time-consuming than it needs to be. - Avoid large legends
Making legends part of your title, or labeling graphic elements directly, may reduce the work needed on the part of the user. - Use intuitive colors and shapes
As Yost points out, the traffic light metaphor of red means stop and green means go does not need to be explained, nor does red versus blue in U.S. political polls.
A related best practice, introduced by Edward Tufte, is to maximize the Data-Ink Ratio.
In his 1983 book, The Visual Display of Quantitative Information, he suggested:
A large share of ink on a graphic should present data-information, the ink changing as the data change. Data-ink is the non-erasable core of a graphic, the non-redundant ink arranged in response to variation in the numbers represented.
Use the 5 second test
If users can’t figure out your viz within 5 seconds, rethink your approach. Here are some ways to pass the test, using Andy Cotgreave's viz on road fatalities as a model:
- Position the most important view goes on top or top-left
- Position legends near their views
- Avoid using multiple color schemes on a single dashboard
- Use 5 or fewer views in dashboards
- Provide interactivity
- Carefully word Titles, Axes, and Units
- Highlight key facts and figures
Ten simple rules for better figures, Nicolas Rougier, Michael Droettboom, Philip Bourne, PLoS Computational Biology, 2014
This article offers brief "do" and "do not" guidelines for making effective scientific figures and shows why they're useful. The first rule--Know your audience.
DataViz cheatsheet by PolicyViz
This cheatsheet briefly summarizes standard considerations for data visualization. A nice reference guide to keep in mind for each data visualization project.
For further information, see these U-MN Libraries ebooks:
Storytelling with Data: A Data Visualization Guide for Business Professionals
by Cole Nussbaumer Knaflic
Information Visualization: Perception for Design
by Colin Ware
The Big Book of Dashboards: Visualizing Your Data Using Real-World Business Scenarios
by Steve Wexler, Jeffrey Shaffer, Andy Cotgreave
Fundamentals of Data Visualization
by Claus Wilke
Data points: Visualization That Means Something
by Nathan Yau
Choosing the right chart type
The first step in effective data visualization is making sure you're using the right chart/graph type for your specific data. Here are some good overviews of your choices and why certain ones make more sense in certain situations.
Consider your underlying question
Jami Oetting, Hubspot
This overview about choosing the best type of graph for data frames the issue in terms of underlying statistical question. For instance, are you comparing values? Showing composition? Looking for trends? Interested in distribution? What you're trying to understand from your data can and should inform the type of graph you use to visualize and communicate the results.
Selecting the right graph for your message
Stephen Few, Perceptual Edge
Data analyst Yan Holtz and designer Conor Healy 
This guide frames the decision of which chart type in terms of data type: numeric, categorical, mixed numeric and categorical, maps, network, or time series. Click through to read more about caveats associated with each type of chart.
Jennifer Lyons and Stephanie Evergreen, author of Presenting Data Effectively
Working with qualitative / non-numeric data can pose a set of different issues. The Qualitative Chart Chooser frames chart choice by what "story" you want to tell with your data.
Larry Silverstein, "From "huh?" to "a-ha": The science of data visualization," Tableau Conference 2017 (slide 52)
Time: on an x-axis
Location: on a map
Comparing values: bar chart
Exploring relationships: scatter plot
Relative proportions: treemap
To encode trends over time, use a line graph instead of a bar chart.
Nathan Yau, Flowing Data
This post provides examples of good and bad charts, distilled into a handful of general rules for making charts. For instance, be mindful of the starting point of the y-axis on a chart so that data representation is truthful to the underlying data trends. For example, he shows the dangers of overusing pie charts:
Choosing colors
Selecting colors for a figure may seem like an inconsequential task and people often stick with the default color scale of whatever visualization tool they're using. This may not be the best choice, as misuse of color in data presentation may confuse the viewer and lead to misinterpretation of results. It is important to think through your color choices in order to best present your data and make the main points of a graph or chart easy for viewers to correctly figure out.
Color blindness (a.k.a. colour vision deficiency, or CVD) "affects approximately 1 in 12 men (8%) and 1 in 200 women in the world." So attention to color is an accessibility issue as well. Color blindness by Elijah Meeks & Susie Lu takes user-specified color input and displays what various colored graph types would look like to people with types of colorblindness.
Limit your colors to ~8. Research has shown that “humans can only distinguish ~8 colors,” so too many colors can overwhelm the user.
For another before-and-after example of this effect, look at the Limit Colors section of Tableau’s Visual Best Practices guide.
What to consider when choosing colors for data visualization
Lisa Charlotte Rost of Datawrapper
A quick overview of what to consider when choosing colors for different type of graphs. One important question is: what do you want the viewer to take away from the graph?
Learn UI Design
This interactive website helps users generate a usable color palette. Also check out the "single hue" and "divergent" choices. A similar website, Colorgorical, lets users specify hue, lightness, and other technical color characteristics when creating a color palette.
Lisa Charlotte Rost of Datawrapper
This post includes lots of links to additional tools for using color effectively in visualization.
Sources
"About Us." Colour Blind Awareness. http://www.colourblindawareness.org/about-us/. Accessed August 14, 2020.
Cotgreave, Andy. "Fewer people are dying on US roads, but seasonal trends persist." Tableau Public, https://public.tableau.com/en-us/gallery/seasonal-trends-us-car-accidents. Accessed August 14, 2020.
Knaflic, Cole Nussbaumer. "Exploratory vs Explanatory Analysis." Storytelling with Data,
http://www.storytellingwithdata.com/blog/2014/04/exploratory-vs-explanatory-analysis. Accessed August 14, 2020.
Silverstein, Larry. "From "huh?" to "a-ha": The science of data visualization." Tableau Conference, 2017, https://tc18.tableau.com/sites/default/files/session/assets/18BI-030_ScienceOfDataVisualization.pdf. Accessed August 14, 2020.