Planning: Important Considerations
There are a few important questions to consider when planning an effective visualization:
Who is your target audience?
In general, an effective visualization should be quickly and easily readable by everyone. However, considerations of audience expertise, requirements, preferences, and expectations are still important in determining your graphic type and presentation stylization.
What message are you trying to communicate?
Designing your visualization with a clear purpose in mind will help you make decisions that leave your audience with a clear takeaway about your data, making your visualization an effective one.
What context is important to understanding the data and your message?
Failing to properly address the context can hinder readability and even result in misleading visualizations.
How do you intend to communicate your message and any necessary context?
With the previous questions in mind, select a visualization type suitable for your data and message. Are there any things you can do to emphasize your message and increase readability?
Without context, statistics and visualizations can be incredibly misleading. A lack of context may be especially hard to notice in visualizations as they are typically designed for quick, intuitive communication. The importance of context can be seen with visualizations regarding COVID cases.
Comparing cumulative COVID cases monthly in different US states seems like a perfectly fine and even illuminating graph. It may lead us to certain conclusions about which states have the fastest spread of COVID or perhaps the most ineffective COVID policy or enforcement.
COVID 19 data from New York Times and average population estimate for 2020-21 calculated using Government Census Data
However, the statistic of cumulative COVID cases in each state does not tell the full story, as it ignores the state population differences which are strongly skewing this statistic. The number of COVID cases per 100,000 people would give us a more accurate idea of the situation in each state because it allows us to look at the number of cases with proportion to the state population. Depending on what we are trying to find and convey, we may also want to consider population density, number of average daily exposures to COVID, and number of COVID deaths, among other contexts.
As visualization rely on perception, you should take advantage of intuition of visual processing through preattentive attributes.
Preattentive attributes are visual characteristics that processed by us incredibly quickly, allowing us to notice things even before consciously focusing our attention on it. They often are interpreted naturally, but may be processed at different levels of precision. For example, we can more easily perceive differences in length and position compared to differences in width and size.
Aside from representing quantitative differences, these can also be used to emphasize information. For example, enclosing a section in a box or highlighting a datapoint in red when the other points are grey will immediately draw the viewer's attention to that data.
Source: Tableau
Best Practices
The following are some guidelines to create effective visualizations:
aria-multiselectable="true" class="panel-group" id="accordion" role="tablist"> class="panel panel-info"> class="panel-heading" id="step1" role="tab">1. Keep it simple
Visualizations should be quickly and easily understood. People have limited attention and memory, so avoiding distracting elements like prominent gridlines, ornamental shading or gradients, and unnecessary 3D can reduce design clutter and help the focus stay on what you want to convey about your data.
- Use familiar chart types along with intuitive colors, shapes, and layouts
- If your visualization is complicated because of the amount of data, try breaking it down into smaller parts, either highlighting only a small subsection or turning the one visualization into multiple
Source: Datawrapper Blog - 10 Ways to Use Fewer Colors in Your Data Visualizations
- While beauty is important in creating a compelling visualization, it should never take precedence over readability
2. Don't forget your labels
- Always include a proper title and any necessary variable or axes labels, including units if applicable
- Consider integrating text into the visualization, such as labeling next to trend lines instead of in a separate legend to the side
- Adding explanatory text for context directly on the graph can be helpful, such as annotating when an event occurred to provide context for a change in a trend over time
Source: Yellowfin Blog - How to Enhance Your Data Visualizations with Context
3. Style considerations
Source: Tableau - Visual Best Practices (Courtesy of The Big Book of Dashboards)
- Don't include too many colors
Source: Datawrapper Blog - - 10 Ways to Use Fewer Colors in Your Data Visualizations
- Colors in your color palette should be easily distinguishable from one another by both people who are not colorblind and people who have any kind of colorblindness
- Don't change style across visualizations that will be compared with each other
- Use large enough font for readability, and use larger font for more important text elements like the title
- Customize. Avoid default styling such as for font and color because they don't stand out and are therefore not compelling
- At the same time remember that fonts affect readability, and that takes priority over how unique it is
4. Avoid misleading
Source: Ten Simple Rules for Better Figures (Rougier et al.)
- Axes should start at 0 and continue with no breaks
- If percentages, axes should continue to 100%
- Data presented on the same visualization (ex: multiline graphs) should be presented on the same scale
- If there are dates or times, they should be chronologically arranged
- Avoid 3D effects or anything else that might distort perceptions
In this 3D pie chart, the slices for items A and C appear to approximately the same size.
However, in the undistorted 2D version, you can see that the percentage for item C is less than half of that for item A.
Source: Wikipedia - Misleading Graph
Here are some resources for more detailed tips: