Eugene Lebedev is a consultant specializing in Power BI. Prior to establishing Versich in 2021, he developed Power BI reports for Autodesk. His reports were utilized by the VPs of Finance and C-suite at Autodesk. Through Versich, Eugene has designed dashboards for well-known clients like Google, Teleperformance, Delta Airlines, and over 200 clients globally.
Data visualization is an essential skill for any data analyst. Clear data visualization ensures that analytical findings are effectively communicated. When the communication style using data is on point, the message resonates and can drive meaningful change within an organization. In this article, I will explore how to choose visualizations that make your message crystal clear. I'll cover principles of effective data visualization, various chart types, and the thought process behind them.
For this discussion, I will utilize Power BI to visualize the data, as it is my preferred tool. However, the principles I discuss are relevant across various platforms: Tableau, Excel, Looker Studio, etc. In this article, I will reference the US bus breakdown dataset from Kaggle. You're welcome to use it for your own data visualization projects.
1. Principles of Effective Data Visualization
Avoiding Clutter
Reducing clutter is my first principle. Clutter diverts the dashboard viewer's attention from the primary message. When clutter is present, the overall communication becomes murky and the various elements blend together.
By clutter, I mean items like:
An excessive variety of colors: Observe how the chart below resembles a rainbow with too many colors. Each time you engage with it, you must first interpret what each color signifies, then remember it, and finally locate that color on the pie chart.
Insufficient white space: The chart below is filled with distracting text: names of the X-axis, Y-axis, and the visual's title. These could be condensed into a single title. The numbers displayed can be found both above the bars and along the Y-axis, providing no additional clarity by being repeated in two places. Additionally, gridlines and sliders add visual noise that does not enhance communication.
Later, we will delve into what elements contribute to clutter in each chart type and how to circumvent it. Keep reading to learn more.
Using Color Strategically
The second principle consists of clear dos and don’ts.
Don’ts
Avoid bright and jarring colors: Consider the chart below. Just a quick glance makes one want to look away. Bright colors fatigue dashboard users much quicker than pastel or neutral hues.
Do not use too many colors: The earlier pie charts exemplify this issue. Many alternative visualizations do not require numerous colors, such as horizontal bar charts.
Dos
Opt for colors that have meaning: For example, green often denotes positive outcomes, while red is associated with negative results.
Maintain consistency with colors across your visualizations: If you represented breakdowns in blue in one visual, continue using blue for breakdowns in others.
2. Chart Types - Selecting the Right Visual
Once you grasp the key principles outlined above, you can begin picking the most suitable data visualization to convey your insights. Depending on the visual you choose, your audience may find it easier or harder to comprehend your message.
It's worthwhile to discuss each data visual type individually to ensure proper selection. Misuse of a visual can result in your audience misinterpreting or disregarding the insight!
Vertical Bar Chart
The general guideline for this chart is that categories on the X-axis must follow a logical order. I typically use vertical bar charts in two situations: to illustrate a distribution or a trend.
For instance, the chart below showcases the distribution of breakdowns by delay. There's a clear order of categories on the X-axis: 15, 30, 45, etc. When this order is clear, vertical bar charts are ideal for visualizing distribution. Alternative scenarios include visualizing age distribution or income group distribution.
> Note: This chart type is not appropriate if there's no clear order among the categories on the X-axis.
For visualizing trends, consider the example below where the categories on the X-axis again follow a clear order, making the vertical bar chart effective.
Horizontal Bar Chart
When there isn't a logical order among categories on your X-axis, a horizontal bar chart is preferable.
Another advantage of horizontal bar charts is that they typically enhance readability of category labels. For instance, compare the two charts below. The vertical bar chart leads to diagonal text on the X-axis when the text is lengthy. Diagonal text may strain your audience’s necks, whereas horizontal orientation makes the text much easier to read.
Pie/Donut Charts
Pie and donut charts are often challenging to interpret. Hence, the only suitable scenario for these is when breaking data into two categories.
When more than two categories are involved, pie charts can lead to confusion. In the example below, data labels were removed from the pie chart. Can you discern which segment is larger?
The answer is shown below. Users will find it significantly easier to interpret length rather than the size of a pie segment. Compare the two visualization methods.
Stacked Horizontal Bar Charts
These charts can serve as a beneficial alternative to standard bar charts for several reasons:
They occupy much less space on the canvas.
Comparing lengths is easier than assessing the size of segments.
The guideline here is to use a horizontal orientation when there is no logical order among categories on the Y-axis.
Stacked Vertical Bar Chart
This visual type is optimal for illustrating the trend of a distribution. How you format this chart should largely depend on the extent of change in data over time.
Consider the two examples below. Which one allows you to better visualize the trend? If the change is minimal, I prefer to reduce the space between the bars.
Line Charts
Since the points on a line chart are always connected, they naturally convey trends. Therefore, line charts should exclusively be used to visualize trends.
Using line charts for anything other than trends can confuse users. One variant worth considering is slope graphs, which show changes between two time points.
Maps
There are two main types of maps: bubble maps and filled maps. Bubble maps feature larger bubbles for bigger segments, while filled maps use darker shades for larger segments.
Personally, I find it challenging to compare sizes of bubbles. I prefer to compare color gradients. In my experience, I almost always replace bubble maps with filled maps.
Pivot Tables
While one of the most traditional data visualization methods, pivot tables remain very effective. Sometimes, when line or bar charts lead to clutter, a simple pivot table is a better choice.
In the example below, I replaced a line chart with a pivot table. See how much cleaner this presentation is?
Adding color to pivot tables is crucial as it enhances readability. In the example below, I designed it to darken colors if the value for the next year rises, replicating the upward and downward movement of lines in a line chart.
I recommend substituting visuals with pivot tables whenever you encounter clutter from multiple lines or colors.
Tables
Often, analysts steer clear of tables because they feel data needs to be "visually presented." However, a straightforward table can clarify data quite effectively.
Consider the screenshots below where I replaced a pie chart with a table. This change declutters visual data presentation by eliminating several colors. Using blue gradients also assists users in visually distinguishing which categories are larger.
I suggest swapping visuals for tables whenever clutter arises from various colors.
Conclusion
To visualize data effectively, it’s important to adhere to data visualization principles. The primary guidelines are to minimize clutter and use color thoughtfully. The rest revolves around selecting the appropriate data visualization method.
Utilize this article to help you choose the right visualization and apply the principles described to streamline your existing reports!
Soon, I'll provide insights on structuring visuals on your custom dashboard. This will help ensure that you communicate insights in an organized manner and tell a coherent story with your data. A story is far easier to interpret than a series of disconnected statements.
