1. Best practices in visualization
Welcome back! I'm Maarten, the second instructor for this course. Now that you are familiar with several basic and more advanced charts, let’s have a closer look at best practices in data visualization.
2. Consequences of bad data visualization
No data analyst starts a day intending to mislead with data. While calculation errors are typically the biggest culprit, mistakes also happen when visualizing the data.
Consequences range from rather trivial such as the creation of ugly charts causing loss of the interest or trust of our stakeholders, through rendering perhaps useful but unclear visualizations which raise more questions than answers, to potentially misleading with data, leading to bad business decisions.
3. Most common mistakes in data visualization
There are many reasons why things may go sideways; let’s zoom in on the most common causes. These could be poor chart choices, misleading chart design or incorrect use of color, shape or size. On the other hand we have neglectful formatting, and finally, wrong handling of missing data. Let’s tackle them one by one.
4. Choosing the right chart for the use case
Before we start any visualization work, we need to consider several things. Firstly, what is it that we want to show? Do we compare and show the relationship, distribution, or composition of various categories?
Does our data consist of many variables and categories? Does it show the evolution over time? Should we show absolute or relative numbers?
We will continue discovering various charts together throughout this course, but make sure to click on the below link and check out DataCamp’s helpful Data Visualization Cheat Sheet!
Remember that the choice of the visualization will also differ depending on the goal; for example, an extensive dashboard might not be the best idea when trying to distill key numbers for senior management, so always consider who the audience is, how data literate they are and what impact you are trying to achieve.
5. Correct design of chart elements
Let’s learn about chart design in this example.
Firstly, we do not know what this visualization is about as both the chart's title and the axes' title are missing. We see an evolution, but we do not know if this concerns volume or money, nor do we see if this is a daily, weekly, or monthly view. Lastly, the trend seems to drop dangerously, but then we notice that the Y axis has an adjusted range. This chart can be improved! Always consider adding chart and axes titles, axis or data marks labels; in most cases, you’re better off keeping the range of axes starting at 0.
6. Conscious use of color
Also, color must be used with care. Use it to highlight key data points, applying clear, distinctive colors coherent within a brand or style. Don’t overuse red, amber, and green as they are reserved for indicating status such as good or bad, increase or decline, etc.
Finally, avoid using color altogether if it doesn’t bring any extra value.
7. Conscious use of shape and size
Shape and size can add extra variables but also add complexity. A good example of using it correctly is the graph presenting the life expectancy and births per woman for every country worldwide. The size adds value as it is used to represent the population and color groups the countries per continent.
Avoid shapes if they cause distraction or when precision is required. Can you tell the exact amount of revenue from circus art in China? I cannot.
8. (Un)told story behind the nulls
Aesthetics and design are important, but so is our responsibility to report the facts. One of the situations when we may accidentally tell only the “happy story” is when null values are purposefully excluded from the visualization. In the following sales report, we seem to generate a continuous revenue stream! However, when we also consider the null values, we notice that on some days, we did not generate any sales at all.
9. Let's practice!
It’s now the time again to put the theory into practice.