This article is about how rules like “avoid pie charts” can be useful for beginners, but also unhelpful in real-world scenarios with more nuance. Instea
you might also use pie or doughnut charts to highlight important categories to direct attention or tell a narrative. For instance, highlighting that global online sales are higher than regional sales for each other country, combined.
This use of color emphasizes how online sales make up more sales than any individual store, combined. We could show the same thing in a bar chart, but it takes up more space and might be less effective.
The previous example also highlights why “data storytelling” does not easily apply to most Power BI reports. You can craft a data story with your visuals, but ultimately, the user is the storyteller.
It is easy to show a screenshot that looks nice, but it is another thing entirely for the report to remain useful after interaction. We will leave that discussion for another, future article.
Despite the bad reputation of pie charts among data or design professionals, many people often still opt for pie charts in their reports for various reasons, like the following:
The biggest problems with pie and doughnut charts are accuracy and precision. While it is easy to quickly identify the biggest and smallest slices, it becomes more difficult with more than three categories or when slices are similar in size.
Typically, when you have more categories, you focus on the top few and group the rest into “others” in a single slice. In Power BI you can do static grouping and binning, but unfortunately, the dynamic “TopN and others” pattern is typically too complex for the average user to implement or maintain.
If you need a deep guide specifically about color in data visualization, then we recommend this article on Datawrapper by Lisa Muth (which is tool-agnostic).
If the purpose of the chart is to evaluate magnitude and compare categories, then a pie chart is not a good option. You generally opt for pie or doughnut charts and other part-of-whole visuals when your focus is on answering data questions about proportionality.
Note that if you are truly interested in the evolution of proportionality over time, you might also consider a stacked area chart or 100% stacked area chart, too (although stacked bars work with discrete data like yearly trends).
