Through a series of client interviews, heuristic walkthroughs, and research, I came up with solutions that ranged from the generic to the unique.
Data Rules
A good visualization does two things: it reduces user bias and increases comprehension. Basically, visuals should allow users to see trends quickly and accurately. Based on these goals for our visuals combined with our understanding of human perception, I was able to hammer out rules for choosing visuals and creating reports.
Straight lines, Simple SHapes
Straight lines and simple visuals are typically better, allowing users to focus on and easily compare data points. After all, humans will always be better at estimating the difference in length between two straight lines than guessing at the difference in area between two pie slices. (Or at least, we will be until we install computers into our brains or evolve.) Readability for visualizations is key in deciding what stays on the page and what gets removed—for example, including data labels provides accurate information quickly, but including them when they start to overlap and bunch together slows down the user's ability to parse the data.
Choosing Accessible and Consistent Data Colors
Data colors also add a new layer of complexity to reports, especially when considering colorblind and low-vision users.
For data colors to be accessible, they need to follow the contrast ratio rule for large shapes, at least 3:1. However, since there may be many data colors interacting with each other on the page, it would be impossible to always get that 3:1 ratio with both the white background and the other data colors. In cases where there are multiple colors in a visualization, the 3:1 rule between adjacent colors should be prioritized, as opposed to contrast with the background, and should be checked with colorblind filters. For these reasons, many data color palettes, including the default palettes for Power BI, tend to follow a light-dark-light-dark pattern.
Furthermore, red and green are popular indicator colors for data but may not accessible for colorblind users. Red-green colorblindness (protanopia) specifically is the most common form of colorblindness, meaning that users with this type of colorblindness wouldn't be able to distinguish between positive and negative indicators. The solution to this has two parts. First, I recommended having the indicator displayed in two ways rather than just via color where possible. This can be done by pairing the color with a shape or text. Second, I adjusted the hues of the red and green so that when viewed under a filter, they were distinguishable.
As one might guess, these rules limit how colors can be added to the palette. So what should I do?
In this case, the best tactic to get an ideal palette was to provide several color options for the client to choose from but also to educate about how they needed to select from a limited number of colors, which would ensure these reports were accessible to users with visual impairments.
In the end, I created a color palette for the data that includes 8 core colors (as allowed by Power BI) plus 22 extended colors for larger datasets. On top of this, after interviewing and researching the team's report usage patterns, I assigned special colors to select metrics that appeared to be the most important and most frequently used.
Accessibility vs. Data Privacy
As one might have guessed from the previous section, there was a large focus on accessibility for these reports (hooray!) especially since The client company has stricter accessibility policies than most and has team members with visual impairments that we were able to speak to. Of particular concern was keyboard accessibility for Power BI, which is a great tool for creating and displaying reports but is not quite as keyboard-friendly as one would like. To the best of my ability I gathered information and resources and provided instruction on how to make Power BI reports as accessible as possible. Certain oddities keep Power BI from being fully keyboard accessible, such as a lack of hierarchy (imagine a web page with no H tags) and tables that can't be traversed the way HTML tables are.
In fact, the non-HTML tables caused a pretty big headache for a user who was blind and couldn't traverse their data properly. As a pretty good workaround, blind users will usually download tables from Power BI as Excel sheets, which are far more accessible. However, much of the client's reports contained Personally Identifiable Information (PII).
Why is this a problem? Well, not only was the client conscious of report accessibility, they also cared about data privacy and international compliance to data privacy laws. This means that under laws like the General Data Protection Regulation (GDPR), users cannot download and keep data that contains PII.
To the best of my ability, I created the design system while trying to balance opposing forces like these. But one of our greatest tools was also our greatest constraint. Power BI, as a tool, provides an easy way to create reports and visualize data, but like many web applications and websites, could still use some improvement when it comes to addressing accessibility concerns. While we were able to provide that user with a workaround, it would have been better if they were able to complete their job as smoothly as their other coworkers out of the gate. As far as I know, Power BI continues to grow and is actively trying to resolve some of these issues, but we're still a ways away.
