With Power BI from Microsoft, you can deploy meaningful visualizations for a variety of audiences. As I have grown to know the product, I have deployed it as a consultant as well as used it for fun. Apart from initial exploratory work, I have learned that greater success with Power BI has always been the result of a strategic, cohesive design regardless of whether I create a business report or something to share with friends. You can experiment, but do it in a way that may help enhance rather than detract from your message.
Here is a framework that I employ when designing visualizations:
- Step away from the software
- Consider the data (and data quality)
- Consider the audience
- Consider the message
We’ll consider each of these points using a sample “report” from Power BI. I have spent too much time looking at invoice, logistics, and sales data recently; so I will focus on something more appropriate for personal time: college football.
The final product shows conference stats for the current season prior to November 21:
2015 Big Ten football season as of November 20, 2015 — Visualized in Power BI
1. Step Away from the Software
Cohesive design does not begin with dragging charts around. Whether using Power BI or something else, a “spray and pray” effort where you throw a variety of charts, colors, and text together on-screen may eventually get you a successful visualization. More often than not, however, you get results that consist of poor quality, an unclear story, and lots of rework. A little planning goes a long way. Therefore, before using Power BI or any other visualization software to create something meaningful, consider a fruitful design process first. Know the capabilities of Power BI and keep its functionality in mind as you plan, but avoid the temptation to jump right in.
Good design starts here:
Poor design starts here:
Don’t use a “spray and pray” strategy for charts
2. Consider the Data (and Data Quality)
There are a lot of data and statistics tracked in sports. For my college football visualization, I selected the basics: wins, losses, and points scored by each team. I also narrowed it down to one conference: the Big Ten.
In terms of data, can I trust my source? Even with a reliable source, is the quality okay? I chose to obtain data from ESPN.com, and the likelihood that team records and scores could be wrong are low. Generally, however, it is not wise to take data cleanliness for granted. Poor data leads to a poor or misleading message for your audience. If needed, Power BI has built-in tools to help cleanse, transform, and model your data.
3. Consider the Audience
Design with your target audience in mind. Keep in mind that sometimes you are not a member of the target audience. For this project, who cares about college football data? College football fans? With data limited to the Big Ten Conference, perhaps only Big Ten fans? Also, be aware of assumptions. Should we assume that everyone knows what a win and loss is? If I use an image of a school logo, can I assume that people know the school, or should I also write the school name? Taking some time to consider who your audience is will help you create a more coherent, targeted visualization.
Can the audience identify Rutgers by logo alone, or is the text helpful?
4. Consider the Message
Over the past few years, there has been a growing focus on using data to tell a story. Whether you call it that or use other terminology, you are leading your audience through the data and trying to provide insights along the way. With my schedule data and audience of Big Ten fans, what am I trying to convey? Perhaps I want to focus on team records, but using the scoring data, can I also attempt to describe what might result in a win or loss? Do higheraverage scores correlate with wins? How can I guide my audience to see such a correlation (or lack thereof)?
Big Ten Records as of November 20, 2015
There are many different factors that result in wins and losses, and I only have data for the records themselves and points per game. In this case, limited datacould lead to incorrect understanding. What if a team’s offense scores a lot of points per game, but the defense gives up more points (such as Indiana)? If you presented a look at offensive stats alone, you might lead your audience to expect that such a team has a relatively high number of wins. Sometimes limited data is enough, but data that is too limited can be misleading.
Present only what is needed for the audience to gain insight and tell your story. If you have a simple message, do not err by presenting too much data just because it is available. In the real world, you need to sift throughnumerous data points and determine what might be relevant or not, thenuse only relevant data to tell the story. You engage your audience better when you present a concise, coherent message.
We are finally at a stage where Power BI is involved! At this point, you translate the fruit of your planning into a rough visual design. That design should consciously match your data and intent with the features and constraints of the software. For example, you can consider your medium — is it a report or dashboard? What chart or combination of charts work best for your audience? Can you find the balance between too bare and too cluttered? Do you have a visual theme, and how will color be used to enhance it? Etc.
In terms of chart types, date and team are good candidates for slicers. Using slicers, I can interact with my data and see results for individual or multiple dates and teams. For presenting both the win/loss data as well as average points per game, linear charts such as bar charts, a waterfall chart, or even a funnel are typically easiest to interpret at a glance. Finally, since football involves a network of teams, I could consider how to visualize each game played throughout the season.
Blurred vision above is not the result of Football Saturday
Once you have an initial design, it is time to execute. For my football visualization, I used the Chiclet Slicer for teams, which allows for custom images (team logos in this case). I also opted to use it for game dates both for consistency and because it is more versatile than the default slicer.
Chiclet Slicers: beautiful, versatile, and should probably be the default in Power BI…
For comparing points, I selected a Clustered Bar Chart to display each team’s average alongside their opponents’. A Waterfall Chart displays cumulative wins throughout the season, and it also functions as a helpful way to show a running total by date. Finally, the overall wins and losses for each team are presented using a Bar Chart.
Linear charts are typically more readable and easy to interpret
In order to visualize how all of the teams in the conference interact throughout the season, I used the new Force-directed graph. This presents each team as a node, with links for each game played, and points scored as the width of each link. Note how this visual could be helpful in an interactive environment, but it may have limitations when simply viewed.
Too cluttered: force-directed graph showing links between teams for the whole season
Now filtered for an individual game day; move clockwise from a node to gauge relative points
After the core charts are in place, final work involves consistent styling and meaningful use of sizing and color. At that point, the visualization is complete — or is it?
Some additional things to consider: Does the visualization convey my meaning as coherently as possible. Did I execute well upon my design considering my data, audience, and message? Did I use an appropriate combination of visuals in Power BI, or is there room for improvement?