Did you know that less than 1% of the world’s data is being analyzed as of today? While we all know data is power, power is only useful when harnessed correctly. Large quantities of useful data are getting lost in this chaotic data ocean, creating a ‘Big Data Gap’. Extracting this huge unstructured/untagged data into a structured form has been one major reason for low data analysis rate. Automating, summarizing & efficiently depicting this data has been another major reason. This is particularly true in the Cloud & Saas Industry where data analysis is extremely important. We, here at Divami, have been working on a lot of SaaS, Cloud and Enterprise platforms and would like to share my experiences with data visualizations.
Some of the insights I have gained through my experiences in Saas, Cloud and Enterprise platforms that helped me make the user experience for each better:
- When well designed, data visualizations can summarize and represent the data conclusions as well as predictions in a simple, yet profound way.
- Visualizations can act as our analytical tools – they can help us interpret, understand and draw conclusions quickly. Also, drill-downs can give us deeper insights.
- They can be eye candy to the users, and can definitely replace plain boring numbers and tables.
Having said that, some of the common design mistakes that I have made during my career that could have been avoided are:
- Overdressing the graphs and complicating them with rich UI approaches.
- Conveying inaccurate conclusions of the data.
- Picking the wrong charts, without understanding the entirety of the data.
- Choosing wrong, complex or multiple scales in charts that led people to misinterpret the data.
- Using complex graphs when the simple ones were enough to tell the story.
In my initial years of designing, I used to select a few complex visualizations to represent the data, not knowing that it was defeating the purpose. Later on, in usability testing, I realized that we as humans take time to process these complex visualizations. So it was very important for me to reduce the chartjunk wherever possible.
The following are the three data visualizations that I have used very frequently over the years. People say these visualizations are overused, some say they have been beaten to death. But I say that they are the most compelling visualizations that help us depict the most complex data in the simplest way.
Over the years, I have realized that the Bar Graph works perfectly when the data sets have to be compared to provide invaluable insights. They are generally very easy to analyze, as the relative height differences of the bars tell the story of comparison. This graph is extremely useful when representing a broader range of data types and their comparative analytics. Any other visualization may make it complex, misleading users to wrong interpretations, which defeats the purpose of a visualization. For users to gain meaningful insights from the graph, these are some points I have thought are essential:
- Ideally, the vertical axis should start at zero. If not, the initial value should be called out.
- The bars should begin at the baseline.
- Ideally, we should avoid sideways labels to reduce the clutter and increase the readability.
- Remove any visual clutter (Increase data-ink ratio, Tufte’s principle)
- Avoid stacked charts, difficult for comparing data
Line graphs are perfect to depict overall trends over a period of time. It helps you understand the changes in the data, especially when the data are collected periodically in a consistent manner. It can also be used to compare the continuous changes across multiple groups of data for the same time period. However, the ratio of the graph can dramatically change the perception of the data. Increasing the height can create an exaggerated hike, whereas stretching the width can underplay it. There isn’t a fixed rule by which we can calculate the aspect ratio of the graph and this is probably an impractical measure, but judging by eye tends to do the trick. Some quick tips that helped me:
- Avoid having any shading or borders
- Highlight each data set with its own unique colour
- Remove all grid lines
- Use direct labeling wherever possible, avoiding indirect look-up
- Do use proper aspect ratio to minimize dramatic slopes effects
Pie Charts are perfect when we are comparing parts of a whole, especially when the parts are limited in their number. I have always thought Pie charts are very simple. We have seen them all our lives, both in our professional and educational settings. In reality, sometimes, they can be very misleading and can have a better alternative. Usability tests indicate that it’s difficult for our brains to make accurate estimates or comparisons of angles. When the slices are similar in size, it’s almost impossible to tell which is bigger. Even when the sizes are different, you can only say that one is bigger than the other, not really by how much. Showcasing precise numbers can be done by using direct slice labels, but this option makes your eyes jump back and forth between the pie and the legend. Some logical rules I follow:
- Don’t have more than 5 – 8 parts in a pie
- Works better in 2D, rather than 3D pies
- Start the biggest section from the “12” point at the top
- Sort your data for easier comparisons
- Be very careful about how you treat “no-data / missing data”
Now that you have learnt from my experiences and will soon enough try them out for yourselves. Just keep in mind that before you start charting, take a step back and ask yourself what are the main questions you want to answer. Choose the right chart type that is best for finding specific patterns and gain possible new insights into your data. Most importantly, remember that the charts don’t need to be complicated and that in fact the simplest charts are often our best solutions.