MBA Students: 3 Principles of Data Visualization

Data visualization is a process in data analytics chain where visual representation is used to communicate a story using insights. What is the purpose? To make data driven decisions. A poorly generated data visual can cause many problems including unwanted and incorrect interpretation, incorrect decisions and, ultimately, poor business results.

Top 3 best practices are:

1. Purpose & Data type-driven and not simply beautification

What is the purpose of the visual? What is it based on? What do you want to communicate? What kind of business decisions are at stake? Who are the KDM (Key Decision Makers)? To what extent are the KDM literate? Every visual has one of the 5 representation mechanisms: distribution, composition, relationship, trend and comparison.

• Distribution: to show how the data items are distributed across different parts. E.g. Histograms, line charts, scatter charts

• Composition:?to show what constitutes the data i.e. what are the major parts? E.g. Pie charts and stacked charts

• Relationship:?to show how one variable relates to other or a group. E.g. bubble charts, scatter charts

• Trend:?to show the trend over a time period. It is part of time-series modelling typically. E.g. Line charts, column charts
• Comparison:?to show how two sets of data compare against each other. E.g. spider charts, bar charts, radar charts

There are 4 types of data:?nominal, ordinal and numeric (including ratio). Ordinal helps to rank, nominal deals with categorical data. We can use Histograms for categorical; bar chart for ordinal and charts like line, scatter, line, box and whisker etc. for numerical data.

2. Consistency of Scales

Choosing the right scales to convey the degree of difference consistently and correctly is important. Many a time you come across charts that accentuate the difference whilst the actual one is not so. Kaggle explains?the scale must be consistent. There is a metric called lie factor. It denotes the relationship between the size of the effect, as seen in the graph to that of the underlying data. The recommended value is around 1.

3. Optimized Real Estate

Just because a slide offers so much space, do not pack it with irrelevant charts. Less is more. Clarity trumps everything. Two useful metrics that can help are:

Data Density measures the quantum of info in a given display space. The mantra is not to depict a large chart with small amount of data.

Data-Ink ratio: measures the proportion of ink used to present actual data to the total used in the entire display. To get a higher ratio, one can think about removing graphics, 3-D effects, background images, grid lines etc. They only clutter the chart and defocus the target users.

Conclusion

Once these 3 principles are used to identify, create and build a data-compelling story, some aesthetics will help.

• Recommended font is sans serif for headings and serif for details.
• People read from top, move across and then down. Ensure the chart/visual can be studied in the same manner.
• Use golden ratio (1:1.6) where possible.
• Stick to not more than 3 colours in a chart. Use one for the base, one for the data and one for any trend or other factor.

If you are interested further, read Edward Tufte, the?pioneer of data visualization, who says "The essential test of design is how well it assists the understanding of the content, not how stylish it is.”