Every day, the world generates 210 billion emails, 4 billion text messages and 90 million tweets. According to MIT researchers, the amount of information that crosses the Internet every second in today’s world is equal to the entire contents of the Internet just 20 years ago. To interpret the data and to make appropriate judgments, leaders need to see the data in context. Visual analytics tools transform large datasets into interactive visualizations that allow leaders to make crucial decisions.
Visual analytics isn’t new, but it’s becoming more critical in the era of big data. Business intelligence, much of it collected by data mining, has become an important field of study at the university level. Students can even earn graduate degrees in business intelligence; if you’re curious, see this website to learn more about earning an online business intelligence masters degree. Visual analytics representations have great value when data is well-organized and when visuals are presented in a logical, uncluttered way.
The Benefits of Thinking in Pictures
IDG Research asked survey respondents to rank the benefits that they receive from data visualization packages. These benefits included:
- Better decision making
- Improved data analysis on the fly
- Better information sharing and collaboration
- Self-service capabilities for end users
- Increased ROI
- Time savings
- Reduced IT burden
For large organizations, spreadsheets and charts of data simply won’t suffice. Imagine a large retailer looking at data from over one billion customer transactions using spreadsheets. Spreadsheet software can construct basic visuals, like simple regressions, but it doesn’t have the power to interpret today’s enormous datasets. Visual analytics software can provide multiple layers of data interpretation and data quality information.
How Visual Analysis Works
Visual analysis begins with a task for which decision makers want to analyze data. After getting the data, a visual analytics program analyzes the data’s structure. After generating a visual representation, decision makers can develop insights. These insights enable them to make better decisions than they would if they were looking at charts of numbers.
Visual data representations depend on the quality of the data inputs. Make sure to take these steps before creating visuals:
- Scrub data. Data cleansing or scrubbing means detecting and either removing or correcting corrupt records. Errors may include incorrect categorization, inconsistent use of measurement units, spelling errors, postal address errors or errors with commonly recognized identifiers (e.g., Social Security numbers).
- Normalize data. Normalizing data enables you to minimize redundancy within and dependency between datasets. When data is properly normalized, you can make a modification to one field in a table, and the modification propagates through the entire database.
- Group data. Raw data should be classified into classes for easier visual analysis. For example, POS transactions can be classified according to the purchase amount. These amounts can then be grouped into classes such as $0 to $100, $100 to $200, etc.
- Integrate data. Most companies have data stored in multiple databases or locations. You may need to bring together data from disparate sources or migrate data to a central location.
- Govern data. For all future data collection and management, develop a single, consistent set of processes and policies for managing corporate data.
How Humans Like to See Data
If you’re responsible for converting data into visuals for decision makers, then you’ll need to choose visual structures that make sense to the viewer. For example:
- Time. To show change over time, present data along an x-axis.
- Location. Put your results on a map. The map doesn’t have to be geographic. Just be sure that you avoid using filled maps for datasets involving more than one measure.
- Different values. Compare values by using a bar chart.
- Relationships. Scatter plots work well for relationship analysis.
- Relative proportions. Go with a treemap (nested rectangles).
As you consider color, remember that humans can only perceive eight colors. Keep your color sets simple and consistent across multiple visuals, and present colors against a consistent background. Also, if you decide to present multiple visuals on a dashboard, then make sure that your viewer can orient to the dashboard within five seconds of looking at it. Use no more than five different visuals on a dashboard, and place the most important one on top or in the top left corner. Finally, make the data interactive, if possible. Tablet computers provide great tactile tools for interacting with visual data representations.
About the Author: Tess Harmon is a statistician providing consulting services to data analytics software developers. She has a particular interest in visual analytics for mobile devices.