Data Visualizations: 11 Ways To Bring Analytics To Life
Data visualizations, used well, can help people make sense of large, complex data. Learn how data visualizations are changing and best practices for making the most of them.
![](https://eu-images.contentstack.com/v3/assets/blt69509c9116440be8/blt61c45e1f56119977/64cb509235bff28d2d3e7c96/1-grid-725260_1280.jpg?width=700&auto=webp&quality=80&disable=upscale)
User demand for interactive data visualization is driving an evolution in the way that complex information is presented.
Rather than accepting static pie charts and bar charts at face value, professionals and consumers alike want interactive tools that allow them to visualize, understand, and analyze data in an iterative fashion, whether at their desks, on the web, or on their mobile devices.
To facilitate the democratization of data and accommodate the increasing volume, velocity, and variety of data, visualizations have improved aesthetically and functionally on a number of levels. Simple 2D charts and graphs have been supplemented or replaced by sophisticated visualizations that accommodate large numbers of data points using additional dimensions, colors, animations, and the digital equivalents of familiar items such as dials and gauges. Meanwhile, software designs and GPU acceleration have made animations, grids, and UI responsiveness faster and more fluid than ever.
"People want graphics that can do cognitive heavy lifting for them and help them explore their ideas. They want to dig into the data in a way that lets them test their own hypotheses and think about their own data biases when they encounter a dataset," said Daniel Faltesek, an assistant professor of social media at Oregon State University, in an interview. "They don't want people to tell them what to think about the dataset. They want to play with it and understand it themselves."
[Can analytics fight fires? InformationWeek finds out. ]
Tools that provide collaborative capabilities enable people within an organization to share, discuss, and debate what the data is telling them. To accommodate deep understanding of the data and individual differences and preferences, these tools typically offer multiple views of the data available so users can easily move between them.
"The ability to slice and dice has been around for a while. It's more exploratory now. You have a lot of data sources, so finding a needle in a haystack boils down to being interactive," said George Ramonov, founder and CTO at meeting planner provider Qurious.io, in an interview. "Now that we have cross-functional teams, it's important to be able to share visualizations embedded in a website or app to allow sharing without all the extra time of putting together an email."
Regardless of how large or small an audience is, good data visualizations speed understanding. Bad visualizations cloud the issues. Here are six ways to best leverage the graphical presentation of data.
There are more data visualization choices than ever before -- more chart types, more gauge types, more grids, more maps. With them, users can get different views of the same data, or use different visualizations to analyze data at different levels of aggregation or granularity.
For better or worse, color choices are virtually unlimited, and there are special effects that help clarify shapes or draw a viewer's attention to a particular element. Because the possibilities are so vast, one can get distracted by what's possible, rather than focusing on what's practical. Software defaults help narrow the possibilities, but they're not a complete substitute for thinking.
"People place too much faith in the defaults in software. A lot more vendors are trying to do smart visualizations based on the data. They're getting better, but you need to engage in the construction of a good visualization," said Andy Flint, senior data scientist at analytics software company FICO, in an interview. "Ask yourself what question you want to answer, and whether the visualization is supporting an answer to that question. If it isn't, it's not really worthwhile."
One problem is that visualizations tend to be very subjective. What's clear to one person may confuse another. However, the availability of so many options means individuals can clutter their visualizations with unnecessary noise.
Powerful computing devices and smart software designs are enabling sophisticated graphics, fast rendering, and immersive user experiences, while behind the scenes the data continues to get more complex.
"In-memory MapReduce, Spark, and GPUs are proven ways to create massive calculation power behind visualizations. You can get a great deal more information in front of someone, and you can interact with it in ways that would have been impossible before," said Andy Flint, senior data scientist at FICO. "It used to be you'd let the computer run for a while and you'd get static output, but if it begged other questions, it would take time to get answers. Now you can get those answers fast [and benefit from] deeper analysis."
It's easy to be seduced by some of the more sophisticated capabilities simply because they're available, although some have learned the hard way that "more" is not necessarily better.
"There was a time when people said they wanted 3D and animation, but some of them are dialing back and thinking about the aesthetics -- whether it's really communicating the information the user wants," said Daniel Jebaraj, vice president at enterprise software component provider Syncfusion, in an interview.
In today's mobile world, users expect to have fast access to information, although organizations sometimes struggle to provide effective data visualizations in a mobile context. Some of them are adapting desktop experiences to mobile formats, while others are taking a mobile-first approach. In either case, the goal is to expedite understanding of the underlying data using intuitive, touch-based interaction.
"The mobile form factor is distinctly different because you have less real estate. The other aspect we see is that [users] don't want to see a picture, they want to play around with it," said John Whittaker, executive director of marketing for Dell Software's information management group.
Because smartphone display space is so limited, designers are taking advantage of gesture-based input to help users easily navigate different views and interact with the data. The constrained space requires information to be presented more simply than on a desktop, so, rather than simply translating a complex desktop view into a simpler mobile one, it is important to consider context.
"Context can take on different meanings depending on your domain. If you have a mobile workforce and you've issued devices to them, you can know where they are and can offer information about customers that are close to them," said Daniel Jebaraj, VP of Syncfusion. "For mobile, you need to make the most common tasks easily accessible."
Data visualizations have evolved to accommodate increasingly complex datasets. As data continues to grow in volume, velocity, and variety, the trick is to visualize the data in a manner that most effectively communicates the main point.
Organizations are visualizing many types of data today and they're going to be visualizing even more types in the future as the "Internet of Yet More Things" emerges. It is unclear what the effect will be on data visualizations, even among vendors who are already selling applications to address the growing demands.
"The possibilities are pretty phenomenal and that complex web of data is begging for smart visualizations, not just numbers in a table somewhere," said FICO senior data analyst Andy Flint. "There's a need to correlate data from these various sources -- routers, fitness devices, the telemetric devices auto insurance companies are promoting."
One thing the Internet of Things will do is provide real-time or near-real-time pictures of what's happening in different places at different times.
"If you manage a large traffic network or oil refinery, where you have a million crossing streets or pipes, and you increase that by one or two orders of magnitude, you've got a lot more real-time data to project in a similar space. It's bound to change the visualization," said Nicholas Marko, chief data officer at Geisinger Health System
Data analytics application providers include plenty of visualization options in their packages, as do vendors selling data visualization components to software organizations. Many types of data visualizations have been developed to accommodate different types of data and datasets, and it is common to allow toggling between them so users can explore the data in different ways.
Different data visualizations also facilitate understanding between people who perceive the same information differently. When information is shared via a collaborative mechanism or via a dashboard or presentation, someone has to decide which data visualization or combination of data visualizations to use. Although a lot has been written on the subject, not everyone understands which visualizations are best in which circumstances, and which can lead to less-than-optimal outcomes.
"If it doesn't have any signal or story, some might say 'it looks terrific,' but they don't get it. If you want the viewer to get your story, you have to keep the question in mind," said FICO senior data scientist Andy Flint. "In the early days, everything plotted had a very heavy set of axes and lots of tick marks everywhere. Aesthetically it was poor and it was also impeding the message."
Statistics can tell any story, and so can the data visualizations that are supposed to reflect those statistics. Even if the presentation is ideally suited to the dataset, it can nevertheless tell very different stories depending on how it's presented.
"If you want to make a point you can do it. You can change the scale or make what you want pop up. That's the reality of statistics and numbers," said Syncfusion VP Daniel Jebaraj. "Users need to get more educated about what's being presented so they can form their own conclusions. Visualizations summarize things, but the summary comes with its own viewpoint."
Sometimes, bias also causes people to restrict the datasets from which they're willing to pull.
"People get myopic with the data they use," said Shay Har-Noy, senior director of geospatial big data at Digital Globe, in an interview. DigitalGlobe is a provider of satellite images, aerial photos, and geospatial content. "We saw this with satellite -- they're only using satellite images. Why? Or tweets. Really? You're just going to get this unofficial, potentially statistically significant sample from this one data source? It's a mistake to get myopic about the data and the methods you use."
Data visualizations are expected to become more prescriptive over time so that users can have greater confidence that the answers to their questions are accurately represented from a visual standpoint. Built-in defaults are common, but even more intelligence is expected in the future.
"One of the things we're going to see more of is advanced analytics getting expressed through visuals without the user asking for it -- putting intelligence into the system to best display the data," said FICO senior data scientist Andy Flint. "More and more, we're going to see those visualizations enhanced by analytic regressions or deep learning neural networks and so forth without the consumer or creator of the visualization having to take an incredibly active role in it."
Qurious.io founder and CTO George Ramonov said he sees an interactive form of guidance.
"You'll be able to construct visualizations, but the systems will guide you to use best practices," he said. "A lot of people are not following best practices, so they don't know how to construct visualizations. But going forward it's going to be more relevant to predict the best visualization. Tools could ask you what you want to achieve [such as] you trying to show relationships or distributions or the composition of things, so there will be more interactivity in terms of data visualization and user experience."
Predictive analytics is enabling organizations to make more accurate business decisions sooner. By using those capabilities, companies are reducing risks, exploiting opportunities faster, and enabling new kinds of innovation. The same trend is expected to translate more broadly to data visualizations so more professionals and consumers have a future view of possibilities in addition to glimpses of the present and past.
"It's becoming easier to model with data so you can learn from it and predict what a person will do or what a value will be. That can be plugged into visual indicators so you have something that summarized past behavior and predicts the future of the environment, said Daniel Jebaraj, VP at Syncfusion. "That will add a lot of practical functionality."
Pretty pictures can sometimes tell a misleading story. When a visualization aligns with a hypothesis, it's easy to conclude that the output confirms the hypothesis, when it's really only demonstrating confirmation bias.
"People will assemble the data that will confirm their hypothesis, but you need to look for evidence that might not confirm your visualization," said Jock Mackinlay, VP of R&D at Tableau. "You can look at something and think it's significant because it looks like it lines up, but you should probably run the statistics to make sure what you saw is right."
Pretty pictures can sometimes tell a misleading story. When a visualization aligns with a hypothesis, it's easy to conclude that the output confirms the hypothesis, when it's really only demonstrating confirmation bias.
"People will assemble the data that will confirm their hypothesis, but you need to look for evidence that might not confirm your visualization," said Jock Mackinlay, VP of R&D at Tableau. "You can look at something and think it's significant because it looks like it lines up, but you should probably run the statistics to make sure what you saw is right."
-
About the Author(s)
You May Also Like