SMB Toolkit: How to Choose the Right Data Visualization

Spreadsheets have been a mainstay in business for so long that some people have trouble letting them go. However, most businesses today have moved past the fleeting pull of nostalgia in favor of more powerful self-service business intelligence (BI) tools. There are many reasons these tools are superior to a spreadsheet, but the most obvious is the capability to easily produce effective and compelling data visualizations. That one factor heavily outweighs any argument in support of spreadsheets for one simple reason: Human brains absorb and process visual representations of information faster and better than they do numbers.

This human penchant for pictures over numbers is far from a recent discovery. History shows us the concept has been successfully tested over time, from early cave drawings to modern universal traffic signs recognized all over the world regardless of language. One glance at picture-based messages and the information is understood no matter your language or math skills. Modern data visualizations take that powerful communication medium to an entirely new level, infusing the rich charts and graphs generated with rich data.

Choosing the right data visualization, however, is not simply a matter of choosing a pie chart over a bar graph or scatter plot. It's not as simple as choosing traditional over avant garde visualizations, like 3D virtual reality (VR) depictions. There is much more than personal taste and preferences at play.

Each element of a visualization conveys a part of the overall message. From color choices and ink ratio to the data itself (and all the layers you can drill down into it), data visualizations are a single, deceivingly simple representation of complex, data-driven business insights.

So what visualization should you choose? Ultimately the choice is up to you to identify the visual representation that best suits the message you're trying to send with the data you're presenting. Here are the things you need to know to make an informed and strategic decision.

Types of Visualizations

I won't bore you with recapping all the tried-and-trues. If you have been in business for any appreciable amount of time or worked with data beyond simply entering it in software, then you're already familiar with the traditional visualization forms: pie charts, bar graphs, line charts, scatter plots, and maps with symbols. These, among other graphic representations, are generally organized by two or three features including time, number of units, and costs or sales.

There is nothing wrong with this group of visualizations. They have been in use for a long time for good reason: they work.

Indeed, their very familiarity means your audience doesn't have to stop and figure out the form before they can understand the message. That's exactly what you want to happen. So why wouldn't you stop here and pick one of these? Or, as is an option in many BI apps, let the software pick one for you?

Because not every analytic output can or should be expressed in such an over-simplified conveyance. A traffic light is reduced to three features because there are only three messages to convey: go, slow down, and stop. If you want to say something else, then you have to add another visual. Perhaps a sign that reads "No right turn on red" or perhaps you'll add another stoplight just for the turn lane. In dashboards and reports, all of those added visuals can pile up and, by the time your audience gets to the end of the heap, they've forgotten how that information relates to the first or other visuals in the pile. That's neither efficient nor instructive. Further, the use of too many visualizations can cause user fatigue. The message is lost on the mind that wanders.

In any case, you know these visualizations already. Let's move on and consider newer forms.

'Twixt and 'Tween

This class of visualizations depicts more features than the traditional group but the message tends to have a single thrust. For example, consider the "word cloud." This visualization measures many words in relation to one another so each word is depicted in size in proportion to its use compared to the other words. Colors can be used to depict subgroupings within the group, or other information, or simply to make it easier to see the individual word sizes in a glance.

When might a word cloud visualization be good to use? There are several use cases including customer/user mood on social media, escalating and/or de-escalating customer issues in call centers, customer inquiries on specific products, product sales, and other cases. Other examples of this type are frequently seen in infographics as they depict data based on a theme.

Vendor-Friendly But User-Averse

Then there are the visualizations that BI vendors are proud to offer but few buyers and users understand. But wait, you might say. If I can't look at a visualization and tell how it works, then how can my audience figure out what the data is telling them?

It's true that sometimes the visualization contains the information but fails to deliver it. Take, for example, Vincent van Gogh's "The Starry Night" that he painted in 1889. The iconic work accurately depicts wind turbulence but no mathematician or scientist recognized that understanding until centuries later. Talk about failing to deliver the information.

"Scientists have struggled for centuries to describe turbulent flow—some are said to have considered the problem harder than quantum mechanics," according to a report in Nature. "Several of van Gogh's works show Kolmogorov scaling in their luminance probability distributions. To the eye, this pattern can be seen as eddies of different sizes, including both large swirls and tiny eddies created by the brushwork."

While any visualization's success depends at least in part on the viewer's perspective and knowledge, sometimes the information is so complex that it requires more sophisticated and exact visualizations. Otherwise, the information gets lost in the transport or translation.

Some BI vendors offer that level of sophistication in their visualization palettes. A Sankey Diagram, which is very useful in describing the flow of information within a data set, is one example.

"For example, this visualization can show the process through which a banking customer transfers money, by measuring the cash flow per transaction. Sankey diagrams are useful any time you want to show information flow across distinct steps in a process," explained Daphne Tan, Product Marketing Manager at MicroStrategy, which produced the Sankey Diagram visualization below.

It may take some effort to teach your audience how to read some of the more sophisticated visualizations. However, it is well worth it if you need to regularly convey more than generalized information and don't want to pull a van Gogh. However, you'll find many audiences already familiar with these more exact metrics and data representations, including statisticians, engineers, and many professionals working in the sciences.

Here's a short description of some of the visualizations in this category worth considering and where you'd want to use them:

1. Arc Diagrams: These diagrams are uniquely capable of representing complex patterns in string data, meaning sequences that often also contain repetitive subsequences. Think DNA and streaming data from the Internet of Things (IoT). You'll find more in-depth information on Arc Diagrams in this paper by IBM Research.

2. Sunburst Chart: Also called multi-level pie charts, these are primarily used to visualize hierarchal data using concentric circles. You can create these in Microsoft Excel, for example. Below is an example:

3. Streamgraph: Microsoft and GitHub describe a streamgraph as "a stacked area chart with smooth interpolation, often used to display values over time." A flowing organic shape forms in this graph and the result can be both exacting and exasperating. Still, it has very valid uses such as displaying high-volume data sets to find trends and patterns over time across a wide range of categories. Yes, this is an open-source visualization so you can get it at the Microsoft Office store or on GitHub.

4. Hyperbolic Tree: Also called a hypertree, this visualization is inspired by hyperbolic geometry and is basically a way to draw a very large tree in a limited space while avoiding making a blob. You put everything on a disk rather than on a flat plane so that branches further away appear smaller. But you can drag those to you, making them larger and easier to examine. Hyperbolic trees show large information with detail and context within one view (as opposed to paging or otherwise summoning and portraying granular details in another view).

Visualizations Sprung from New Technology

There are so many different types of visualizations available today, you might expect that every conceivable means of visually depicting data has been made available already. Alas, no. New technologies and use cases inevitably spawn new visualization forms, too.

Augmented reality (AR) and virtual reality (VR) systems immediately come to mind. BI vendors are already working on unique visualizations for these systems. One example is Vantage Data Centers' new data visualization system in a 3D, VR-enabled virtual tour system. It looks like this:

"We first launched the concept3D platform in May 2017 to help us promote our new data center facility in Santa Clara which, at the time, was under construction. The platform is incredible when you're trying to market a building that doesn't exist," said Steve Lim, Vice President and Head of Marketing at Vantage Data Centers.

Data appears as an overlay on the screen in VR but that venue alone would be too limiting.

"In the near term, we anticipate most of our clients and employees using the system without VR on their desktop or mobile. It's impressive to see for the first time, and there's great potential for how this system can help us with operations and accessing real-time data from anywhere in the world," Lim added.

Choose According to the Task

Each type of visualization is built for a specific analytical task such as distribution, composition, relationship, or comparison. Make sure you understand each task and choose visualizations accordingly. For example, understanding product sales on holidays such as Christmas is a relationship study. Good visualization choices for that would be scatter plots, word clouds, and Venn diagrams.

Understanding whether coats or tires sell better is a comparison depiction. Bar graphs, pie charts, bullet charts, and line charts are good choices here. Depicting market share and competitive analysis is a composition task. Consider stacked bar/area charts, pie charts, waterfall, or any of the tree maps, depending on how much information in context you need to display.

Distribution tasks involve understanding what types of goods are sent to which stores and/or stored at which warehouses, as well as visualizing how resources are distributed by governments per various demographics. Good visualization choices include histograms, strip plots, and box plots.

"In this case, we want a view where we can see all data at once and try to find the range of values, shapes, or outliers," explains Patrik Lundbald, Visualization Advocate at BI and visualization software company Qlik .

Checklist for Choosing a Visualization

1. Know Your Audience: Pick a visualization your audience is most likely to find relatable and engaging. So, if funny hotdogs in an infographic depict your sidewalk vendor sales best, go with that. But don't skimp on information if you're conveying it to an audience steeped in statistical, data science, engineering, or other top-shelf skills. Pick a visualization that will deliver the detail and context they need to act on the information, without having to sort through a seemingly endless pile of related visualizations.

2. Make Clarity Your Top Priority: Be clear and be concise, even with highly detailed and complex information. Your goal is to produce easy-to-read visualizations even if the content is anything but.

3. Pay Attention to Every Detail: So you'd like the bar graph in this BI app to convey this piece of information. But do the bars correctly relate to one another or is the scale off? Details matter. Everything in every visualization is telling a story. Make sure you're telling the story you meant to tell.

4. Plan to Prevent User Fatigue: Too many visualizations tire the viewer out, as do unfamiliar depictions or overly complex graphics. Deliver the information in a precise and short narrative so the viewer stays engaged and remembers what they learned. Limit the number of visualizations in dashboards and reports.

5. Test Visualization Forms: Visualizations are like jokes. If you have to explain them, then you failed. It has to be able to convey the information with minimal text. Before you begin using a visualization routinely, test it on people who aren't close to the subject matter. Choose people who must find the information in the visualization rather than those who already know it. Are they enlightened or confused? If confused, then choose another visualization form or prepare to educate your audience.

"Unless it's specialized information requiring deep knowledge in artificial intelligence, blockchains, petechial hemorrhaging, or quantum physics, the visualization best serves the reader when it can be interpreted on its own, not just with the article context," said Mark Nicholson, Vice President of Marketing and Business Development at NiceJob, a social media/customer review reputation-building company.

6. Remember van Gogh: Complex information can be lost in a deceptively simple depiction. Therefore, a simple visualization may not be the right choice. Focus on conveying the information, that's the important thing. Also, van Gogh taught us that colors are not the only or even the best way to quickly convey information. Van Gogh's "magnificent brushwork made use of a property known as luminance, a measure of the relative brightness between different points. The eye is more sensitive to luminance change than to color change, meaning we respond more promptly to changes in brightness than in colors," reported NPR. Use varying levels of brightness as well as color to highlight information or show movement.

7. Learn New Vendor Visualizations: Ask for tutorials, examples, and other information on any visualizations offered by a vendor that you do not understand. It's better to learn on the job than to stick with the visualizations you already know. Why? Because technology is changing and, as it does, even newer forms of visualizations will appear. It's like never updating or upgrading your phone. Sooner or later you won't be able to reach anyone.

8. Sometimes Automated Is Best: Some BI vendors put a lot of thought into their automated visualization feature. One example that comes to mind is Salesforce Einstein Analytics . The company has years of experience in customer, sales, and marketing analyses, dating back to their early customer relationship management (CRM) days. Their automated visualizations reflect that experience. So, if you're powering through sales and customer data day after day, relying on Einstein to handle the visualizations is a smart and practical solution. There's no reason to reinvent the wheel.

9. Consider the Narrative: Choose visualizations that enhance your narrative, that tell a story. Otherwise, you're back to depicting numbers and your business colleagues or boss isn't going to absorb and retain the information as well. Make sure representations are in context, use the correct measures (e.g., absolute values versus relative), and check for scale. Use colors to highlight important points but limit the number of colors you use. The visualization itself should not be the viewer's focus, the content should be.

10. Keep Your Task in Mind: Remember that visualizations are designed for certain tasks and use them accordingly. However, simple is almost always better than complex. The goal is to find the best, fastest, and clearest means to transfer information from machines to humans.

This article originally appeared on PCMag.com.