When you look at the image below, do you feel a sense of urgency?
It would be surprising if you did. This is showing a user interface for a transmission control room where operators monitor and manage power grids. Although it may not seem like it, there’s a big problem depicted in the table. The problem is here:
In this example you’re shown a table of four violations, and it appears the most urgent of them has a value of 845 kV, which is 35 percent outside of the highest limit for that type of equipment. The story this data tells looks unassuming from the computer screen, but in the power grid itself could look something more like this:
Needless to say these values result in an emergency. Luckily most operators in a control room are rigorously trained to interpret the data seen in this type of UI. Operators in the US are required to undergo training simulations on a regular basis so they’ll be prepared during high stress situations. However, even a well-trained operator can be significantly impacted by having to search through data that isn’t relevant to the task at hand. Information overload is a real problem in UI design, and can be detrimental in a situation where people are working in a high-pressure emergency situation. In the book ‘Set Phasers on Stun’ we see many ways in which a poor user experience has had negative and even disastrous impacts on past technologies ranging from armed tanks to medical devices. But in today’s software renaissance where a growing body of evidence around best practices and user research exists, it’s now harder to excuse antiquated table-driven UI’s such as the one in our example above.
Although the UI in the example is quite complex, by applying a different design it can be nicely put into perspective. See if you can spot the same four violations with this image:
This is the same information as the previous UI, but imagined differently. Here we see a bright red icon to indicate our potentially hazardous bus bar. We’re shown its location, (the “BERM” station), that it’s a security analysis violation (which means it hasn’t happened yet but needs attention), and that this violation is dealing with voltage. That’s four key pieces of data shown almost entirely without words. Even a layperson can see the four violations on the screen and decide which ones to address first. This illustrates the benefit of data visualization for situational awareness. When relevant information is depicted visually, it can help operators quickly digest what they’re seeing and act on it.
It’s often said that a picture is worth a thousand words, and data that’s presented as a picture is no different. The success of the visual often lies in its simplicity and its ability to be true enough to the data to make sense but abstract enough to be understood quickly.
Take for instance the concept of generation control in a power grid. Operators monitor power grids to make sure the energy being produced is the same as the energy being used, otherwise called the system “load.” They do this by keeping track of the frequency of each. Two visuals for load and generation frequency might achieve this, but is that really what an operator wants to see? A glance at the deviation between the two might be more useful. Take for instance the below image:
We can see at a glance that the frequency of generation and load is deviating from the acceptable range by quite a bit, and is well on its way to timing out. We can also see that the deviation has been turbulent over the previous couple of hours, timing out at somewhere around T-4 hours with a low value, only to shoot upwards and out of range on the high end of the scale at 0.8 Hz. Even minimal training would be sufficient to glance at this image and know what it means.
So what makes a good data visual work?
Generally it works when something apparent such as colors or shapes are used to represent the data. The three states in this example are shown with three different colors, and can easily be used to draw attention to trouble spots from the previous 10 hours.
If the visual element is too abstract to convey relevant information, we’d need to add just enough information back in to be explicit. The color does a good job of telling us if we’re out of range, but what constitutes out of range, and how far out of range are we? In this case we consider normal values to be between -0.5 and 0.5. These numbers are required to understand the context of the number 0.8 and therefore add context to this particular visual.
Of course it’s important to remember that a visual can quickly become overcomplicated by data if it attempts to show too much information. A good visual is a summary, not a replacement for detailed data. Our frequency gauge example shows three colors for three states, but if we needed to show ten or twenty statuses with their own colors it would overwhelm the operator rather than help him. It’s helpful to decide on a couple pieces of key data and limit your visual to showing just those.
We’re also lucky in that we live in an interactive world, and a successful visual can also become a mode of navigation to the more detailed data. When an operator sees that frequency deviation is out of range he may want to know how many sources are causing it, which can be summarized in a tooltip. If he wanted to know more about our low frequency alert from four hours ago, clicking that section of the visual might bring up a table of sources with the offending source shown at the top. The most important information should be apparent at a glance, and subsequently important data can be shown incrementally at the user’s whim using a variety of interactive features. And, the same rules for static visuals also apply to interactive ones; a paragraph packed into a tooltip for instance will seldom be read.
These are just a few of the ways you can present data in a way that has meaning for the user, but there are many methods for devising and testing the impact that visualizations and interactions such as these examples will have on your specific user base. It’s advisable for companies who prioritize situational awareness to take visual design and research seriously as they improve their software. As we’ve seen from past examples, a convoluted UI can result in time wasted, property destroyed and in some cases lives lost.
A good UI may look simple in the end, but as we’ve seen it requires work to make something complicated look simple. Each visual requires a critical look at the data and the operator’s needs, a decision on the key indicators to show, how they are most meaningfully shown, and what (if any) data should be accessed from within the visual itself. It puts more work on the software designers, but results in less work for the operator, and ultimately, a safer, more streamlined business.