As access to high quality data becomes more common, more team members across the business organization are facing the challenge – and opportunity – to make more data-driven business decisions.
From traditional spreadsheet and data visualization tools to modern dashboard aggregators, many tool chains exist to help business users find the meaning in their data.
Data visualization tools that present a queryable visual interface offer a new approach to data analysis – allowing the end business user to explore the data and discover patterns without waiting for data analysts to run custom queries.
With all that in mind, here's Yet's introduction to data visualization and our top three characteristics of great visualizations for data discovery.
Data Viz – What and Why
For our purposes, data visualization is any visual representation of data. An Excel sheet generally doesn't count, but charts that Excel can generate do. Good data viz is structured in such a way that patterns in data become easier to see and understand.
Interactive data viz empowers the business user by putting them in the driver's seat, giving them accessible tools for data discovery. The bulk of data collected today is never analyzed; data discovery tools helps businesses find the trends and patterns that can help drive success and improvement in their organizations.
Before we dig in to modern interactive data viz, let's take a look at some historic examples to get an idea of what goes into a good visualization.The Tree of the Two Advents, Joachim of Fiore, 1202
This genealogical tree example is from the 1200s; this classic tree structure is still used in visualizations today. The genaelogical tree represents connectedness as well as time, all in relationship to the focal point here which is genetic descendants.
Our second historic example comes from the 1500s and is a map of the relationships between geometric shapes. This networked structure is in some ways similar to the tree structure above, but functions more like a network graph. De Savigny is mapping shared geometric properties of shapes in a hierarchical graph.
Diagrams, Florence Nightingale, 1853
This example represents mortalities by cause of death in military hospitals; Nightingale would go on to use these visualizations as a tool as she advocated for improved sanitary conditions in care facilities. Polar or radial graphs like this are good tools for showing proportionality across different measured properties.
Each of these visualizations was designed to help show the relationships and underlying patterns in data. In this way, the work we do today visualizing data is not that different from the work that was done hundreds of years ago. We still must make decisions about what kinds of graphs best represent the data we are analyzing, what connections we want to show and how to visually style the end view.
The challenge today is to do all of that while designing flexible interactive systems for data visualization. Instead of designing one view of one data set, we work with varieties of data sets and data modalities. We design not one single visualization, but systems of data visualization and exploration.
Design for Data Discovery
Visualizations designed for data discovery have requirements that go beyond static graphs. Here are a few of the characteristics important in designing viz for data discovery.
Interactivity allows the user to ask questions. A single fixed view of data is informative, but not explorable. The best interactive visualization tools let the user ask questions through direct interaction with the visualization itself.
This visualization shows every active satellite orbiting the earth. In this view, satellites are color coded by launch vehicle. Scrolling down, the visualization continuously represents fixed spatial arrangement, but allows the user to view different contextual information like launch vehicle, purpose, operator, etc displayed as color, by clicking on highlighted keywords.
Yet Analytics xAPI LRS Timeline
This next example is from Yet's xAPI LRS. We chose to build several interaction types into the timeline:
- Drag to scope time window
- Search and scope to view different actors, verbs, objects
- Hover to get date and statement count detail on the visualization
Building the control interface directly into the data visualization lets the user explore the data intuitively. Visual cues like hovers, highlights, and other reactive elements keep the business user focused on and connected to the data while they explore connections and patterns.
Multidimensional visualizations help the user see relationships between different variables. In truth, nearly all visualizations represent at least two dimensions of data – but many data sets have more than two dimensions, and smart design decisions are needed to keep multidimensional data views coherent and readable.
If you've got only two dimensions to represent – let's say value and price – a simple X / Y graph will do you just fine. But how do you organize overlapping dimensions when you've got four or more layers to consider? You need to move beyond just X and Y and start to consider other visual dimensions like color, shape, size, spatial organization, line thickness, and contextual elements.
Buzzy Drinks, Dirk Aschoff and Klaas Neumann for Scientific American, 2013
This visualization uses spatial placement, color, and shape groupings to convey information about caffeine density in different consumables. The organizational choices that the creators made present a relatively complex set of overlapping data points in a highly digestible format. The graph presents multiple high level groupings, solving for the problem of overlapping groups with transparent shapes. These high level shapes help us identify overall patterns in the data that would otherwise be difficult to see.
Yet Analytics xAPI LRS Statement Frequency
This is another visualization from the Yet xAPI LRS; here we’re conveying frequency of statements by actor, with an internal breakdown by object. An interesting detail here is that with the color system Yet uses – yellow for actors, red for verbs, and blue for objects – you can see that there are Actors (yellow) showing up here as Objects in statements. Color in visualization design can function much like color in wayfinding systems – consistent use of color can be used to give viewers quick contextual information.
Yet Analytics xAPI LRS Statements Over Time
Here we’ve got a graph showing stacked statements over time - note the visualization provides high level information, revealing details on mouse-over. The overall shape of the visualization shows us trends and breakdowns, and the hover gives specific details. This prioritization is a good way to keep your visuals from getting overly noisy.
This last two examples both show stacked breakdowns using color shades; by connecting the segments in the visualization with the labels below, we're connecting the user to specifics and giving them the ability to see both the high level and the more granular breakdown while avoiding cluttering the graph display with too many labels.
3. Visually Efficient
Efficient visualizations use graph types that match their data sets. Some data sets have obvious visual graph styles: percentages as pie charts, growth over time as a line graph, etc. More complex data sets often need custom graph designs, which can take many forms but which always need to connect back to what is most important about the data.
A good visualization typically conveys information in a dramatically smaller and more accessible footprint than the same data presented tabularly. Different graph types imply different kinds of relationships – it's important to use the ones that work with your data, not against it.
This is an example of an extremely space-efficient visualization representing annual weather patterns in different cities. One year's worth of weather data is represented in a single circle. Temperature ranges are both represented as variance in and outside the circle and as color; precipitation volume is shown as gray circles oriented around the year's circle. Visual cues that connect to familiar metaphors – red is hot, blue is cold; clouds are gray; time is a circle – help make this visualization quick and intuitive to read.
This is a custom visualization that Yet designed for the 2016 HP Global Learning, Economic, and Social Index as a summary indicator for each country. Each bar represents a scaled indicator in a larger group – Learning indicators are blue, Economic indicators are purple, Social indicators are orange. The center number is the country’s weighted score.
As in the Weather Radials example, the compact form also makes visual comparison between two different 'scores' quick and easy.
Yet Analytics xAPI LRS Outliers Graph
This graph represents behavior and performance outliers among actors in an LRS; we’re using a radial space to efficiently display individual outliers relative to standard deviation from the mean. The interface is clean with names placed near their respective dots; on hover the user gets more information.
Decisions made by visualization designers shape what the end users will see in the data. Well designed graphs are structured to help make the important patterns in the data highly visible to the end user.
Help people answer questions.
Ultimately, data visualizations exist to help people answer questions. As designers and programmers, it's our job to give them the tools to discover patterns in their data. The importance of data and the ability to use data to make smart decisions is only going to grow – so start thinking now about building the right tools to make that data useful.
Looking to get insight into your people data?