Data visualization takes something complex and makes it simple and easy to understand. Large data sets can contain insightful gems of information, but those are hard to unlock from the raw data alone. Often, a well thought out chart, graph or table will make those insights easier to understand and extract.
At edo, we are always looking for ways to improve our data visualization skills. Recently, we had the opportunity to attend a talk by Edward Tufte, noted statistician and professor at Yale. Tufte is passionate about beautiful and meaningful data presentation. Because so many areas of our edo Marketplace rely on summarizing large datasets, finding new and better ways to explain and display our data has become one of our passions as well. Tufte uses six principles for analyzing and presenting data, and that structure has served our team well.
Show comparisons, contrasts, and differences
Making intelligent comparisons is why we do statistical analysis in the first place. One of the first visualizations I worked on at edo was the Network Analyzer, a page to help potential customers compare their business to our network of cardholders. The bubble charts make it easy to compare the relative size of cardholders or dollars spent per group, and the color coding keeps things consistent when switching between the graphs.
Show Causality, Mechanism, Explanation, Systematic Structure
A good data presentation can show important comparisons within the data, but a great one can help you understand why the data is a certain way, and what conclusions you can draw from it.
There’s a great example of this from another excellent visualization website called Perceptual Edge. A graph from the website of SAS provides an example of how to use their analytical software. The final iteration of the design leads to explaining seasonal patterns and provides a contextual reference to the freezing point.
Showing more than one variable with multivariate data
While we like to think of a lot of things in black or white, it’s a truth that nearly everything that is interesting enough to try to understand is multivariate in nature. In Tufte’s book, The Visual Display of Quantitative Information, he presents a great example from Charles Minard titled Napoleon’s March. Published in 1869, it is famous for weaving in five variables all onto a two dimensional map.
Completely integrate words, numbers, images, diagrams
Why separate words from your graphs and numbers from your diagrams? Graphs should be annotated with words and numbers: it provides clarity and doesn’t limit the scope of how you may present your data.
This is a personal favorite of mine, it encourages creativeness in the design of your visualizations. In this graph to the right, we wanted to show bar chart comparisons, but the exact dollar amount annotated on top of them also provides accuracy.
While citations are critical in the academic world, it’s just as important to source your data no matter the environment.
The worst offenders of this rule are frequently infographics. Often, there is no link to an article or infographic’s source data, dramatically reducing the trustworthiness.
At edo, we include sources in our infographics, so you know exactly where our data is sourced.
Content counts most of all
Probably the most important principle of all. The edo team’s motto is, “Content is king.” And rightly so, without good content, you’re pretty much limited in what you can do. At edo, we don’t have that problem, the data we get is great and one of our best assets. Uninteresting visualizations are a data problem, not a design problem.
Looking forward, there’s a lot of good challenges on our roadmap to apply Tufte’s principles.
We’re particularly excited about performance reporting of campaigns and deals, showing the data in user-friendly ways. On advertiser dashboards, we are exploring ways to give them a quick overall glance at campaign progress – all to help our partners leverage data to make smarter business decisions.
Continuing that idea, we’re also building a more detailed deal performance page that drills down into every piece of data we have, and present a thorough analysis of how well the deal did. This is a first draft, and a lot of it isn’t hooked up to real data yet. We like to make working prototypes so we can fully envision how a visualization would perform, and after getting feedback and a few iterations, we’ll get the data wired up, confident that we’re providing it to the most effective visualization possible.