I worked on a small team to build a Tableau-based data visualizations to allow users to sort through large tables of data about US National Park soil and vegetation. We followed a user-centered design process in which we interviewed and tested our designs with users, including experts in the field. Ultimately, we created a pair of dashboards in which users could filter large amounts of data to spot correlations and get details about individual items. The dashboards were built using data visualization best practices (such as using proper visual encoding methods and supporting ideal data interactions) learned in a graduate information visualization course at UW.
For this project, we were tasked with choosing a data source and creating a Tableau visualization by following a user-centered design process and utilizing ideal data visualization techniques.
We began by locating a data source, which required researching the context of the data (how and why it was collected), finding related work, and exploring visualization possibilities in Tableau. We wanted to find a data set that was extensive and valuable but without strong existing visualizations. Eventually we chose a series of data sets from the United States National Park Service's Inventory and Monitoring Program. The data sets we chose contain vegetation data collected for several national parks across the United States. Each spreadsheet contained a list of land plots with all kinds of value for each, from topographic data (elevation, aspect, slope, etc.) to vegetation data (dominant plant type, alliance, etc.).
For our design process, we sketched out potential visualizations and dashboards, as well as explored the data extensively in Tableau. We learned what kinds of visualizations we would be able to support and created some prototypes that allowed for some interaction and exploration.
We conducted expert user interviews, including with one of the scientists from the National Park Service who was behind the data collection, by asking about the tasks they used the data for and by sharing the visualizations we had created at that point.
Our user interviews helped us refine our designs and identify new tasks to support. We iterated several times to ensure that our designs would fulfill the needs of the users we had interviewed.
Once we had a stable version that we felt met our users' needs, we conducted basic usability testing with both expert and non-expert users. With our expert users, we attempted to validate the utility of the dashboards for the tasks we needed to support. With non-expert users, we were mostly interested in testing the interface design since they couldn't comment on the usefulness of the dashboards. Through our testing, we found that our users could accomplish the tasks that we had provided them with, but that some of the labeling was insufficient and that there were some minor issues with the UI. We were able to resolve the minor issues that were uncovered during the usability testing without much trouble.
In addition, we wrote an in-depth paper documenting the entire process and results, and presented our visualization to our class.
Overall, I learned a lot about visualization 'best practices' and about how to use Tableau for creating visualizations.