You want to create a data visualization presentation that is engaging and persuasive. The crucial first step is to ask good questions. Sometimes you already know the specific issue that requires greater understanding. Other times, questions are loosely defined or yet to be discovered. This article covers data set exploration and the search for insights.
HOW DO YOU FIND A GOOD QUESTION?
When analyzing a dataset, think of yourself as a journalist. Know the subject matter. Find something that is of interest to you. Make it relevant to your audience.
The first step is to understand the data. Get familiar. Identify any data quality problems. Discover first insights. Finally, detect interesting subsets that warrant a closer inspection.
Tableau hosts an annual competition called Iron Viz. During this event, competitors are given large datasets. Everyone is using the same tool, Tableau. It’s up to the competitors to find insights.
For the 2017 Iron Viz, the competitors needed to tell compelling stories using U.S. housing data from Zillow.com. The dataset had millions of rows of data, lots of detail, and lots of metrics across time and geography. The dataset included median house prices from 2000 to 2017 by city and state. All three competitors chose to focus on the housing crisis of 2006 to 2011. There are many ways to query the data for insights on the housing crisis.
Here are some possible avenues of inquiry:
- Were fluctuations in housing prices the same at every price range?
- Were fluctuations in housing prices the same geographically?
- Which cities experienced the greatest price drops?
- Were any cities or states immune to the home value decreases of 2006 to 2011?
- How did each state compare to the national aggregate?
- Just how inflated were home prices leading up to the housing crisis?
As it happened, the winning data viz had an animated time sequence which vividly showed by state the increasing home prices followed by plummeting home prices.
For the animation, see the Youtube video, www.youtube.com/watch?v=lP7r_G1k0FU&t=2085s, start time 34:20.