At the end of our 4-month intensive training to become Data Consultants we practice our learned skills through the so-called "Dashboard Week". It is quite a fun invention! At 9am we receive a dataset and have until 3.30pm to clean the data, sketch out a dashboard, build the dashboard, write a blog post, and prepare a small presentation where we walk through our work. And we take 1 hour lunch break too! So we really have 5.5 hours to build a dashboard.
It is fast-paced and a good lesson in letting go of your perfectionism and scoping down!
The Data
For Day 1 we received a dataset on Slovakian Youth's preferences. The data was based on a survey and most questions where answered on a Likert scale from 1 (Strongly Disagree) to 5 (Strongly Agree). The dataset also included a lot of information on various different topics such as music, movies, spending habits, hobbies, phobias, health, and demographics.
The Data Cleaning
It is impossible to build a dashboard with all of that information. Therefore, I decided to only focus on a subset of the questions and used Alteryx to de-select any columns that I did not need. I early on decided that I wanted to look at music preferences and mental health so I only kept questions that focused on these topics.
Because the data was in a very wide format (each question had a separate column), I decided to transpose the data to have one column for different music preferences and then another column with the survey response. To not explode the data later in Tableau, I also had to create a Record ID so that each music preference can be mapped to each respondent.
Because multiple questions dealt with positive and negative mental health aspects, I decided to combine a few questions that very asked in a positive way (e.g. "I am always full of life") or negative way (e.g. "I worry about my health") and take the average across those questions to create scores for positive and negative mental health.
In the following you can see my (very short) Alteryx flow:
The Dashboard Sketching
Even though my final dashboard did not end up looking anything like my sketch, it is good practice to sketch out your dashboard first. This will make building charts easier, you will have a better understanding of the data, and you can imagine how different charts interact with each other.
From my initial sketch I only kept the bar chart on the left, which depicts the Likert Scale for all different music genres and the average across all respondents.
The Dashboard
I then spent most of the time building the actual dashboard. As with most data visualizations you spend a lot of time exploring the data, seeing what is possible, and what ways may convey the information the best. By exploring the data, I also slowly got an idea for a story. In my bar charts I saw that even though the music genre "Rock" is the most popular, the respondents who liked "Country" music reported greater positive mental health. This insight was what I wanted my dashboard to focus on.
When building charts, I discovered a way to visualize data that I had seen before, but never tried out. I decided that Dashboard week would be the ideal opportunity to try out this chart. I plotted each respondent's survey response as a little dot in a larger circle for each of the music genres. I then colored each dot according to the survey response, e.g. whether people enjoyed or did not enjoy a particular music genre. This way, the user can see quickly which genre was more popular than other genres.
A darker outside of the circle indicates greater popularity. If the center of the circle (the light pink) is bigger, then more respondents answered that they do not enjoy this particular music genre.
To practice some Dynamic Zone Visibility, I also implemented a parameter where the user can choose between different visualizations without overcrowding the dashboard.
The final result looks like this:
What do you think about this exercise to build a dashboard in only 5.5h? Feel free to start a conversation with me on LinkedIn: https://www.linkedin.com/in/lisa-hitch/
Screenshots by Lisa Hitch in Tableau 2023.3