Week 6 of The Data School Down Under introduced DSAU5 to survey data. With excellent guidance from Tram Trinh and some supporting resources (e.g. Data Revelations Blog), we learned how to prepare and visualize single response, multiple response and Likert survey questions. At sessions end, I was left pondering about other survey questions. What about the Conjoint Analysis? So in this blog, I will first give some context into the value of the Conjoint Analysis, and later suggest 3 charts to visualize the Conjoint Analysis
What is a Conjoint Analysis?
Conjoint analysis is a statistical process that measures utility. In simple language, it tries to calculate the importance of different attributes for a certain decision. These attributes may include factors such as pricing, delivery times, branding and quality. You might be thinking, isn’t this accomplished with a Likert scale? While you are correct, Likert scales have the option of scoring all questions “strongly agree” for instance, meaning you will not know what attributes need to be primarily highlighted to sell to their desired product, and which attributes could have fewer resources allocated.
Why Should You Visualize a Conjoint Analysis?
Tables are a great way to get accurate information, but for the audience that doesn’t understand statistics, visual charts allow users to get a snapshot of the analysis very quickly and easily. When searching the web for charts that visualized the conjoint analysis, the best I found was a bar chart that sized based on the utility score. While it gave a clear picture of the importance of that survey, it looks disappointing.
How Would I Visualize a Conjoint Analysis?
While there are some variations in Conjoint Analysis, the main components are the attributes as dimensions and the utility scores are measures. Knowing this, I propose 3 charts that add value compared to a simple bar chart.
- Line Chart (Bump Chart if Cluttered) – Adds Time as a Dimension
Due to the over-arching purpose to highlight attributes importance, using the utility score (or using a rank calculation on the utility score if the line chart is cluttered), and plotting this against time gives instant insights on how the market has developed. Unfortunately, this still looked unimpressive, which lead me to…
- Copy the Premise of Survey Charts (Likert or Dumbbell Chart)
Segmentation is an important consideration when making decision making, so instead of looking at a variable of time, you could use the other survey questions to form categorize, and using these categories separate your conjoint analysis. Once having this, you could make a chart with each segmentation on the vertical axis, the utility score on the right axis, and shapes denoting each of the attributes. This gives an easy comparison between which attributes are the most important across different target audiences. The other approach is to once again use time, but this time format it into a dumbbell (or comet chart), showing how the market has changed from a previous dimension (usually time).
While these two approaches further enhanced understanding, I still thought there could be a different method of visualizing conjoint analysis data.
- Radar Chart
Radar charts have the purpose of comparing dimensions in multi-variate data. You can separate each radar chart for a certain demographic, and quickly notice which factors need attention. An added advantage of these charts is their compact nature, however, looking at area or length could be inappropriate when the utility scores are close together. Another issue that may arise is having negative utility, but I would use color in order to format the negative, instead of having a weird axis for the radar chart. The next question is how would I build a radar chart, and this will be covered in my next blog, “How to Create a Radar (Spider) Chart in Tableau”.
Thank you for taking the time to read my blog post. If you want to continue this conversation, feel free to connect with me on LinkedIn. Also, don’t forget to register here if you are interested in applying for the data school down under.
