Last week in Data School was focus on survey data: how to prepare the raw data to use it in Tableau, how to analyze it and best visualization practices. Having spent the last 8 years and a half working in Market Research companies in Spain I was really interested in learning the best practices to analyze survey data in Tableau, and Emma Whyte, our coach for this topic, did an awesome work teaching and showing us the best practices in survey data analysis with Tableau.
The first issue we have to solve when analyzing survey data is to prepare the dataset for Tableau. Survey datasets have a distinct row for each survey, but to use it in Tableau we’ll need a different row for each answer to each question and for each of the respondents. So, instead of a file with N rows = sample and the different questions in columns, we’ll need a file with a smaller number of columns, but much bigger number of rows. The basic columns we will need are a respondent ID, the question ID, response value, response label and respondents weight if we need to weight the sample.
To make all this data preparation, Emma teached us how to use this useful Tableau Add-In for reshaping data in Excel. But if you are going to work regularly with survey data and want to avoid human errors as much as possible I’ll recommend to use Alteryx to create a workflow that realize all this preparation, so you can also re-use with every survey you work in.
Finally, It’s very important to be very careful with the measure you are using for your analysis. If you have a liker scale of 5 items, and you start your analysis using the SUM of the answer value, you’ll be overweighting the higher values of the scale. This means that it’s very important to dedicate some minutes in each type of question (liker scale, single answer, multiple answer, etc.) to think the type of calculation you will need and how to do it correctly, using count of respondents, our count of values, or when to use a sum or records or sum of values. Spending extra time to think and check each calculation is always crucial, but even more in survey data, that is a bit more tricky to analyze in Tableau.
If you want to check out the dashboard that Nicco, Ravi, Sasha and myself create last week analyzing survey data from Surrey County Council, have a look at Ravi’s blog and Nicco’s Tableau Public Profile.