4 Replies Latest reply on Apr 4, 2018 10:57 AM by Stephanie Matthews

    Survey Data w/o Unique IDs

    Stephanie Matthews

      Hi, All!

      I'm a Tableau newcomer and am attempting to get our institutional (higher ed) survey data into Tableau. We currently house all survey data in SPSS, and I've been trying to read up on Steve Wexler's survey data pivot guidelines. My question is this: If our survey respondents don't have unique IDs, will that method work? And, as a follow-up, if it will, do we need to move the data out of SPSS?

       

      We collect hundreds of survey responses from students each month, and they take a survey for every course they complete. So, while they don't have a unique ID in the database they likely occur in the data multiple times, but are assigned months, course numbers, etc. for each survey as well.

       

      I have been able to do some work with the SPSS databases, but have the sense that my options are limited with the data in its current format.

       

      Thanks for any and all insight you can provide.

        • 1. Re: Survey Data w/o Unique IDs
          Dawnita Gilmore

          I think it all depends on what you are trying to do with the data. SPSS has some statistical offerings that may, at first, seem quicker/easier to complete than in Tableau. However, it all depends on what you want as an end result. I have former experience working with Higher Education Institutional survey data (your dataset sounds like course evaluations). But needing a unique ID for the student or the course all depends on what you are trying to accomplish in either database.

           

          Are you just trying to run descriptive statistics or something more like a t-test? That would be a determinant for me when I completed that type of work.

          • 2. Re: Survey Data w/o Unique IDs
            Dawnita Gilmore

            I also just landed on this resource from a different post and perhaps it can help you to make that choice as well. Reshaping the Tableau data often helped me to transform survey responses so that I could emulate some of the tasks I saw that SPSS could accomplish. I hope this is useful too.

             

            Reshaping Survey Data with Tableau 9.0 – Data Revelations

            • 3. Re: Survey Data w/o Unique IDs
              Sasha Hanna

              Hi Stephanie,

              Because you need to pivot the data to create all sorts of visualisations in Tableau, you will require a respondent ID to keep the responses per person unique when aggregating these:

               

              number of responses = countd(respondent ID)

               

              As for taking the responses out of SPSS, I personally use Alteryx to obtain the metadata of these files and do the data preparation. This saves me a lot of time in processing those files as they are always the same structure and require the same prepping, so I can just run a workflow. Ultimately you need three elements prior to visualising: question text to create a question helper file which you can then join onto transposed answers both in text and numeric formats. Getting Survey Data “Just So” – Data Revelations

               

              Hope this helps!

               

              Cheers,

              Sasha

              • 4. Re: Survey Data w/o Unique IDs
                Stephanie Matthews

                Hi Dawnita Gilmore,

                We use SPSS right now for standard summary tables (N, Mean, SD for each likert question) and multiple response sets for the "check all that apply" questions. We have to maintain those databases regardless of how we ultimately get the data into Tableau, for accreditation reporting. I was hoping to get a similar look in Tableau for VP dashboards but with a little more flare - something they could filter by  month and then dive deeper for program-level insight.

                 

                My understanding of the need for a unique ID is if I want to join any other datasets to my course evaluations (another year's database, or another survey for comparison of the mean or whatnot). With our surveys being anonymous that's nearly impossible. I don't need to aggregate anything by student, so that's why I'm looking for a workaround to the unique ID issue.