1 Reply Latest reply on Jul 21, 2017 11:29 AM by Cathy Bridges

    What Constitutes Peer Review of Data?

    Michael Szymanski

      Hello

      Any ideas on designing a peer review process of milestones for data source and data content availability for Tableau?

       

      I have an idea for data milestones (Review Draft, Review Created, Reviewed, Review Complete) in that order.

       

      Some other data points to collect would be Data Source, Data Type, and Date Added.

       

      I am open to all ideas submitted on this topic.

       

      Thank you.

       

      Michael Szymanski

        • 1. Re: What Constitutes Peer Review of Data?
          Cathy Bridges

          Hmm... I have a feeling this is going to be something that is specific to your company and data.

           

          Some considerations might be what kind of data sources are you using (should someone review SQL code for a report?)? Do you have any supplemental "sources of truth" against which you could compare a new data source? Are you concerned about performance in Tableau? These are all things to think about.

           

          I can tell you that in our world, when a table was transferred from Excel to our data warehouse, we had someone connect each data source, create data tables in Tableau, and compare them side-by-side to see if the numbers matched.

           

          If we have someone writing code for a view or report that Tableau is connected to, another coder reviews the code, a Tableau designer tests the functionality of the data in Tableau, and yet another person tests a version published to Tableau Server to confirm functionality and performance.

           

          Then there's the end user, who may have additional concerns about the data. Maybe the data is correct, but for whatever reason they don't want to visualize it the way it actually is - they may want other filters or to change the scale, or whatever. And they may also find that the data doesn't "make sense", at which point you must determine what the data "should be" and figure out how to remedy it.

           

          Lots of variables in this discussion, I'm afraid.