2 Replies Latest reply on Mar 30, 2016 1:27 AM by Ian Burdin

    Cohort analysis

    Ian Burdin

      Hi all


      Im trying to create a cohort for the number of logins that people make to a website. Having read around this topic, it's clear that my workbook is missing a dimension for "first login date" or something similar.


      I've also read that I need to produce the DATETRUNC and then DATEDIFF as per the article below.


      Tips for Cohort Analysis | Tableau Software


      I've attached a workbook. The login is event 210 within the "event list" dimension but I don't know how to create a date field from this. The data is for a two month period between Dec 15 and Jan 16.


      So I'm looking to put people into the following cohorts, by frequency of login:


      - several times per day

      - once a day

      - once a week

      - once a fortnight

      - once a month

      - none of the above

        • 1. Re: Cohort analysis
          Dan Sanchez

          Hi Ian!


          For the initial cohort analysis we can find the first login date by using an LOD calc like this:


          { fixed [User ID] : MIN([Date Time]) }


          We can then create a line chart using [Date Time] and then drop our cohort calc onto the Color card.  For this one I've set the date level to be Week:

          cohort line chart.png

          In terms of bucketing the number of users that fall into the different categories you outlined, I think that's going to be a little bit more challenging.  I'm still working on that one but wanted to at least get the first part uploaded for you


          Thanks Ian!

          1 of 1 people found this helpful
          • 2. Re: Cohort analysis
            Ian Burdin



            Many thanks for that, very helpful.


            I'd be interested to hear if you can get the data into buckets


            Now that the first login date is created, am I now able to use a DATEDIFF between this and the date/time dimension?


            Anyway will wait to hear back from you, thanks again