3 Replies Latest reply on Nov 14, 2017 4:06 AM by Norbert Maijoor

    Building your own forecast models in Tableau

    Wolfgang Erb

      Hi guys,

      I recently started playing around in Tableau with creating my own forecasting models as the functionality in Tableau is limited at best and only allows you to forecast single items. My company has over 4000 different products listed and forecasts need to be generated for each of them. Furthermore, the statistical forecasts then need to be reviewed by the sales departments and adjustments need to be made.

      To this end I am planning on abusing Tableau as a data generation tool (I know, I know..) so that I can export the generated forecasts in csv format into an access database for editing since the editing part is not in Tableau's sphere at all.

      Right now I am playing around with a straight line regression model (y=ax+b) with seasonal indices where x is my time period (So the final forecast formula is: y=(ax+b)*SF). Using fixed level of detail calculations I worked out my a, b and SF factors per product and then created a calculated field per forecast month to calculate the forecast. The one drawback currently is that I have four levels of nested fixed level of detail calculations that absolutely destroy my computers performance. Anyone here interested in this line of thought that would be willing to dig a little bit deeper with me to optimize my current calculations and build further more advanced forecasting models?

      I'd be happy to share with you all the current work I have done so far on the matter.