Statistical Forecasting can convey some very meaningful projections for end users. This is especially true when other factors besides the past influence projections. By letting end users input variable values, we can allow for ‘what if?’ analysis, and better understanding of the projections.
In the demo, end-users are prompted for input values into the model. Months to Forecast, Average Discount, Average Units Sold per Transaction and the distribution of company-wide sales amongst departments are used as input values to the model.
Those inputs are sent from a Python/Flask web application to an R Server where an individual ARIMA-X model is run for every Sales Region in order to forecast Company-Wide Sales. The results of the model are then sent to a Postgres database that Tableau Server is connected to. After submitting the input variable values in the application and the process has run, the results appear, being rendered from Tableau Server.