How to auto generate forecasts with Tableau

Excerpted with permission of the publisher, Wiley, from Tableau Your Data!: Fast and Easy Visual Analysis with Tableau Software. Copyright 2014, all rights reserved.

Tableau Desktop users can create forecasts with a couple of mouse clicks. The resulting figures can be exported, revised and possibly added to your datasource -- providing a fast and easy way to model the future based on past history.

How Tableau creates forecasts

Tableau generates forecasted values by using time-series data that is included in your worksheet. Figure 6-1 below shows a time series chart that includes forecasted values.

Time series with forecast
Figure 6-1: Time series with forecast

Forecasted values are presented in a lighter color than the actual values. The forecast values in Figure 6-1 can be added by right-clicking within the worksheet, and selecting the Forecast/Show Forecast menu option. You can also generate the forecast from the Analysis menu/Forecast option.

Forecasting Options

Tableau can forecast data in a variety of ways and will automatically select what it feels is the best method. If you dont want to accept the default, edit the forecast model by right-clicking on the worksheet and selecting the forecast options menu. Tableau provides the following forecast trend models:

  • Trend and season
  • Trend only
  • Season only
  • No trend or season
Tableau forecast options menu
Figure 6-2: The Tableau forecast options menu

Depending on the amount and granularity of the historical data, each option will generate different results. Trend and season will generate the most volatile forecast data. The forecast options menu includes several other variables that can be adjusted. You can see the forecast options menu in Figure 6-2.

At the top of Figure 6-2 you can see that by default, Tableau will generate a 12-month forecast, but it is possible to forecast a specific number of periods into the future. The number of periods that Tableau forecasts is dependent on the date range in your view and the data aggregation level presented. The Ignore Last setting allows you to omit incomplete historical data so that it wont skew the forecast results. Checking the Fill In Missing Values With Zeroes box will prevent null values from corrupting the forecast.

Review and present forecast quality metrics

You can examine the quality of the forecast that Tableau generates by right-clicking within the workspace and selecting the forecast/describe forecast option. Figure 6-3 shows the summary tab.

The summary tab provides details regarding the precision and quality of the forecasted values. You can choose to express forecast precision as number or percentage ranges. Forecast quality is described as poor, OK, or good. Clicking on the Models tab exposes more detailed quality metrics that you see in Figure 6-4.

Tableau forecast summary
Figure 6-3: Forecast summary
Forecast model description
Figure 6-4: Forecast model description

Tableau's forecasting model weights recent history more heavily. The statistical models for the different quality metrics presented in Figure 6-4 above are defined in detail in the Tableau Desktop manual. The smoothing coefficients for alpha (level smoothing), beta (trend smoothing) and gamma (seasonal smoothing) refer to the amount of smoothing applied. Values closer to one are smoothed less than the lower values. If the value is very close to zero a lot of smoothing was performed.

Computerworld's IT Salary Survey 2017 results
Shop Tech Products at Amazon