One possibility is to create two worksheets with your sales forecast. In one worksheet, forecast using all your data. In the other sheet, filter your latest X months of data and forecast. Now create a dashboard and drag your old forecast on top of your current forecast.
Also, notice in the attached example, I changed the default for both forecasts to ignore the last zero rather than one months of data. You can see this in the Forecast Options dialog.
compareForecasts.twbx 995.3 KB
A follow-up question on your answer to Matthew:
I have actuals for total counts of calls during the period of 01.01.2013 through 12.31.2013 and for the 2014 YTD. I want to show the forecast of what 2014 would have looked like if I did not have the actuals for this year. In addition, I want to show the difference between the 2014 forecast (started at the end of the 2013) and the actuals for 2014. The issue is that I need to show this in a single graph... I made a dash board of two graphs, similar to yours in the attached workbook above, but I really need to see if Tableau has the capability of mixing together the forecasted outlook and the actuals into one graph.
Thank you so very much!
As far as I know (and I wrote it), it is not possible.
Forecast of variable length time series in the same worksheet is not yet supported.
If you created a calc which just filtered some trailing values in your original measure and concatenated them in the same worksheet, forecasting would zero fill the missing values to equalize the time series length and give you a terrible forecast.
Superior behavior would be to forecast missing trailing values rather than assuming they are zero which is commonly only a correct assumption within the time series rather than on either end.