I don't know if this will help you exactly, it is a bit of a hack, but you may be able to find a way to get it to work for you.
Under some circumstances, tableau is willing to perform a calculation in a cell during a table calculation EVEN IF THERE WAS NO DATA THERE!
So what I have done is choose an extra dimension from your dataset to set across columns to create a number of cells that might be N forecast periods out from where you currently are.
Then you can use the INDEX() , (or FIRST() or LAST() functions probably) to see which forecast interval you are on.
Then you can use more table calcs to look up your project's start and end dates, plus recurrence interval in order to perform your calculation.
Here is a workbook that contains several intermediate calculations to demonstrate the effect.
You can do it all with just a single calculated field if you are interested.
That's great stuff David! I had a feeling I needed to involve an index, just not quite sure where.
That is exactly what I'm looking for from the side of the calculation, but now I need to figure out what I can set along the columns shelf so that these dates will extend further than the 3 values that are currently there.
I really appreciate the insight.
I agree you'll need to find a solution to figure out how many records you'll get as far as the prediction.
You could do a number of things, including for example adding a set of rows at the bottom of your data set.
I have done this before, you could say, add a new column called "Recurrance Interval", and populate a few rows containing numbers going from say 1-20 or 1-31 (daily for 1 month) and null in all other columns.
OR if you use your real dataset, you might have more different values in the column than exist in this sample, and it coudl resolve your entire problem.
1 of 1 people found this helpful
Posting from my phone here while wiring for the bus to get through the storm. David's solution (I haven't looked at it, just read his post) looks like one solution.
You could also check out http://drawingwithnumbers.artisart.org/tableau-data-blending-sparse-data-multiple-levels-of-granularity-and-improvements-in-version-8/ for an example of using domain padding to pad out data, using a single extra row instead of columns. Demo #2 of http://drawingwithnumbers.artisart.org/creating-a-dynamic-parameter-with-a-tableau-data-blend/ also does this. In addition, Joe Mako did a demo of using domain padding explicitly for forecasting in a Think Data Thursday awhile back, I don't have the link handy.