Previous Period Stock + Today's Movement = Today's Stock

Brendan Broughton
Brendan Broughton Active Partner
Third Anniversary 100 Comments 100 Up Votes 25 Likes
edited April 2020 in Platform

Hi All, I have a question which I've asked some others about offline and haven't found an answer too, so I thought I'd try the community.

 

I have a 'Stock' cube (by day) and a 'Movement' cube (also by day) - dimension of each are identical, and there is sparsity (64bit)

 

To calculate today's (closing) stock the dataflow was set up as

 

Stock (previous period) + Movement = Stock

 

This procedure needs to be run for multiple (15) days at a time to allow for backdated stock adjustments following weekly audits.

 

The issue I have is that the procedure is only updating the stock correctly for one day.  So if I have 15 days I wish to update for, the procedure will need to be cycled through 15 times, and each time it is run it will correctly calculate the next days stock. 

 

below shows what I'm getting...

image

Stock has been correctly calculated to 17/3/2018, but the Stock value showing for 18/3/18 is the 'movement' for 17/3 and 18/3, but not including the Stock for 17/3.

 

If I run the procedure again I get the following

image

Stock has now been calculated for 18/3,and 19/3 stock has changed accordingly.

 

When doing similarly structured dataflows previously (eg. in generating Balance Sheet) it has calculated all selected periods when using the same approach, although they have been monthly and they have also been smaller dense cubes.

 

I'm wondering why in this instance this method doesn't work. Is it a limitation cause by day dimension, the number of dimensions/size of cube (cube has approx 10000T cells per year), or is it to do with the sparsity that is in place (Density factor is <0.01)?

 

Things we've tried unsuccessfully;

  • offset rather than 'previous period'
  • turning off 'High Performance' (this may work, but the calculation time goes from 2 seconds to >1 hour, so it is not a suitable approach)

 

Calculation and screen performance appears to be fine at present, so our current work around plan is to use a loop to calculate each day (rather than re-structuring the cube as we're not sure if that is what is causing the issue).

 

Thanks

Brendan

Tagged:

Answers

  • Samuele Marchetto
    Samuele Marchetto Active Partner
    Third Anniversary 100 Likes 10 Comments 5 Up Votes

    Hi, I had the same issue with a smaller cube and sparse structures. I think that Board take a snapshot of the "stock" cube at the beginning of the procedure and don't update it day by day. For this reason you obtain only the value of the movement (at the first run, the stock cube is 0 for the day after 17/3). This behavior has changed after Board 7 (in Board 6 it worked as you desire). But this is only a my theory, we should to have the confirmation from the Lab.

    In this case there could be two approches:

    1. extract the cubes and run the calculation via ODBC (but Text ODBC are no more supported)
    2. using two service cubes and Yearly cumulated value function:
      1. srv1: keep the stock value for one day only (the first day of the period to calculate)
      2. srv2: keep the movement values for the 15 days only
      3. calculate the stock = srv1 YTD + srv2 YTD

    To evaluate the performances of the dataflow.

    Bye

  • Brendan Broughton
    Brendan Broughton Active Partner
    Third Anniversary 100 Comments 100 Up Votes 25 Likes

    Thanks Samuele Marchetto

    Good to know it's not just me. Seems like it might be the day causing the issue then. It's been a while since I have to set up a daily incremental balance, but it seems to work ok with month.  I did consider aggregating as per your second approach but i'll just loop the procedure for now until I hear otherwise.

    Cheers

Categories