Page 1 of 1
Feeder for cells with value based on rule
Posted: Wed Jan 17, 2018 5:04 pm
by kenship
Hi,
I have a small cube with cells which their value is populated by a few rules.
How to write feeder for these cells that is not based on another cell?
or is it better to omit skipcheck and feeder?
Thanks.
Kenneth
Re: Feeder for cells with value based on rule
Posted: Thu Jan 18, 2018 5:36 am
by gtonkin
This would be a very rare case as generally the values being used in the calculation should be coming from an assumptions or driver cube which could then be used to feed the result.
I worked on a cube recently which had values for a probability of default per client rating band-15 maturity buckets needed to be derived using a formula.
I chose to load to probabilities into the cube alongside the ratings and then feed to ALL buckets rather than hard-coding into rules.
Whenever a client tells me that the values will never change, I immediately build a model to handle changes!
In terms of feeders, I would try and feed from an assumptions cube unless you can populate the dependent values into the cube (or another cube) like I did above with my probabilities.
In terms of skipcheck vs no skipcheck/feeders - I would go with "it depends" - if you do feed and you get it wrong, performance can be worse than no skipcheck/feeders (depending on cube size). Going to suggest you try both options but don't forget that you may need to restart the model to compare results.
Re: Feeder for cells with value based on rule
Posted: Thu Jan 18, 2018 1:16 pm
by kenship
Thanks.
For now I'll probably skip the skipcheck. The size of the cube is so small that there's no noticeable performance issue.
gtonkin wrote: ↑Thu Jan 18, 2018 5:36 am
This would be a very rare case as generally the values being used in the calculation should be coming from an assumptions or driver cube which could then be used to feed the result.
I worked on a cube recently which had values for a probability of default per client rating band-15 maturity buckets needed to be derived using a formula.
I chose to load to probabilities into the cube alongside the ratings and then feed to ALL buckets rather than hard-coding into rules.
Whenever a client tells me that the values will never change, I immediately build a model to handle changes!
In terms of feeders, I would try and feed from an assumptions cube unless you can populate the dependent values into the cube (or another cube) like I did above with my probabilities.
In terms of skipcheck vs no skipcheck/feeders - I would go with "it depends" - if you do feed and you get it wrong, performance can be worse than no skipcheck/feeders (depending on cube size). Going to suggest you try both options but don't forget that you may need to restart the model to compare results.