Hi there, this might help point you in the right direction… though it’s not perfect.
The matrix is defined on the tProbability Matrix] table. All of those records are linked to a single record on the tJunction] table. In turn, all of the records on the tOpportunities] table must also be linked to the same record on the tJunction] table to get this to work properly.
The TJunction] table contains a rollup field which puts all possible combinations of {Deal Type} and {Stage} into an array. Then, on the tOpportunities] table, that array is parsed via a MID()
formula to find a matching {Deal Type} and {Stage}, and return the associated percentage.
Hope this is helpful to get you started!
Hi there, this might help point you in the right direction… though it’s not perfect.
The matrix is defined on the tProbability Matrix] table. All of those records are linked to a single record on the tJunction] table. In turn, all of the records on the tOpportunities] table must also be linked to the same record on the tJunction] table to get this to work properly.
The TJunction] table contains a rollup field which puts all possible combinations of {Deal Type} and {Stage} into an array. Then, on the tOpportunities] table, that array is parsed via a MID()
formula to find a matching {Deal Type} and {Stage}, and return the associated percentage.
Hope this is helpful to get you started!
Great, thank you!! This seems like a far more elegant solution compared to my 31 line (!!) function of nested-IFs clustered within a SWITCH function’s options. Will try it and revert.
Regardless, you’ve definitely moved my solutions-toolbox forward a couple of notches. Thank you!