question

Is your probability the frequency the product become to be part of process A, then maybe B, then maybe C and finally not a Process at all and get deleted?

Then I would use a dempirical statistical distribution. For each different probability I use a single dempirical global table - 1st column the probability in percent, 2nd column the logical answer true or false as 1or 0.

Depending on the logical true in a conditional decide I make the product to be an item of the process type A, B, C or nothing at all inside the OnCreation trigger of the source

int itemAssigned = 1;
item.labels.assert("Process","none");
if(dempirical("globalTableProbabilityName",getstream(current) && itemAssigned)
{
  item.Process = "A";
  itemAssigned = 0;
  // assign additional labels like outgoing port, process times and splitting value to more labels
}
// next conditional decide to choose if the item becomes a B process type

If you choose later to use the switch - case structure you should think of to assign numbers instead of characters or strings as label values.

Otherwise if the sum of all probabilities is 100 then you use a switch -case structure directly with a single dempirical global table. The statistical dempirical distribution returns 1,2 or 3 from the second column of the global table.

switch(dempirical("globalTableDempircalName",getstream(current))
{
  case 1:{
    // setting labels for process time, split value, ..
    break;}
  case 2:{
    // setting labels for process time, split value, ..
    break;}
  // and so on
}

Then you can use the label values to send the item to different output ports, set the process time or split the item in separators.