When modelling business processes (we use BPMN which will be executed via BPEL), to what extent should the analyst worry about implementation of that process vs capturing the business view. For example, I have a BPM which at a high-level: calculate payments for a period, create a schedule of instalments. Where I have multiple periods I could model my process as follows:
1a.1st pass through the process for year 1, creating a payment schedule
1b.2nd pass through for year 2, creating a 2nd payment schedule which may actually mean adjust the 1st schedule to add new payments and dates
OR
2. loop through the processing of both years and leave the activity 'create schedule' until all the processing is done so that only one schedule is actually created.
From a process view, option 2 seems 'neater' as it is only one activity of 'create schedule' (reviewers have looked at option 1 and ased why the need to go to that activity multiple times when the end result is you actually just want one output). From an implementation view though, option 2 may mean problems with managing state, as effectively the information from processing year 1 would need to be held in state until year 2 had been processed and all that information is then fed into the activity 'create schedule' which would need to work out which is the most recent and relevant info to use (there could be many periods worth). Option 1 allows for each subsequent period to build on /replace the previous period so that no matter how many loops through, each 'instance' of a schedule is effectively replaced with a subsequent one.
To what extent should the process analyst need to worry about implementation? Or is it actually just a process modelling style issue? Preference for one over the other?
Thoughts greatly appreciated.