Activity granularity in process mining: a challenge

Here’s a great article on the biggest challenges in BPM that you should read if you’re interested in the future of process mining (do yourself a favour and avoid those inane listicle articles if you’re actually serious 👍 ):

https://lnkd.in/eiWrwqay

There are several interesting problems raised, but the one that resonated most with me was our issue of selecting and extracting the appropriate ‘activity granularity’ in process mining. (Page 7 in the article if you’re following along.)

When creating event logs, the analyst has an inordinate – and almost always unacknowledged – amount of leeway in determining what the final analysis can deliver. Often, due to extreme time pressure, activity granularity is selected based on what’s easiest to extract from a source system, which may not be what’s most valuable to the business. Value is left on the table (literally 🤓 ).

As suggested in the linked article, imagine a ‘multi-granularity’ selector bar in a process mining tool. That would mean more work on the event log, but would make analysis of the target process more useful to more stakeholders. Think about how cool it would be to move between the different levels of a target process in one place.

Right now, it seems to me at least, many process mining analyses are ‘brittle at birth’ because the event log creation method is too often ill-considered, strung together in a messy pipeline, then left untended until its totally broken and needing a fix.

Better process mining means more resilient, more flexible, and multi-granularity event logs.