Prescriptive Analytics in IBC?

Dear Celonis community,

I was wondering whether prescriptive analytics / recommendations (which are not out-of-the-box) are possible in Celonis? E.g. is it possible to define a goal (a particular performance indicator) and suggest possible actions to choose from (e.g. option A, B, C) in order to reach that goal?

Example: As a production manager I would like to cut throughput times by 15% - What should I do next in order to best achieve my goal?

Thanks in advance!

Best,
Daniel

Hi Daniel,
yes, for sure, this is what Celonis Machine Learning is all about! Since you are talking about non out-of-the-box prescriptive analytics/recommendations, this is how you could go about implementing such a use case (really high level):

  • Find impacting factors (e.g. features of your data) on your goal, throughput time in this example
  • Investigate how the performance increases by estimating how big the impact would be when you are changing this impacting factor (e.g. through sampling the data).
  • Define actions that you want to take to increase performance. This can be tricky, since the relationship between the action to take and the decrease of the throughput time might be non trivial
  • Create rule or ML model that classifies your cases as something needing action or not needing action
  • Setup Action Engine Skill that notifies users about cases needing action
  • Connect the Signal resulting from the Skill to the previously defined Actions
  • To measure performance, you could track the relevant metrics in the Transformation Center

For an end-to-end demo of such an implementation, have a look at the following video:

Hope this helps
Nicolas

1 Like

Thanks a lot Nicolas, that helped a lot!