Skip to main content

Celonis Product Documentation

Multi-Event Log

Note

This section covers everything relevant to Multi-Event Log technology in the Celonis EMS.

Motivations and Opportunities

Questions you can answer with Process Mining:

  • How much time do you need to sell a product after it was ordered from you?

  • If you realize you don't have this product in stock but have to buy it from another vendor, how long does it take until it arrives in your warehouse?

  • Are you paying the vendor according to the best possible payment terms?

In the Process Mining world so far, these are three different perspectives on the data that would require three separate Data Models, three different Analyses and a lot of mental overhead for analysis. You can combine the three perspectives into one with the new Multi-Eventlog capability.

Limitations of traditional Process Mining

Traditional Process Mining follows one record type through the system. For this record type, all events related to the record are collected in an activity table and linked together with a case ID.

In general, an entry in an Event Log consists of:

  • Case ID: identifies the record that we follow through the system.

  • Activity: describes what happens to the record.

  • Timestamp: describes at which point in time something happened to the record.

That means each Event Log is specific to a certain record type, e.g. there can be one Event Log for everything related to Purchase Order Items and another Event Log for Sales Order Items.

The Event Log - or rather the record type we chose as Case ID -defines the perspective and section of datafor the visualization of the as-is process. A Data Model containing only a single Event Log limits the analysis to this one perspective. With such limitations, we can only analyze one process at a time with a certain focus.

For record types in the purchase, sales and production the following limitations can be observed when using single-perspective Data Models:

50731069.png

Figure 1: Single-perspective Data Models allow for in-depth analysis of single processes but lack inter-process insights

Opportunities with Multi-Event Log Process Mining

Modern organizations of course run a variety of different processes in order to achieve their desired business outcomes. Those processes do not stand alone but interfere with each other. Possible inefficiencies and their root causes are often correlated among different processes. As an example, inefficiency in the order process (O2C) may be partially or fully caused by an inefficiency happening in the purchasing process (P2P) inside the same organization.

So, while analyzing one process helps to find inefficiencies within that particular process, it does not provide insights into how inefficiencies correlate with events in related processes. Analyzing multiple process perspectives at a time bears the opportunity to capture such dependencies and remove operational friction more holistically.