Perspectives
A perspective contains a set of object types, event types, and relationships that you want to analyze. It's called a perspective because it's a particular view on the object-centric data model. Perspectives can contain Celonis object types, event types, and relationships, or custom types and relationships that you have created, or a combination of both. They let you apply different viewpoints to the same modeled objects and events, so you don’t need to ingest and transform the source data again to model each process.
We've prebuilt perspectives for the Accounts Payable, Accounts Receivable, Order Management, Procurement, and Inventory Management business processes using the Celonis types and relationships. When you enable one of the Celonis processes, you’ll get the perspective that Celonis needs to identify and use the appropriate object types, event types, and relationships. You can add object types, event types, and relationships to extend the Celonis-supplied perspectives. You can also create complete custom perspectives using the perspective builder.
Perspectives are important not just to save time on analysis, but also because the query context is important to Celonis. In the object-centric data model there are often multiple relationships between a pair of object types. To construct a PQL query, Celonis needs to know which relationship is relevant for the analysis. A perspective includes only the relationships that need to be used in the analysis, so that the query context is unambiguous.
Perspectives can include standalone helper objects that provide reference information to PQL functions, such as a factory calendar table, workday or weekday calendar, and quantity or currency conversion tables. The Celonis catalog processes include the CurrencyConversion and QuantityConversion helper objects. You can also create and include your own reference tables, and other types of helper object such as lists of values. Previously, all the object types in a perspective had to have a path of relationships between them, but now a perspective can include distinct groups of objects that are connected to each other but not to other groups.
You can restrict end users' data access by setting data permissions on a perspective in the data pool, in the same way as you would for a case-centric data model. For more on this, see Data permissions for object-centric process mining.
Event logs
Some Celonis apps and tools, such as the Supply Chain Network Visibility App and the object-centric versions of our Starter Kits and applications, work with perspectives directly. Other Celonis apps and tools, including Process Explorer and Process Adherence Manager, as well as the case-centric versions of our Starter Kits and applications, work with event logs that are created from perspectives.
All of the information that you need to construct an event log for a particular case is held in the objects and events that you've created using your business data. Event logs pull data from your objects and events into a format focused on the events that happen to a single object through its lifecycle.
We automatically create an event log for each object in a Celonis-supplied or custom perspective, using the events which are connected to that object in the graph. In your Knowledge Model, the automatically generated event logs start with the prefix
el_, a namespace, and an object name - for exampleel_celonis_SalesOrder. These are to get you started, and you might not need to create any other event logs besides these.For extended or custom perspectives, you can define additional event logs manually. Custom event logs let you include further events that are indirectly connected to the lead object, through another path of object to object relationships. You can choose what object type is the case for a custom event log and what event types are included in it. In your Knowledge Model, the manually created event logs start with the prefix
el__, with a double underscore, and no namespace - for exampleel__MyEventLog.When you enable a process from the Celonis catalog, we'll install some custom event logs to support the most common use cases. In the Knowledge Model, our custom event logs are named in the same way as yours, with the prefix
el__, with a double underscore, and no namespace - for exampleel__SalesOrderActivities.You can set any object as the lead object in event logs, including standalone objects and objects that are part of a distinct group that isn’t connected to other groups. You can have multiple different event logs for the same lead object, including its automatically generated event log and any number of custom event logs.
You can select attributes of an event type to create subsets of the event for analysis. For example, you could analyze each specific delivery method for the ReleaseDelivery event separately. You can use any number of attributes for any number of events in the event log to subset them. We set up a separate column ending in
_ActivityDetailsfor events and attributes, so you can also use the event log’s_Activitycolumn to analyze events without subsetting them.If you're using any applications and features that expect a single event log rather than multiple event logs, such as Analysis, you need to create a custom event log and mark it as the default event log for the perspective. You can't mark an automatically generated event log as the default. If you don’t set up a default event log, then the perspective won't have one, and PQL operators that expect a default event log (for example
ACTIVITY_TABLEandCASE_TABLE) won’t work.The event logs create activity tables in the Celonis Studio for backward compatibility. You can create views to analyze these tables, or write custom PQL to work with them.
You can also dynamically create an event log in Studio using the PQL operator
CREATE_EVENTLOG.
(We called event logs “projections” in the first release of the object-centric process mining user interface and information, so if you’re looking for that term, you can just use “event logs” now.)
Data extractions and transformations for object-centric process mining explains how data is ingested to the object types, event types, and relationships to create objects and events.