Trace vs. Case Nomenclature

RuneRasmussen

The IEEE Draft Standard for XES does not contain a metadata definition for a 'case' – only one for 'trace'. The ProM Lite user interface refers to cases in the Dashboard where surely Trace is meant, e.g. 'Events per case' should be 'Events per trace' and 'Event classes per case' should be 'Event classes per trace'.

hverbeek

Dear Rune,

Yes, these terms are frequently mixed up. In general, we consider a case to be the entity that flows through the process in real-life (like a complaint, or a claim), which leaves a trace in the information system. As such, cases and traces are closely linked. If you only have the log form the information system, then the only thing you know about the case, is the trace.

Note that business users might prefer the term 'case'. The XES Standard is more technical, and hence uses 'trace'. I hope this clarifes it a bit.

Kind regards,

Eric. 

RuneRasmussen

Eric,

Thanks for your quick reply.

I understand that 'cases' could be the entity that leaves a trace but this is not covered by the standard - see page 7 of the draft standard in the link above. In this standard the entities that leaves a trace is merely described by the attributes of the trace (which can be used for filtering, grouping etc. in ProM).

So, it is confusing when the interface says 'Events per case' because (as you say yourself) if we only have the log we only know about traces and noting about the entity that left the trace.

Standards are necessary because definitions of words are important! They can mean different things to different people; the entity leaving the trace is potentially a confusing term. E.g when analysing a claims handling system should this be 'a claim' or 'a customer' or 'an account'? We could have several claims each with many customers and each of them could be related to many accounts (i.e. each of the relationships would be modelled as many-to-may). When you say that business users would prefer the term 'case' then we need to specify which business users we are talking about; the claims handlers would probably prefer 'a claim', the finance department might prefer 'an account' and other stakeholders will want 'a customer'. In the draft standard these would all be attributes of the trace object and we can filter or group by them in ProM.

My guess is that the word 'case' comes from a previous standard definition from March 2014 where a trace is described as 'the execution of one specific instance, or case, of the logged process'. The June 2016 draft standard still has remnants of this by saying 'A trace component represents the execution of a single case, that is, of a single execution (or enactment) of the specific process.' but this is not included in the metadata structure on page 7. Here the word the 'case' seems to be used as a synonymous to trace. Notice the confusion of whether 'case' means 'instantiation of the logged process' (aka. trace) or 'the entity that leaves the trace'. That is why we have standards to clarify our terminology.

My suggestion is to update the interface with the words used in the new standard.

Best Regards,

Rune