Message flows in BPMN 2.0 for a question-answer process

Question

I am just learning BPMN 2.0 modeling language. I have to design a call center process where an operator asks several questions to a customer. I am wondering if I need to model a message flow between customer and operator for every question (as the exchange information) or if a message flow is only needed if real documents are exchanged (e.g. an invoice is sent). Thanks!

Answer 1

Let's look at the definitions in the BPMN 2.0.1 specifications (highlighting from me):

Message: An Object that depicts the contents of a communication between two Participants. A message is transmitted through a Message Flow and has an identity that can be used for alternative branching of a Process through the Event-Based Exclusive Gateway.

Message flow: A Connecting Object that shows the flow of messages between two Participants. A Message Flow is represented by a dashed lined.

So the message is not necessarily a business document that is exchanged, but any kind of communication, including spoken exchange.

The granularity of the messages depends on what you want to show in the model:

• If the question/answer is just a detail in a larger process, you may want to show one summary interaction corresponding to the questions and another in reverse direction for the answers, and consider that all the questions are asked at once and all the answers provided at once, even if this does not fully correspond to reality.
• If you're interested exactly in the details of the question/answer because for example different participants are involved at different stages of the Q&A, you may well single out each question and each answer separately according to the real flow. But be careful, since too detailed flows might be difficult to maintain, so some degree of abstraction is recommended.

If your interest is in very detailed dialogue scripts between a customer and the same operator, BPMN might not be the best tool: more specialised conversation flows, job stories or intent based scripts could be more appropriate (see more about the topic in this article on chatbot requirements)