The BPMN specification presents lots of technical definitions and rules, but it does not teach you how to create process models that are effective in their primary mission-maximizing shared understanding of the as-is or to-be process. To do process modeling effectively, you need to go beyond the spec and learn a basic methodology, best practices, and specific diagram patterns to use in common situations. That's what we teach in our Process Modeling with BPMN course, available online at BPMEssentials.com in public classes at BPM Institute events.
To illustrate the point, here are ten tips for effective modeling in BPMN.
1. Make the process logic visible in the diagram. This is absolutely fundamental, but is routinely ignored by beginning modelers. The BPMN spec describes various shapes and connectors that print in the diagram, plus supplementary detail in attributes visible only through the modeling tool or in the detailed documentation that can be generated using the tool. But maximizing shared understanding doesn't come from individuals privately examining your model through the tool, nor from digging through 100 pages of documentation. It comes from a group sitting around the table looking at a printout of the diagram, discussing it, usually thinking about how it could be better.
Effectively that means two things: First, label everything in your diagrams--not just activities, but subprocesses, intermediate events, gateways, sequence flows, end events and message flows. Some required attributes, like the duration of a timer event, may not be displayed in the diagram. If not, add a label to the event that replicates that information. If you can't see it in the diagram, it doesn't really count.
Second, show exception handling logic explicitly in the diagram. Unlike many traditional notations, BPMN gives you the tools to do that even if you're not a developer.
2. Make your models valid. While the diagram is the key output, a process model is more than a drawing, and a modeling tool is more than a drawing tool. A real modeling tool has the semantics and rules of BPMN baked in, and gives you a Validate button that can display a list of errors when you violate the spec. A free BPMN stencil in Visio can't do that. If you want others to understand your models, you need to start by making them valid, so you should use that Validate button and learn to fix the errors.
3. Make your models hierarchical. What makes BPM different from traditional management disciplines is its emphasis on viewing the business from a cross-functional end-to-end perspective. Capturing processes in flat models that take up thirty feet of wall space does not allow that end-to-end perspective to be consumed all at once. Instead, we teach a top-down methodology in which the top-level diagram shows the whole process on a single page, and uses subprocesses to expand process detail at nested diagram levels, so you can zoom in and out of your model to describe any level of detail. It may print out as multiple pages, but internally the integrity of a single model is maintained.
4. Label process activities VERB-NOUN. BPMN describes processes in terms of activity flows, where activities are actions. They represent work done in the process. Activities are not states, not business functions, not use case interactions. To reinforce that, we ask students to consistently name their activities using the Verb-Noun construction, like Check Credit or Validate Order, not Credit Check (a function) or Valid Order (a state). A resource is going to be performing each activity, and the name of the activity should describe what action is performed.
5. Specify task types. One of the BPMN attributes with no standardized representation in the diagram is the task type. The spec defines several task types, but there are really two that are important to distinguish: user (human task) and service (automated task). Fortunately, most BPMN tools distinguish task types with icons inside the activity shape--but you need to specify which type you mean.
6. Don't use a task to route work. Another common beginner mistake is to insert tasks like Send to Manager, followed by a sequence flow to a task in the manager's swimlane. That sequence flow is already routing work to the manager, so the Send to Manager task is redundant. Just get rid of it.
There's a second problem here as well. Best practice is to reserve the keywords "Send" and "Receive" in task names to send and receive task types, which are equivalent to message events. In BPMN a "message" means a signal between the process and some external entity. Here manager is not an external entity, but a participant in the process. So it's not a message and should not be labeled Send. If you want to communicate something to the manager without routing the work itself, you could use a task called Notify Manager. Practices like this may seem petty at first, but in the end they make it easier for everyone in your organization to understand immediately from the diagram what is going on.
7. Distinguish success and failure end states in a subprocess with separate end events. Each path in a subprocess that is enabled must reach an end event before the subprocess is complete. You can draw a single end event for the subprocess and route all paths to it, or you can draw multiple end events and route specific paths to each one. Since there is an implied "join" of all the end events, technically it doesn't matter. But there is benefit in drawing and labeling separate end events for each distinct end state of the subprocess, particularly if some end states represent "success" and others represent "failure" or some type of exception.
You can follow the subprocess with a gateway that tests the end state to see whether to continue the process or do something else, like end it or loop back to a previous step. Matching the label on the gateway to that of the end event makes that linkage clear in the diagram. Alternatively, an end event on an exception path inside the subprocess can rethrow the error, which is caught by an attached error event on the subprocess boundary. This also can end the process or follow some other exception flow. Again, matching the labels of the throwing and catching events makes the connection obvious from the diagram.
8. Use subprocesses to scope attached events. Intermediate events attached to a process activity mean if the event occurs while that activity is running, abort the activity and proceed down the sequence flow out of the event, called the exception flow. A neat trick is to wrap a sequence of activities with a subprocess for the sole purpose of defining that scope for the event. For example, if you have an order handling process with steps A through Z and you want to allow the customer to cancel or change the order without penalty any time between steps B and G, you can wrap the sequence from B to G in a subprocess and attach a message event and particular handler exception flow to that.
9. Standardize on specific diagram patterns to distinguish types of exceptions. BPMN provides a business-friendly notation for describing exception-handling behavior. Even though the BPMN spec gives the modeler great freedom, best practice is to learn specific diagram patterns to distinguish each type of exception, and use them consistently. In our course, for example, we teach distinct patterns for modeling internal business exceptions, system faults, timeouts, solicited response exceptions, unsolicited events, and others. Again, the key principle is understanding exactly what is meant just from the diagram itself.
10. Use message flows consistently to show business context. In addition to the activity flow of your own process, BPMN lets you show the interactions between your process and external processes as dashed connectors called message flows. Message flows typically represent requests, responses, and unsolicited events exchanged with the external process. In your own process, message flows connect to specific activities and events that indicate exactly how your process responds to an incoming message flow or generates an outgoing message flow. Since you do not control or even know the internals of the external process, it is common to just attach the message flows to the boundary of the pool representing that process.
Message flows can add valuable business context to your diagram, but it is important to use them consistently. For example, if you are going to show any message flows from and to a requester of your process, you really should show all of them, and show them consistently in each level of your model. That is, if a particular message flow is shown in a subprocess nested three levels down, it should also be shown in the top-level diagram, and labeled the same at every level.
None of these ten tips are mandated by the BPMN spec, but if you learn to apply them consistently throughout your organization, you will find it much easier for others to understand your models--and for you to understand theirs--immediately just from the printed diagram. And isn't that the real goal?
Bruce Silver is an independent industry analyst covering BPMS technology and the author of the 2006 BPMS Report series on bpminstitute.org.
Timely, incisive articles delivered directly to your inbox.