Modelling a unique process with multiple and different roles (for example A, B, C) where some activitys don't apply to some roles.

Hi Patricia,

My take would be to generate a generic process which does all the activities and then create variant processes which model the specifics for each process.

This way you do not create individual 'stovepipe' models which can not be compared to each other.  However, I'd also wait to see what the experts opine as well.

Salu2

Hi Patricia,

It seems to me that you have a problem scoping your model.

For EPC models, one key thing is essential: One Time, One Person, One Place.

Within BMPN, this however, is not required and provides confusion for modelers somewhat.

Make sure the activities are as role exclusive as they can be. Most of the time, the reason that someone else can or must do a specific task, is because of an escalation. This is not the same process! So make sure you are as specific as possible in declaring processes and their activities.

However, there is a way to model this, using split and join gateways. In the current example I've added a WRONG interpretation on purpose, to challenge you how you need to resolve your issue:

The case is modeled as described - but the description is incomplete.

On the left hand side you see a split and join activity for two scenarios for roles A and C going through the same processes. They are defined in such a way that they would have a different start event. (For example a telephone call and an e-mail starting a sales process).

In the right hand side, we see a split (exclusive) with two end events for roles A and B. The two end events imply that there are two different end-states possible based on the role handling the processes; a third end-state is possible with the end event for role C.

There are many ways in which your scenario can be modeled to be sufficient, however more information is needed.

For example, is there only one type of outcome? (then you should use a exclusive or join and one end-state, with a responsible person for this end event in the correct lane)

Are there escalations for the roles? (Then model with intermediate events and show that it is an escalation and rename the processes to resemble the escalated activity chain).

Are the processes or activities on the same level of detail? (Are there activities and processes that are represented as subprocesses? Model with care, and apply them within the correct process descriptions)

Are the incoming events one and the same? (Use split exclusive or signs and make explicit when what flow is required)

There are more aspects, but given the limited amount of information, this is it for the moment :)

And the answer of Mr. anthony horner is exactly the other issue:

You can create alternative versions of processes in ARIS as well. It all depends on what you need to represent and why.

The rule for BPMN is usually to combine the roles to represent your scenario above. You want to have only one activity per lane. So, to overcome this, your model would look like this...

It is much neater.

Damian

Sorry - I have made a mistake in my above narration... Not One activity per lane but rather One role per activity. Of course you can have multiple activities in a lane!

Patrícia,

I fully support the solution suggested by Damian. Alternatively you may consider using Function Allocation Diagrams assigned to BPMN Tasks and there assign roles as appropriate. Then you can have a nice view on the resposibilities using Matrix model, RECI report and the other ways to generate a two dimentioanal view on asserted relations between two object types. The benefit of this approach is that you can specify the type of assignment of each role, something that BPMN alone does not provide.

I like Ivo's idea, because the lane object in the spec is not well specified. This creates a lot of confusion and most people miss the fact that you do not have to use lanes and pools to create a legal BPMN model (if you do not model a pool it is assumed that everything in that model happens in exactly one pool; lanes can be omitted without any implications).

In addition to this, you are using the "correct" object types and can avoid to create mappings between lanes and roles in a matrix model for example. This makes things a lot easier.

@ JEF

With my knowledge(so short)

About Your Questions:

1. Yes - the Start Events are different

2. All the End Events are the same -> Archive Done -  and done by a Role D

 That can be solved with a Join and the End in the proper Lane.

3. There's no escalation. Depending on the Lane (Role) some activities are Centralized  or Descentralized. Meaning some Activities (exactly with the sme definition) are done by diferent Roles. Somes Activities are always done in a BackOffice, depending on the Role.

4. Almost all Activities will have Subprocess and... where some are reusable.

And my comments.

- All the Splits depend on the Role. Meaning I would have to have a decision before each one. Because Splits are Logical (not decisions) The Split after the Start event  in Role C, doesn't feel right. I would design a Task before the Split.

- The Process for Role B only happens with Role A Start Event. That might not be true...

PP

@ IVO

I assigned  BPMN Allocation Diagrams 1.x to each Role in the Lane, and not FAD - although they're not the same in the end I think the result is the same.

So I have:  One Role depicts several Organization Units (therefore Locations).  With the relation Role x Organization Units (with different OrgUnits for each Role), I'm able to know where all the activities happen.

As you suggested, I'll try the aproach using BPMN Allocation Diagrams 1.x  for Tasks x Roles. Or should I use FAD?

 It's another solution: It reduces the number of Lanes, and I can get in the end those nices views you mentioned. The downsize - is the visual look when you need to publish Documentation for End Users.

Do you know some resources where I could get some Reports about the views you mentioned. I've one, but it only gives Lanes X Process.

Great conversation here, with the help of @Damian and @Ivo!

This could become a long post...

First of all I'd like to point you to this website, the home of BPMN:

http://www.bpmn.org/#tabs-docs

Here you will find full documentation and example scenarios on how to tackle certain issues.

There are some points to consider about how I purposely modeled some things wrong and some things in a certain way.

The ideal result business process model is that it provides clear documentation on how a starting event (state) leads to a final event (state). Preferably, the model is simple and elegant, removed from redundancy and (visual) clutter.

This means that you will need guidelines on what to model within one view.

My current background is that I use the database driven ARIS environment, where objects are reusable by nature. This also is where @Ivo is referring to, using the same processes in a Function Allocation Diagram and assign multiple roles to a process to make sure they can perform these tasks and show in a RACI diagram based on the relations established between role and process who primarily and secondarily are responsible or carrying out a process.

Based on what you are using BPMN modeling for in what tool (some people use ARIS Express to model a whole chain of processes within one model) you need to define where to draw the line:

Use the basic 5-level detail pyramid prescribed by IDS-Scheer (Software AG) with Level 4 (and 5?) EPC Diagramming exchanged for BPMN and adhere to the correct level.

Choose a free-flow detail level to model, but have the risk that it becomes unclear why you sometimes model something, and sometimes not. This approach would allow you to focus on multiple roles or scenarios; however, this is not best practice and might completely wrong.

You provided some hints for me in your answers:

1: As you have described, given two different start events leading to the same end event with the same process activities, it would be recommendable to split this into two models if this is done by one role. In ARIS express, you have to model all processes again, or use the same model and save it as a different instance.

2: This also provides a clear hint: there is an activity about archiving something. Let's dip into two basic points here: processes are aways activities: doing something. At the end of a process something is done, describing the state. 'Archive done' sounds like something is archived, which defines the state or end event, and would name the process before the end state 'archive something'. So the role D in your case would have this last process in their lane, and would have the end event in their lane as well, because role D is repsonsible for this state to be reached.

3: Your central and decentral description also requires some attention: regularly I advise clients to consider if these activities are actually the same: they are not done in the same place and could be literally thousands of miles apart. That the roles share the same system and do the same activities on a system does not mean that this is treated exactly the same, because the person performing the task has hidden attributes (being central or decentral) and this would best be represented with two different pools(!) making explicit that the decentral and central are different. This is really formal, so with business process owners / experts you could also make the decision to model only one version and name the process '(de)central process'.

4: Subprocesses are tricky, BPMN allows for mixture of detail levels, just as I've described earlier. I will address the issue with your first comment here as well. Where the pyramid approach gives you various detail levels, with level 4 being processes and level 5 having process steps or subprocesses in full detail, this is based on Event-Process-Chain modeling. EPC modeling also has its rules, one of them is regarding splits and joins. Whenever a event-process flow generates a split, all following connections joining in a process or event are lead by a join. The rule for this is that each event or process only has one ingoing and one outgoing relation (arrow) to allow correct simulation on script level. So if you would like to simulate your flow just like Markov chains and other network calculations, you need to model properly.

To come to your final comment, I'm not sure how to handle this, as you have described only a theoretical, incomplete scenario for my understanding. Therefore, I would think that the hints I've picked up have proven that although you can agree that the processes are intrinsicly equal, the model should be split into either two models or at least multiple pools defining the two main scenarios: central and decentral activities.

@Damian, how do you assign the lanes to multiple roles in your database? It is really clean and elegant if this type of combinations are possible for representation.

I've added an older example to illustrate how the differences are between modeling solutions.

There are some fundamental design connotations to consider, also for making business processes measurable for performance. Would you describe processes as generic as possible or do you want to be specific?

Are you measuring the problems solved, the problems solved in 1st line or 2nd line support? Only measure the # of escalations? With this type of questions, the organization also learns what is important to know and what not, i.e. make problems solved by channel explicit or not.

Therefore, you need to discuss and think about what you want to achieve with the models you create.

Also, regarding the modeling notation, you could model this with an exclusive OR gateway as well after "Assess Problem", instead of an escalation event. However, this case is a typical example of an escalation, so therefore modeled this way.

@Damian, how do you assign the lanes to multiple roles in your database? It is really clean and elegant if this type of combinations are possible for representation.

The crux in this scenario is the definition of a lane. Lanes create sub-partitions for the objects within a Pool, which can take on any desired classification e.g. Roles, Technology, Org departments, Location, Position, Org Structure, Application being used, etc.. Since BPMN is new to ARIS and they have not fully integrated it with other model types, for whatever reason, they have at least integrated it with the FAD (Function Allocation Diagram) as Ivo, above, has also suggested. This is very useful as it allows us to use the modelling richness of BPMN and combine it with the Architecture Object richness of EPC.

So how does this all work?

Take my BPMN model as before but with an update on the Lane names (so we can see the application of all this)...

I have replaced:

-Role A with Manager (Client Executive)

-Role A/B with Front Office Management (Client Executive + Client Administrator)

-Role A/B/C with Front Office Staff (Client Executive + Client Administrator + Client Executive)

Now Manager is a genuine Role and can also be a Position or Position Type, Employee Variable, in fact any classification to do with roles / actors / stakeholders/ etc. What is important is that the Lanes all use the same Object Type. Let's use Position Type in this case.

Now, getting to the above question of how I assign lanes to multiple roles?

The short answer is... I don't!!! Instead, I create a Function Allocation Diagram to do all the work for me i.e. build all the relationships I need to all the relevant objects.

So, lets take Activity One in the above diagram, assign a new Function Allocation Diagram to it, and start to populate the relevant Activity with all the relevant data you require...

So, now we can see that Activity One has a Position of Manager and a Role of Client Executive (I'm not going to discuss all the relationship types e.g. RACI - which can be done here). I could add other objects, which are not part of the BPMN specification e.g. Risk, KPI, Control, etc. and they would all be related to Activity One. If you had to now navigate back to the BPMN diagram above, select Activity One, and look at the relationships you would see...

...two new relationships that are not visible on the BPMN diagram. We do, however, have two instances of the Manager object but they are different types (better naming of objects would resolve this issue). The important thing is that object relationships are preserved across models and this is important for reporting on objects. You now go on and assign FAD's to each Activity and thus build a Process architecture structure that has both the modelling richness of BPMN and EPC.

The FAD for Activity Two and Four would look like something below...

Now the Lanes are assigned to multiple Roles via the common object - Activity. I admit it isn't as clean as I would like it i.e. I would have preferred to assign the lane object directly to an Organization Diagram or whatever the lanes are representing (Location - Geography breakdown, Application - Component breakdown, etc.) but we still get what we need from the models.

Damian

@Jeff

For now the main goal of the project is end users Documentation. So  I want to describe processes as generic as possible,  but  in a way that the most 'visible' bottlenecks are there.

This conversation is really interesting.. Based on something  you wrote , here's another approach. Centralized versus decentralized.

Here is a BPMN for Descentralized Process (Happy path?) Each Lane is assigned to a BPMN Allocation (BPMN-A)

Some events are for measure,

For the first Lane Descentralized Participants here's part of the BPMN: :

From here I can go to the Organizational Chart( with locations,  and so on):

This is something I cannot do with FAD - it's not available) . Lane versus FADs

With the Tool I have and BPMN 1.x:

- I can't use FADs to map Lanes to Real Organization Units. Only BPMN. A

- I'll use FAD  for Roles. Example for Company X: Person Type A in Organization Unit B, may have several Roles depending on the Process/Activity.

Finally:

1.  How I can I model a Loop (hope you can see it in the diagram)

2. I know that with your help I found the way to my my answer. Tomorrow I'll say something.

Comments please.

Tks

PP

Another solution:

Rob Davis - ARIS Design Platform: Advanced Process Modelling and Administration

"7.2.3 Multiple Relationships in Row/Column EPCs

By placing a resource object in the header cell we automatically create a relation-ship between that object and every Function in the column using the relationship defined for that column. However, we may also want to establish additional relationships between other resource objects and specific Functions. We can easily do this by placing the resource object in the main area of the column and specifically connecting it the Function. Because the object is not in the header cell, it will have no implicit relationship defined and hence it can be connected in just the same way as you would using a standard EPC.

[...]

So we can use the column to define relationships that apply to all the Functions in the column, and then specifically connect separate Organizational unit types or Application system types where an additional relationship is required for that particular Function. For example, look at the model shown in Fig. 7.4."