Tuesday, October 02, 2012

Manifesto for BPMN 3.0

Dear Readers,
   Having worked extensively for 6- 7 years on BPMN 2.0 specification, and its amazing success, I would love to see more work to address the requirements of more real life processes being added to BPMN. The first prototype of .bpmn file was written in second half of Dec 2004 on a late night, using little more than XML and a XML editor. Its great to see the success of the standard and it speaks volumes about the hard work done by the entire team. Steve big kudos to you!!! Any estimate how many person years were invested, my guess is about 200 person years at-least.

   Here is my set of recommendations for BPMN 3.0 -
1. Global Event support
An event, of critical importance,  can occur at anytime during the execution of a process instance. e.g. Pardon the extreme: Death of a loan applicant while the loan application is being processed.
It would be painful to model every task to be aware of such an event occurrence. Therefore, it should be possible to have a global event(s) defined which all activities in a given process are aware of and can react to if needed.

2. Make Process Artifacts first class.
Current support for notes, documents etc is restricted to association with activities. There are scenarios where several activities in a process may result in updating a document or based on some content of the document, might influence the decisions made during a process. Therefore, Documents should be treated as a first class citizen similar to data objects.

3. Refine Execution semantics in cases of Signal and Global events
Taking a cue form 1, we need to refine the execution semantics to support such events.

4. Allow for rework
Often times, process instances may need rework of certain steps. This is especially true for human tasks but is equally common for service tasks, the specification should specifically allow for rework of process instances from specific tasks.

5. Allow for Skip
As in 4, there are cases where certain activities may be skipped both in normal cycle as well as during rework cycle, so execution semantics need to provision for skipping certain tasks.

6. Allow partially defined processes
To support design by doing paradigm, and also allow lazy folks like me :), it is important to allow users to model processes which may not be complete.

7. Runtime Modification - Change the state of instance -
Execution of process needs to allow for change of state of any given instance, including go back and restart a task or skip some tasks in some instances, or activate certain branches - branches that were never activated given the conditions at that time but need activation now.

8. Runtime Definition Modification -
Change the definition of executing business process and instances thereof. A common use case, where processes can never be right the first time and often times there are inflight instances already when we discover the discrepancy. Allow the definition to be changed while instances are already in flight.

9. Digramming should support layering
There can be multiple level of detail in a BPMN diagram. One specific to Executive users, one to department heads, one to business and one that actually executes on the wire. Since BPMN supports diagram - there should be support for layers of diagram. I see the level of detail, I want to see. I know this can be left to tool vendors but this can probably be the non-normative part of the spec. "Non-normative" is the OMG speak for not mandatory and binding - but dont quote me on this definition.

10. Bubble up user participation in activities and consequent decision management.
Currently decision making is made based on condition expressions with data variables. However, process participants and their decisions are equally critical and there should be first class support for these decisions.

11. Pre and post execution rules support.
There should be pre and post conditions for execution of each task. Before we assign a task to a user, or even create a task, we should allow for user defined rules as an entry condition to the task; similarly before a task is considered to be done there can be exit criteria defined.

12. Add checklist for activities including ad-hoc
As a post condition for ad-hoc tasks - there should be an option to add a checklist, one which can be modified by users when needed.

13. Milestones - Add a notion of milestone.
Since several business processes can be long running and may even take months to finish, and also for supporting intermediate signpposts, we should add the concept of a milestone. Yes one can argue that even today this can be modeled as an association but it makes sense to make it first class member of the process definition.

Friday, September 07, 2012


Folks I see that there is a lot of confusion between my WYDIWYE and yours WYDIWYE.

From my experience, I have seen tons of BPMN models modeled in Visio / Signavio / ARIS and the user expects these models to execute on a BPMN 2.0 engine.  Alas, execution is when the real WYDIWYE challenge comes out. (There are tools that compile process models to other execution languages, for example to Java. )   It’s quite likely that unless you have severely limited your modeling richness, that your compiler will reject your process model! Does your tool truly support WYDIWYE?  Which is to say, does the product you are using support your business process pattern?  Or are you limited in the patterns you can choose to run your business?

Maybe we should call the feature in question “WYCDIWYE”, where the “C” stands for “can”, as in “What-You-Can-Draw-Is-What-You-Execute. Which workflow patterns does your tool support?

Regardless of political affiliation, where modeling freedom is concerned, Roubroo users can all quote Barack Obama - "Yes we can!"

Build ACM from BPMN 2.0?

My comments on Adam Deane's post:
Lets see what we can agree on?
I hope we all agree in value in process participants being able to change process even if they are “cheap butts…” as suggested by Max. I can fully understand why social BPM gets the attention it deserves.
Max you define BPM as model preserving execution while my vision is that you can still preserve the model (or a template of it) even if you make changes to it while execution is in progress. All ongoing executions have to be supported by an underlying definition to exist and also support audit. This vision aligns closely with Stephen White’s vision of what is coming next in BPM, found here http://www.column2.com/2012/09/bpm2012-stephen-white-keynote-on-bpmn/. And further you can make changes in “retrospect”, which is one step ahead even of design-by-doing, as highlighted by Bruce Silver in his review of Roubroo.
If you look at Alexander Simrin’s blog post here :http://improving-bpm-systems.blogspot.com/2010/12/illustrations-for-bpm-acm-case.html
you will see that its a continuum from structured to unstructured.
The focus of this conversation is use cases for adaptive case management not the degree of WYDIWYE, please take a look at IOR merge semantics before we can conclude either way. Obviously, the design tools prohibits users from drawing a process model which it can not execute – even though the vendor may cliam WYDIWYE. Here are some process patterns as food for thought : http://roubroo.com/resources/process-patterns
Back to the topic – Knowingly or unknowingly, we have all agreed that we are ALL using BPMN 2.0 as the modeling paradigm some with extensions and some without. If you are using something else, then I would like to hear.
Roubroo is using the BPMN 2.0 execution semantics to deliver the adaptive business processes. And this is where the big Question is – do we invent a new way for doing adaptive processes or build on top of the execution semantics which have become the de-facto standard.
Another question: what happens once the goal in ACM is achieved – does that sequence or sequences get repeated? If yes then its a flow, if not – then I am interested in understanding the ROI – but that is probably a separate topic.

Friday, August 31, 2012

Case for ACM using BPMN?

Reading through Adam  Deane's post and comments I realized its important to understand where the industry is headed without the three letter acronym jargon. And more importantly, how can folks leverage the work done to build existing specifications. I very much appreciate the real world examples Adam listed. I saw the commentators telling what ACM is not - I want to hear what it is. Thanks to Max for giving such a good list of what to expect from an ACM:
And my comments follow the "-"
A) business performers can create the processes to meet well defined goals.
- Well who can argue with that - and that is true for any process; the process owner should "OWN" the process. WYDIWYE is critical must have. Dont shoe horn the process to fit what the tool can do.

B) processes are signed off by the process owner and rated by customers.
- Absolutely - it means other folks that consume the process must agree with the definition of the process.

C) They can follow a flow but do not have to.
- When they dont follow a flow what do they do? My point is whether we like it or not, there is a flow.

D) They can be modified during execution where allowed.
- Absolutely. I am assuming the question here is whether you allow state changes or also allow definition changes.

E) Goal completion can have very strict criteria.
- It should have a measurable criteria.

F) Modified processes can be saved as new templates. (adapted)
- Absolutely - in CS corridors  the word is versions.

G) each process instance can be audited.
- Absolutely true. If an instance cannot be audited then we are generating code and reverse engineering audit info from code to tie back to the process initially created.

H) linked objectives and targets allow business perfomance monitoring.
- The key performance indicators should be linked with business performance.

Well a product can do all of these, it can use BPMN as a notation to track what is happening?  Would we call it BPMS, Dynamic BPM, Intelligent BPM or ACM ...? Check out Bruce Silver's review of Roubroo.

Thursday, August 23, 2012

Cloud demands a new way to run Business Processes

My invoices / purchase orders / quotes / leads are stuck!
Does any of these sound familiar? Have your ever had a situation when you are working with a cloud service and another application and you realize that your business is stuck somewhere in between because of an IT issue either on your IT or the cloud providers IT? How many times you have been interrupted because xForce upgraded their API or your ERP was undergoing maintenance or the DB went down over the weekend towards the end of the quarter – just when you needed it the most and you could not provide that crucial analysis or report.
If any of these scenarios strike a chord then you are looking at the new way of doing your IT business.  Where services will go down or will  do an upgrade with or without notice and you have to redo some steps while not redoing others – sounds familiar? I want to push all my quotes again to SalesForce, but I don’t want to submit all the invoices again into my SAP. Welcome to a world where when you define a business process, it may not work exactly as expected because of a myriad of reasons – some you can control some you cant. How do we get around these? We have seen the argument of on-premise vs on-cloud play out and currently the answer lies somewhere in between. And that is going to be the reality of our world for at-least the next decade.  How do we prepare ourselves for this reality and ensure that we are not sitting in conference calls between Friday and Sunday AM.
What you need is a solution that lets you make these various applications talk to each other and be prepared to hold the line when one of the parties is not reachable. Also it must allow for reconnection when the call drops – almost all of the things that you expect from … common sense!
I remember talking to a tire selling company last year when they had put a system in place which was holding a few hundred  of invoices in flight but there was no way to move them forward. Engineers on both sides spent hours trying to “fix it”, understand how we got there in the first place but it was a tedious task. Those invoices were critical for the sales team to hit their monthly target but there was little respite from the “system” issues.
When we started building out the next generation business process platform, we were acutely aware of these requirements. Our belief was confirmed when talking to our early adopter customers that they have this pain and it is a recurring pain. Therefore, we built a very agile business process engine which can react to such events as and when they occur. Earlier research around BPMN execution semantics came in handy for our design. And the result was Roubroo – a light weight, cloud based graph enabled business process platform which could execute structured and unstructured processes, accommodate ad-hoc process definition and instance state changes with a WYSIWYG control – much like the control panels in more evolved machines like Hydroelectric plants – more on this another time.

Monday, July 23, 2012

BPM 2012 Demos

Hey guys - A quick note, there is the most significant BPM academic event BPM 2012 is coming up soon. It is hosted at the University of Tartu this year and Prof Dumas is the General chair. Industry stalwarts Alistair and Stephen White are the other people that I have had the pleasure of working with at the BPMN 2.0 task force. The most interesting part is that there are a whole bunch of  industry relevant research papers. I have my favorites but I will hold on until the conference.  Here is the list of accepted papers. Roubroo wont be present in this years conference, however, we are plugged in to what is the latest and greatest in the BPM world; in terms of paper review and more.

Saturday, July 07, 2012

Crossing the chasm

What do businesses do? Or what makes the businesses successful.
The ability to deliver a  quality product (goods / services) efficiently  and consistently. The reason why Walmart is such a phenomenal success as a business is because it can deliver a specific product efficiently, hence at a cheaper price, than any other competitor. Success in business means being efficient and have economies of scale. And I can speak some more here but lets desist the temptation:
Why did I start with the above because a business is a business so that it can deliver "" repeatedly. We have annual IT budgets that run into billions so that businesses can execute their function - deliver a specific product to their customers again and again. The importance of the word repeatedly and again is where I am going with the reason why businesses need business processes. Whether they execute in a person's mind ( a SMB owner) or in a BPM system, there are business processes that repeat on a daily basis so that business can deliver the product or service custoemrs expect. Making these processes efficient and improve them rapidly is the goal of every CEO.

At the same time, there are professionals who have to deal with unique problems everyday - a doctor tending to a patient in ER, a lawyer dealing with a criminal case, a golf player on Pebble beach golf course with El Nino winds for a change...most of these examples are unique and being able to document each of these processes might :

1. Defeat the purpose (the patient might die) while the process is being documented.
2. Wont be cost effective - a doctor documenting the case (not patient history notes) and how he arrived at the differential diagnosis and then choose a specific antibiotic over another will not be cost effective as this decision making intellect has been honed over at least 10 years of med school and practice thereafter. It would be impractical to design for every permutation and combination.
3. It would not be repeated - I dont think the speed and direction of the wind at pebble beach is going to be the same again.
4. Each of these is unpredictable, you never know what to expect.

I think that the reality of businesses lies somewhere in the middle of these two. There may be certain processes which are so mundane that any investment would be a kill (and I wont give an example on purpose); there are certain processes critical and core strength of the organization, the procurement process for an automotive manufacturer, or order to cash come to mind immediately, which can benefit from a BPM approach. Even within these critical processes, 60-80 percent of instances follow a well defined path ,depending on maturity model of the organization, certain instances may need adaptive processes or segments thereof, or processlets as I would call them, and then again there are cases which happen in rare instances. 
As a CIO where should I put my money in - 
a) in those 60-80% process instances which can be automated, done faster better and cheaper; 
b) in those 20-15% instances where parts of the process are defined but not the whole process, 
c) or in those capabilities which are rare and in between - wherein even after the investment I might find that it was insufficient.
I think the choice is really between 1 and 2. But what if, <...pause>, there is a software that can do both. One that can execute my 60-80% process instances, support cheaper operational models for these, continuously measure performance of key tasks, improve  the process based on feed back ....
AND  with the turn of a switch, take those specific 20-15% instances and create an adaptive process model, modify them on the fly and execute segments of process as and when needed.
Time for my sales hat - please feel free to ignore
Fortunately, such a system exists and that is called Roubroo. It lets you model your business processes using BPMN 2.0 - executes them as per the semantics defined and alo lets you modify specific instances as and when needed. Please reach out to intro@roubroo.com for details.

Friday, June 08, 2012

The BPM pattern challenge?

Dear Readers
   I have been doing some deep thinking about the value proposition of BPM. The modeling environment, the execution engine, the admin console, the user tasks engine, the reporting and the list goes on...
In this laundry list of features, users may often overlook the very reason why they needed a BPM tool in the first place. The reason people decide to do BPM is to run their business - with accuracy, promptness and absolute fidelity; and they want the ability to change their business based on new market conditions "VERY" rapidly.
    If this is the reason why you are doing BPM - to run your business process with absolute fidelity - then dont be fooled by the laundry list of features - go for the kill. Can the BPM tool in Question run your business process with "ABSOLUTE FIDELITY" - can it "CHANGE RAPIDLY"?
   I argue that most tools cannot. And I am doing so with the experience and conviction that I have.
   And the reason we started building Roubroo was exactly the above - to be able to execute what your business process is with absolute fidelity, sometihng that we call "WYDIWYE" for the past couple of years. Sometimes, it is difficult to articulate the value of process fidelity and WYDIWYE - but probably it is easier to articulate the value of not having it. Ask the following Questions:
1. Does your tool let you model the process you want to draw? Or does it draw extra gateways etc as you model, trying to "help" the modeler?
2. Does your tool execute the process as is?
3. Do you have to call in your IT folks to code wait conditions or other checks to work around the limitations or as they say "features" of the said tool?
4. How rapidly are you able to add or remove steps?
5. How easily can you change the order and dependency of the steps - change the flow logic of your process?

I think if the answer to any one of these questions is No or the estimate of change is more than a matter of day (I wanted to estimate minutes but to be more benevolent, or couple of days) - chances are you are dealing with the wrong tool.

And the reason is not that someone made a mistake, its simply that these tools were built without the correct execution semantics. This basic and fundamental power is buried under layers of tooling and feature lists which is hard to argue around - every vendor claims they have all of these items in their feature list. What they dont tell you is what is at the core of the engine. What are the mathematical fundamentals around which they built the process engine?

I will leave today with a few examples and I will ask you friends to enlighten me if your BPM tool can build these patterns in a matter of minutes or not?

Take a look at these patterns - I have built only 4 for now, to see if the tool you have chosen can support your business process with "ABSOLUTE FIDELITY". And mind it these are the simplest of examples with no more than 4-5 steps required to demonstrate the BPMN 2.0 pattern. When we start modeling the real business processes there will be combinations of these patterns at play.


Sunday, April 01, 2012

Activiti - PVM - a closer look

Recently a team member pointed me to this white paper evaluation of PVM - the process virtual machine which is the heart of jBPM and Activiti. Apparently, the support for various patterns is pretty week in PVM:
Who is betting the future of their workflow on Activiti? Take a deep look at Roubroo

Maturing BPMS to Case management


This article is an excellent summary of what to look for in a BPMS that can allow for the flexibility of supporting case management. The ideal solution which kill two birds with one shot. It will create a set of defined process(s) for the 90 % of throughput but will also allow to turn this into a case for the remaining 10%

Learn to walk before you can run

On one of the linked in forums I saw a quote, learn to walk before you run in the BPM context and I was really happy to observe the power of simplicity. We are busy talking about and creating new acronyms, S-BPM, X-BPM, Y-BPM and what not. But have we and our practitioners learnt to walk? Or we are saying we will give you a car which can only be run by our trained chauffeur and the chauffeur will cost you more than the car.