Practitioners within the IT industry have been misusing terms and concepts consistently to promote their eccentricities. These are either data-centric or process-centric. For instance, objected-oriented programming uses the concept of a class to represent an executable unit of code (through abstraction, encapsulation, inheritance, and polymorphism), which an application can create, use, and destroy for a computational purpose. Is class data or process in UML? It is both. A class encapsulates data and methods (a form of information hiding) through proper abstraction of real world entities and events. Since computers became available commercially in the 50’s, graphical modelling languages proliferated and underwent several re-incarnations through various symbolic mechanisms. This article addresses UML, BPMN, and BPMS from a business user perspective, at a cursory level.

Retrospective on Graphical (Visual) Modeling Languages

History repeats itself all the time. In the early 70’s, structured programming evolved into structured design, and finally structured analysis. Object-oriented programming began in the early 80’s, which led to object-oriented design, and then object-oriented analysis.

The first de jure standard graphical modelling language was flowchart (uni-level model), certified by ANSI. It was the mechanism for the scope definition of a problem domain, and the use of logical steps to arrive at the solution through a computer program. When a problem would become more complex, it was extremely difficult to verify the correctness of a solution. Nothing is as constant as change. A change in requirements necessitated a monumental effort in the update of the flowchart and verification of its correctness.

Dissatisfied with the inflexibility of flowchart, the US Air Force led the effort to deliver a series of IDEF (multi-level model) modelling languages to support its ICAM (Integrated Computer-Aided Manufacturing) undertakings. This took place alongside commercial efforts (e.g. DFD, Structured Chart, MERISE, etc.) in the mid 70’s and early 80’s. As the DoD (Department of Defense) was an influential standard driving organization, IDEF was the only de jure standard, certified by NIST, to replace flowchart.

When RDBMS (Relational Data Base Management System) became more affordable and efficient, Information Engineering led the effort in promoting a data-driven approach to software design and development. The process-driven approach fell by the wayside. As systems became more software intensive, the US Air Force commissioned SEI (Software Engineering Institute) of CMU (Carnegie Mellon University) to conduct research on software management, engineering, and deployment in 1992. To reduce cost, risk, and time in delivering systems, COTS (Commercial Off The Shelf) software & hardware were augmented by systems engineering, concurrent engineering (i.e. Integrated Product and Process Team), and quality management in CMMI. Thus CMMI (Capability Maturity Model Integration) became the yardstick to determine and measure the capabilities and performances of an organization. The erospace and Defense industries were also contributors to CMMI, in addition to the DoD. When ISO released ISO-9000:2000 (for Quality Management Certification), the focus was process-centric, with emphasis on continuous improvement rather than just documentation. CMMI and ISO standards reinforce each other; the process-driven approach came back in force. Business Process Re-engineering (BPR) was also a driving force to bring process back into the picture for organizational improvement.

When imaging technology, including fax, could produce electronic documents with decent quality, many organizations began to deploy Workflow Engines. These not only saved physical storage space, but also reduced time significantly in routing documents within a premise or across the ocean. The Workflow Engine was never designed and developed for BPM in mind. With the release of the Internet to the general public in the mid 90’s, electronic document sharing and routing became a de facto mode of operations for many organizations and individuals.

Recognizing the benefits of the object-oriented approach in the early 90’s, programming languages were objectified, including the old work horses FORTRAN and COBOL. Proliferation of ways in applying the object-oriented approach began to get out of hand. In the mid 90’s, Rational led the effort to consolidate multiple object-oriented approaches, and UML became a de facto standard, primarily geared toward software development efforts. After the OMG took over the effort to develop UML, ISO adopted it as ISO-19051 in 2005 (UML 1.4). It should be noted that ISO and NIST have applied UML in their joint standard efforts in many areas before ISO formally turned it into a self standing standard.

BPMI.org was formed in 2000 to promote standardization in the management of business processes. The concept of process management was not new though. It began in 1961, when IBM released GPSS (General Purpose Systems Simulator). The prime users of the GPSS were Operations Research and Management Science practitioners. Before coding commands to run GPSS, a user would draw a graphical model to depict entities, events, activities, storage spaces, staging areas, queuing disciplines, etc. to represent a system or an operation, such as gas stations, checking out counters, traffic control systems, and traffic flows in a network. Some similarity exists between BPMN and the graphic models in GPSS and IDEF3. BPEL was not a new concept either. One can find many equivalent languages in real-time devices such as real-time process controllers, cell phones, PCs (the ROM-BIOS), and many field-programmable firmware devices. BAM may seem a novel idea, but SCADA (Supervisory Control And Data Acquisition) systems have been in existence for more than 20 years for real-time process control applications. The only difference is that BPMN and BPMS may elevate the concept of real-time process control to a more visible level in the management of day-to-day business activities.

UML (Unified Modeling Language)

UML underwent several versions from 1.0, 1.1, 1.2 (never released by the OMG), 1.3, and 1.4 to 1.5. The latest version is 2.0, which includes more elements to represent different views and allows extensions that incorporate other graphical notations into a model diagram. With ISO, NIST, and CMMI as the prime driving forces for standardization, many professional disciplines have joined forces with the OMG to enhance UML and use it to extend professional standards. Examples are:

1. ISO-191xx in GIS (Geographic Information System) standards.
2. ISO-10303 in STEP/EXPRESS for Product Data Description and Exchange.
3. SysML by INCOSE (International Council on Systems Engineering).

As a result, UML facilitates the consolidation of modelling languages into one, but also introduces potential complexity and inconsistency. For instance, the class diagram in UML consists of 3 compartments; one for class name, one for attributes (data), and one for methods (operations). In SysML though, block replaces class in UML with 4 compartments. The top 3 compartments are identical to a UML class; the additional compartment is for constraints.

In an organization that has both IT and Systems Engineering practitioners, business users need to deal with the eccentricities carried by these 2 professional disciplines. That is, UML from IT and SysML from Systems Engineering. There is a good chance that both disciplines develop very different models for a system, and that can be extremely difficult — if not totally impossible — to reconcile. Anomalies in the usage of UML still persist. Examples are:

1. For IT, users (actors) are outside the boundary of a system. This contradicts the well established definition of a system. That is, users are elements of a system.

2. The activity diagram in UML is a modified version of flowchart, which was deemed inappropriate for communication purposes between business users and system practitioners (i.e. IT or Systems Engineering) throughout the 60’s and 70’s.

3. UML facilitates communications within a discipline but not across disciplines.

4. Class diagrams are used for both objects and data, which are essentially very different entities. An object is an executable piece of software code, while data is for the storage and retrieval of encoded characteristics of an entity or event.

With many professional disciplines coming on board to help evolve UML, it is hoped that UML will remain a horse rather than turning into a camel. As the old saying goes, a camel is a horse designed by a committee!

BPMN (Business Process Modeling Notation)

BPMN was created as the visual (graphical) front-end for BPEL (Business Process Execution Language). In a way, we can deem it as a potential replacement for IDEF0 (Activity Modelling) and IDEF3 (Process Modelling for Simulation), and other simulation modelling languages (e.g. GPSS). Apparently, the designers of this language approached it from a pure IT point of view. They abandoned the age-old practice of a process consisting of activities, and activity consisting of tasks. In BPMN 1.0, we have process, sub-process, and task. Activity was dropped for no good reason. BPMN consists of “activity on node” constructs in a network-like structure of events and flows. It also supports the construct of layering or decomposition. In the absence of activity, business users would have a hard time correlating and reconciling the following:

1. How does BPMN map to Activity Based Costing/Management/Budgeting for performance measurements that percolate up in the hierarchy of Balanced Scorecards?

2. How does BPMN correlate with BAM (Business Activity Monitoring) utility modules within a BPM 2.0 BPMS?

3. How does BPMN map to the simulation capabilities offered by a BPM 2.0 BPMS? And that is assuming that there is enough functionality to collect actual performance data, determine best probability distribution fit from historical data, and select the best stochastic model (e.g. queuing, Markov-chain, chance constrained programming, etc.) for optimizing business processes on the fly! If this is at all feasible, will the BPMN diagram be updated automatically to reflect the optimized processes, as opposed to the original process pattern?

BPMS (Business Process Management System) in BPM 2.0

There are a number of functional elements incorporated into the BPMS of BPM 2.0 that support:

1. Application development without programmers.
2. Rule definition to govern process execution.
3. Role assignments to human resources who own the process.
4. Business activity monitoring in real-time.
5. Simulation for potential continuous improvement.
6. Optimization of processes on the fly.
7. EDI in X.12 and EDIFACT.

These capabilities are indeed very impressive. A number of questions remain though:

1. Does the rule engine becomes part of the ontology, and does it support the use of fuzzy logic in addition to binary logic?

2. Business process management is more than managing transaction-oriented activities (i.e. supply chain and demand chain). There are a lot of other supporting activities that are carried out within an enterprise, as articulated by Professor Michael Porter in his seminal work on Strategic Management (i.e. Competitive Strategy and Competitive Advantage) in the mid 80’s. Can BPMN and BPMS support the management of activities in “firm infrastructure” and “technology development” as depicted in the Value Chain model?

3. If optimization can really be realized on the fly, is there a need for simulation? By the same token, BPM 2.0 should be able to replace 6-sigma quality control programs in every enterprise. When the adaptive process optimization runs long enough, one person can manage a large enterprise without any help from anyone. Wouldn’t this be a revolutionary enterprise management breakthrough?

Harmonization of UML, BPMN, and BPMS

In the 50’s, MIT and Stanford were the first institutions in the US involved with AI (Artificial Intelligence) research. A number of scientific results were commercialized. Examples can be found in robotics (e.g. manufacturing and logistics), vision or image processing (e.g. OCR, CAT Scan, and MRI), rule-based expert systems (e.g. symbolic integration in mathematics, creation of industrial chemicals, and LSI circuitry designs), and semantic network. Semantic network was the foundation for Chen’s E/R modelling, as well as frame-based ontology. In spite of tremendous efforts put into AI research, natural language processing and cognitive science still need ground breaking advancements that can truly lead to programmable decision making with partial information.

Now that the work on BPMN has been folded into the UML working group at the OMG, it is hoped that UML will evolve into a more robust and simplified modeling language. Similarly, the future BPMS will incorporate more capabilities that can truly support cyclical activities in planning, programming, budgeting, execution, real-time process control, knowledge management, and decision making. In essence, a good foundation has been laid for better enterprise management through UML and BPMS.

About the Author: Francis Ip is a pragmatic Systems Engineering practitioner with over 35 years of combined experience in Strategic Management, Enterprise Architecture, and Program/Project Management. He is a contributing writer on IT|Redux.