I'll preface this entry by stating that I'm not a "Business Rules" expert by any stretch. I write this as one who has experienced an implementation from the trenches and who is about to dive into the deep once again at my latest account.
I arrived on a project that needed an injection of technical leadership. The project was basically a complex settlement system for a large energy market. There were hundreds of steps required to create a bill for a customer with numerous regulatory considerations. The complexity of each step could range from a simple atomic calculation to an extremely complex set of "business rules" with 10 to 15 pages of single spaced specifications. In general it was determined by the client that a business rules engine should be used to construct these steps. The goal would be to allow the business users to maintain the rules for the new system. I a'm not going to explain business rules basics, but going to assume youÂ've read the basic marketing available. This article contains the implementation experience that I pass on to you.
Business People Should Write Business Rules?
The ability for a business person to directly control the business rules, this is the Holy Grail when implementing business rules engines. The thinking here is that a visual tool can allow business users to simply "draw" a business flow. Additionally, by bypassing the Business Analyst, developer and associated typical software development life cycle (SDLC) that rules can be constructed more quickly and managed more easily. It is true that Business Rule Management Systems (BRMS) can accelerate and streamline the implementation of business rules, but buyers should understand some of the realities of adopting the technology and adjust their expectations accordingly.
It has been my experience that even the most advanced rules engine with the slickest interface requires training and, in some respects, structured thinking about how to organize and construct the rules. There are two factors at work here. First, visual rule environments provided by Business Rules Engines are not "Visio". There are structural constraints enforced during the definition of a valid visual rule flow. How long before a business user becomes frustrated because they just can't draw the "picture" they can clearly see in their minds eye.
The second issue with business users and direct manipulation of rules is rule decomposition. Rules engines that implement a Rete algorithm
allow large numbers of simple rule statements to be automatically organized into a network of nodes that allow efficient execution of highly interdependent rules. Business users typically do not think in terms of rules, they think in terms of business policies.Merriam-Webster
defines policy as
"a high-level overall plan embracing the general goals and acceptable procedures"
So in essence policies are much broader than individual rules. Someone must decompose the various business policies into discrete rules and structure them appropriately in order to represent these policies in a form that can be executed by the engine. Although rules can be structured such that they can be easily read and understood by business users, rules engines are not "Natural Language" processors. They require strict syntax in order to operate. An example would be the Yasu QuickRules engine. The user interface allows rules to be entered either visually or as text that adhere to a VB like syntax. Rules that do not conform to the syntax can not be saved and therefore never executed by the engine. The product gives relatively vague error messages, so good knowledge of the syntax and operation of the engine are typically required.
So if the business person is not the most appropriate person to implement
business rules, who is? We could naturally gravitate towards the programmer as an option, but my experience is that a typical programmer does not easily become an effective business rules designer. There is the natural tendency for a programmer to try and use the engine's rule language like another general purpose programming language. This usually ends in frustration because general purpose "programming" is not the goal of a rules engine and therefore can seem a poor substitute for languages targeted for the general purpose space like Java or C#.
Business rules in general should be recognized as a separate disciplinep. Special roles such as rules analyst, designer and administrator should be created. Talent for these roles can be sourced from disciplines like business analysis, software development and operations, but there must be the appropriate investment in training, mentoring and time to allow existing employees to grow into these roles. Think back to the investment required to move functional programmers to a more object oriented methodology. It took significant investment and effort for a successful transition. You can expect your results to match your investment.
Business Rules Require Facts (Data)
Business rules operate on facts.Facts are essentially the data provided to the business rules for execution and also derived during their operation. Facts are compiled, or loaded, by the calling application and passed into the rules engine for evaluation. The Fact Model is the entire set of facts that has been exposed for use to the rules engine. This model is typically constructed incrementally over time as facts are added to support new or modified business rules. The complete fact model can be developed in advance of implementing any rules, but in practice this can be very time consuming.
There are two issues that may arise in the management and transfer of facts (data) to the business rules engine. The first is version control. New facts required because of rule changes result in associated changes to the calling application. Additionally this requires that the modified application be deployed along with the new business rules. Something that we would hope to minimize as part of a business rules strategy. So in this case, the SDLC overhead can not be avoided and synchronized deployment of business rules and application software must be managed appropriately.
The second issue arises from the overhead of moving data in and out of the rules engine. Depending on the amount of data and the location of the engine (remote?), this could be a relatively costly endeavor. There are a number of engines that can execute within the same memory space as the calling application and therefore reduce the overhead of moving data between the two. But again, depending on the amount of data required it may be questionable as to whether data, not explicitly required for rule execution, should be loaded at all. Some engines have the ability to read and write application databases directly which can simplify the data movement process, but this mixes IO overhead with rule execution. This feature also removes the ability to use optimization techniques like write-through share memory caching in applications.
Business Rules Portability?
With industry support for standards such as the Java Rule Engine API or Java Specification Requests (JSR) 94, applications can be developed that can plug in different rules engines. JSR-94 standardizes the way rules engines are called, but does not standardize how rules are written. We can plug a new rules engine into an application quickly, but moving the rules between engines can be problematic. There are initiatives out there like RuleML that attempt to facilitate interoperability among vendors, but none have reach critical mass. Do not count on JSR-94 to ease replacement of a rules engine significantly.
When constructing a suite of business rules, it is important to be able to verify that we achieve the desired results. Testing all rules in isolation will not necessarily give a definitive answer as to expected outcomes. Dependencies between rules may have unexpected effects. Yes we can do "what if" scenario testing, but the real work is in defining a comprehensive set of test cases that exercise the dark corners of the rule network. This was our most difficult part of the project, generating the volume of scenarios and associated data to ensure coverage across all the business rules.