Jim SinurThis is really a fun topic. It used to be that there were only programmer like representations of rules because business rules engines operated on the “IF THEN” structure (AKA Modus Ponents for you AI folks). Today with business rules engines and business rules management systems (BRMS) focusing in mostly on explicit rules for managing agility, there has been a migration to other forms of rule representation. There is a battle brewing as business rules become more important going forward.
![MCj04370760000[1]](http://blogs.gartner.com/jim_sinur/files/2010/01/WindowsLiveWriterRuleRepresentationThePotentialShowdown_A544MCj043707600001_2.png)
Rule Representation for Business Professionals
I find that most business types like more visual representations and there is a variety out there. Some business folks prefer decision tables because they are kind close to a spreadsheet metaphor. Others prefer a tree like structure that implies a hierarchical metaphor that is similar to analytical break down thinking. Other still yet like a graphical relationship view somewhat like process models. There is no clear winner here, but I see things tipping toward decision tables.
Rule Representation for IT Professionals
I find that IT type like precision and tend to like a text like representation that mirror programming languages and the deep AI guys that want to alter the sequence of rule firing really want the linguistic style so the same rule can be used in forward and backward chaining situations with support truth maintenance that generally uses more of a programmer/mathematician kind of approach.
The issue the use of different representations comes to a head when collaborating on initial rule creation and ongoing maintenance because of the divergent styles between the business types and the IT types. Just like folks settled on a common way share process models (BPMN) there needs to be some compromise here as well. There seem to be two streams. One is the rampant use of decision tables recently and the recent emergence of the semantics of business vocabulary and business rules (SBVR). Which one will win will depend on how easy it is to communicate back and forth easily. I’d say there is consensus about rule representation; only tendencies.
Category: BPM Business Process Improvement Business Rules Tags: BPM, Business Rules
16 responses so far ↓
1 uberVU - social comments January 20, 2010 at 5:38 pm
Social comments and analytics for this post…
This post was mentioned on Twitter by JimSinur: Business Rule Representation: Let’s Get Ready to Rumble http://bit.ly/8CoL1E #brms #bpm…
2 David Wright January 20, 2010 at 8:42 pm
What about straightforward statements of rules? Non-procedural, definitely.
“A person may place an order for pizza, only if they are 18 years of age or older.”
“A Customer can place an order on credit, only if their credit history is ‘good’ or better.”
“A Customer’s Credit is ‘good’ only if they have made 2 or less late payments in the previous 6 months.”
“Inventory is re-ordered when stock falls below the defined threshold for an inventory item.”
“State Sales Tax is calculated as the current state percentage plus any applicable county percentage.”
These are understandable by any audience, and fit with guidelines defined by the Business Rules Organization.
If there still has to be some choice made, business-favored representation has to come first; IT can transform that into any internal representation they need to implement rules by any means.
3 Mark Norton January 20, 2010 at 9:58 pm
Jim, your focus on decision presentation metaphors is really timely and very relevant. I have to disagree with decision tables though – decision tables cannot win this showdown because they lack dimensions. This is also a critical limitation of Excel – a rows and columns metaphor is just not rich enough in an XML world. With the richness of data structures that XML allows, rules metaphors have to follow suit.
4 Jim Sinur January 21, 2010 at 8:53 am
David,
I like your representation as it is linguistic in nature, but the technology to support linguisitics is pretty new and hasn’t hit the market in mass yet. Quite often it is complex and nested logic that defues simplicty. I wish it was that easy.
Jim
5 Jim Sinur January 21, 2010 at 8:58 am
Mark,
In order ro give dimension, you have to combine mutiple tables, so you get driven back into complexity. Good point.
Jim
6 James Taylor January 21, 2010 at 6:48 pm
Rule representation is an interesting topic for sure. You have vendors like Corticon, IDIOM and Innovations Software (Bosch) developing a single, rich metaphor to represent all the rules in a decision while others like FICO, IBM/ILOG, InRule, Oracle and SAP use multiple metaphors. While Corticon uses a decision table or rule sheet, IDIOM and Innovations have their own tree/flow combination. IBM/ILOG, FICO, InRule, SAP and Oracle all support decision tables as well as rule syntax and rule flow with IBM and FICO adding decision trees. Add in the support that InRule and IBM/ILOG offer for authoring in Microsoft Office and the range of choices is pretty overwhelming.
The bottom line is this: Do the metaphors make it possible for you to manage your rules and do they enable effective collaboration between IT and business people? If they do, they are candidates and if they don’t they are not.
Personally I don’t think any one representation is a silver bullet – a vendor that understands the problem and develops its metaphor(s) thoughtfully can develop what people need in a number of different ways. I don’t expect to see one way dominate any time soon.
JT
Disclosure: All these vendors were clients of mine in 2009
7 Tom Debevoise January 22, 2010 at 4:16 pm
One should remember that the decision table and the ‘rule-tree’ or decision tree are equivalent. In fact, in a future release of Visual Rules we will have a feature that converts a decision table into to ‘rule-tree’ (we call it a flow rule).
We have found that decision tables become unwieldy when they grow very large. Frequently there are additional conditional parameters that direct the assignment to results.
That said; no one has mentioned ‘loops’. We frequently must loop through records to compute a min/max/average for a vector of numbers. This is particularly true for risk score cards.
Business Rules and the BR Approach have principally evolved from three sources: Artificial Intelligence (Expert Systems and Data Mining), Data Modeling and Business Process Management. For better or worse, as with any technology, these early beginnings influence current practices and perceptions. Future developments in business rules will drop unnecessary technology and practices while preserving the benefits of the status quo.
With expert systems it is not enough to define a business attribute in the knowledge base— you must define and connect it in terms of the facts. On the surface, this sort of formalism sounds desirable yet, it creates more work … far in advance of executable applications
The legacy of Expert/System RETE-based business rules systems has fostered a very cumbersome implementation methodology. The reasons for these extreme symbolic derivations have eluded the practicioners.
We believe that a sequention execution approach can cast off the unnecessary complexities of the expert system it offers a more direct approach to the implementatio of business rules.
8 Mark Norton January 22, 2010 at 9:36 pm
Tom claims that “One should remember that the decision table and the ‘rule-tree’ or decision tree are equivalent. In fact, in a future release of Visual Rules we will have a feature that converts a decision table into to ‘rule-tree’ (we call it a flow rule).” I would be interested to know whether the converter will convert rule trees into decision tables, given that they are equivalent.
I agree that this is achievable for simple problems, but simple problems are rarely the subject of decisioning implementations – and anyway, for simple problems Java or C# are also perfectly adequate rule definition languages. On the other hand, a decision to underwrite even a simple class of insurance business will require many tens to hundreds of distinct decisions (assessments and/or calculations), each of which will be related to many others in a precisely defined way. This cannot be represented in decision tables; nor can it be reliably represented in natural language; and it is expensive to represent in a computer language. This is the class of business rule that I think Jim’s challenge is aimed at.
Similarly David’s simple rule statements appear to be presentable in natural English. But my own experience in translating narrative decision statements (usually written by IT professionals) into Idiom is that the narrative is ALWAYS idiomatic (by the way, that is where the brand name ‘Idiom’ came from – the creation of a rule definition ‘idiom’ is required for each bespoke problem domain) and is almost always ambiguous. This is because you cannot have idiomatic phrasing that does not risk ambiguity when read by outsiders (this is by definition – the meaning of the idiomatic terms are localized and cannot be interpreted without exact knowledge of the rules of the idiom). The antithesis of idiomatic phrasing is highly structured narrative that rigorously adheres to pre-defined terms – and this cannot be equated to ‘natural language’. It can however be equated to an IDIOMatic language.
9 Tweets that mention Rule Representation: The Potential Showdown -- Topsy.com January 24, 2010 at 6:22 am
[...] This post was mentioned on Twitter by James Taylor and Leonie Norton, Björn de Visser. Björn de Visser said: @jimsinur – Rule Representation: The Potential Showdown http://bit.ly/90U3rM >>Added Comment to Comment on the Comments! [...]
10 Carole-Ann’s Predictions for 2010 « Only Decision Management, Technically Speaking January 25, 2010 at 6:01 pm
[...] the on-going requirement for lean, and the vocal involvement of key industry analysts such as Jim Sinur from Gartner and Mike Gualtieri from Forrester, I expect that more attention will be given to the choices of [...]
11 Paul Haley January 25, 2010 at 10:57 pm
Which vendor(s) support the full logic of Mr. Wright’s statements without mapping them into multiple rules in their metaphor? I think the answer is “none” and that we should be more candid about the limits of each metaphor.
Too many of these metaphors combine the use and content of knowledge. And too many of these metaphors lack adequate expressiveness that leads their users into reengineering efforts once they fully understand their limitations. For example, two dimensional rules typically limit quantification and disjunction or nesting of connectives. This places burdens on users to map their natural logic into the metaphor and then to map a body of such natural logic into a procedure that exists only because of a lack of expressiveness. Too many rule metaphors force people to think in terms of antecedents and consequents or actions instead of requirements or truths. It’s not that the former don’t exist but that they do not always exist in knowledge as people naturally conceive it. And too many of these metaphors will result in vendor lock-in while limiting functionality. We have PRR and a dialect of RIF that is enough to cover almost all of rule vendor functionality but is far too weak to implement SBVR, OWL, or more than this most primitive subset of RIF, for example. The truth is that rule vendors generally do a poor job of managing knowledge independently of its implementation and actually require additional implementation in order to cope with expressive limitationns.
Now I’m not arguing here that natural language is the be all and end all or that higher-order defeasible logics are a panacea but that we should understand the limitations that we accept with metaphors and inform users of the limitations and risks they accept with each.
There are metaphors discussed here that don’t make much sense to me but I respect that others think they are great. But they each have limitations that I don’t see being discussed other than by Mr. Norton.
12 Declan Bannon January 27, 2010 at 8:12 am
Interesting article and as a follow up on Mark Norton’s point around the use of rules engines in the insurance industry he is spot on when he says that it requires many tens to hundreds of distinct decisions (assessments and/or calculations), each of which are related to many others in a precisely defined way and that these cannot be represented easily in decision tables. Here at Allfinanz we’ve worked with numerous clients worldwide and our experience has shown that they also cannot be easily solved by using natural language or by programming and indeed there have been some expensive lessons learned in trying to use this approach
Some software vendors for the life insurance industry have solved this problem by identifying those complex problems that are specific to the industry and breaking them down into repeatable subsets. By focusing on a smaller set of problems 3rd and 4th Generation Automated Underwriting Rules Designers (http://www.lifeandhealthinsurancenews.com/Exclusives/2009/12/Pages/The-Evolution-of-Automated-Underwriting-Software-Systems.aspx) allow business users create and maintain rules using intuitive, easy to understand GUI screens. There is no need to have any level of technical or programming knowledge and after minimal training, business users with simple word processing skills can develop and change complex underwriting rules without any IT involvement. However standard Business Rules Engines struggle with the complexity of rules and they become increasingly more difficult and costly to maintain and change as the business needs change over time because in reality they still quite technical in practice and therefore requires close collaboration between the business and IT resources. This is because Automated Underwriting solutions specialize in life insurance underwriting, while BREs are designed for general business environments rather than focusing on any particular vertical industry.
This last point is important and I think I will spend some further time thinking about a more formal response which will generate further debate.
13 Jim Sinur January 27, 2010 at 8:55 am
Declan,
The appraoch you mentioned assumes that all the rule types and rule patterns are known. This means someone can predesign a container that is stable..What about Unkown unkowns that may appear under unlikely business conditions? Just food for thought and debate:)
Jim
14 Declan Bannon January 28, 2010 at 7:54 am
Jim. Fair question. I think it’s not so much a matter of rule types and patterns but more the smaller parts – questions, outcomes etc that can be specified. If that is done properly then the rule types and patterns can be made up of them. We have taken this approach and so far and in our particular area of expertise, the life insurance underwriting process, we’ve found that the emphasis is on the multiple outcomes and paths through the rules and as they are being written by business experts i.e. underwriters there are no “ unknown rules or patterns “ that can’t be covered. Well certainly I can’t think of a scenario that we haven’t been able
to cope with so far.
15 Jim Sinur January 28, 2010 at 10:12 am
Declan,
I was thinking of things like ajustment of blood chem rules for unknown diseases like we had happen in the 80s or unthinkable fiancial conditions for financail underwriting of the payer (like hyper inflation). I do believe that the insurance industry does have some great planners though
Jim
16 Declan Bannon January 28, 2010 at 11:14 am
Ah ok Jim. The good thing about the 3rd and 4th generation tools is that they gives the business user the ability to react immediately to changing circumstances and to adjust the rules accordingly without IT iintervention