In-Depth
Business rules are back
- By Colleen Frye
- July 1, 2002
Today's business rules engines (BREs) and Mark Twain have something in
common: News of their death has been greatly exaggerated. Or as research firm
Gartner puts it, business rules technology is back from a ''near-death
experience.''
Business rules technology standardizes and automates much of an
organization's business and decision-making processes, removing to some extent
the human element that is prone to inconsistency and error. Business rules
determine what action should be taken based on a company's internal guidelines,
business practices and/or regulatory compliance. For example, business rules
might determine a loan applicant's credit eligibility.
''Think of [a BRE in the same way] as database management software manages
data -- so a rules engine manages business rules across applications. It's an
enabling technology to build flexibility into an application, and to manage
business rules separately and apart from that application,'' said Colleen
McClintock, Ilog JRules product manager at Ilog, Mountain View, Calif.
But early manifestations of business rules technology -- several products
have their roots in the realm of artificial intelligence (AI) and inference
systems -- were complex, expensive to run and maintain, and not
business-user-friendly, thus their ''near-death experience,'' noted Gartner's Jim
Sinur, vice president and research area director.
The time is now ripe for the current generation of BREs and rules technology,
observers say, due to the confluence of several factors. First, empowerment of
the business analyst ''is a trend we've been seeing over the last five years,''
said Peter Urban, senior analyst at AMR Research in Boston. ''BREs that empower
the business analyst represent a growth market. People are looking for this kind
of stuff; they don't want to go to IT every time they want to add logic to
applications.''
Gartner's Sinur likens the acceptance of business-user-friendly BREs to the
adoption of reporting tools targeted at business users, who can now run their
own reports rather than going through IT. ''I think the businesspeople are
ready,'' he said. ''I think IT folks will resist because it's also threatening.
But just like letting business users do their own reporting, BREs let them
change the rules.''
Another factor driving acceptance of BREs, according to Sinur, is improved
technology, including enhanced usability, scalability and performance, as well
as less costly maintenance. AMR Research's Urban agrees: ''It's taken a while for
the software to become usable, with graphical user interfaces rather than
programmatical [ones].''
''Expert systems were a lot harder to work with,'' added Mark Allen, president
and CEO at Corticon Technologies Inc., San Mateo, Calif. ''They required
programmers and knowledge engineers to analyze the rules and their
interdependencies. The other area was integration; [expert systems] weren't
friendly to integrate with databases and existing systems. Usability,
auto-analysis and integration are areas we've targeted with Corticon Decision
Management Software.''
As businesses move their applications to the Web, they are looking for
products that comply with the latest standards and technologies, such as Java,
XML and Web services, and vendors are delivering. In addition, the general
business climate -- do more with less, faster -- is helping to drive acceptance
of BREs. By enabling business users to respond more quickly to change,
organizations will have greater agility and improve their time to market, and
thus their competitive advantage, said Gartner's Sinur.
Ken Molay, director of product evangelism at San Diego-based HNC Software
Inc., the developer of Blaze Advisor Java-based rules management software, said
business rules technology provides a value-add for organizations on three
fronts. ''The first [front] is time to market -- saving time on getting whatever
it is into production and out to market, both from initial development and later
for making changes. The second is improving consistency across the organization
by making the same decisions the same way, with one centralized set of rules
that apply to your applications. The third is saving money on the development
work needed by being able to allow business-level users to make changes. If we
can get those changes off the plate of the IT group, they can concentrate on
mission-critical tasks.'' (HNC and Fair, Isaac and Company Inc., a developer of
predictive modeling, decision analysis, intelligence management and decision
engine systems, recently agreed to merge under the Fair Isaac name.)
Indeed, for CIOs like Ed Leveille, putting more application development
responsibility in the hands of business users is a goal. ''The concept I have in
developing applications is trying the best I can to get coders out of the mix,''
said Leveille, vice president and CIO at Providence Washington Insurance Co.,
Providence, R.I. In 2000, Leveille was charged with developing an Internet-based
commercial lines quoting system to be used by internal employees and agents. ''I
was looking for user-friendly tools that would have a nice GUI where the users
could do some of the business logic,'' he said.
Leveille decided to go with Ilog JRules. ''Any product I pick has to have most
of the current standards and an open architecture,'' he said. ''And the company
has to prove themselves as being sustainable; Ilog fit all of those
[requirements]. Another benefit of Ilog was the XML-based messaging, which fits
within my strategy of communication and integration of all my components,'' he
added.
Providence Washington Insurance has been piloting the use of Ilog in two
lines of business over the last 10 months. ''We've virtually automated the
underwriting guidelines manual, taking out the need for someone to look up code
or rules in this thick manual; the system guides users through the
classification and pricing of a given risk,'' said Leveille.
''We're also using the Ilog engine to score our risk and, based on that score,
to adjust the premium up or down,'' he added. ''And we use it to automatically
slot a piece of business into the proper special program and the correct pricing
tier. For example, if you have good loss experience, you might go into an A+
plan and get our best price.''
Opening the applications up to insurance agents was the driver for automating
previously manual business processes, added Leveille. And because they were
targeting agents, usability was key. ''Opening up our system to 5,000 or more
people, the training had to be almost zero, which is another benefit of the
[Ilog] rules package,'' he said.
According to Leveille, the JRules engine, rather than being embedded into the
application, ''is called as a service by our application. At various points, the
application sends it a message asking for some kind of decision; Ilog [then]
sends the message back giving us the answer. Our information is in a SQL
database, and the Ilog information and objects are stored in their database. So
there are two content stores: mine is the insurance information, and the content
Ilog has would be questions and answers to whatever the business rule is.''
The benefits of using a BRE at the insurance company are twofold, said
Leveille: ''The quality of our data and the quality of our underwriting will be
greatly enhanced,'' he noted.
Choosing a BRE
Companies like Providence Washington
Insurance have various options to weigh when choosing business rules technology
today. Packaged ERP applications have some embedded business rules technology.
And enterprise application integration (EAI) solutions may also offer some
business rules capability. ''Rules in ERP packages cover basic rules, but if you
have a complex financial scenario, for instance, you would need to extend
rules,'' said AMR Research's Urban.
BREs are ''more sophisticated and you have the ability to manage rules across
multiple stovepipe applications,'' explained Gartner's Sinur. ''Business rules
affect more than one system, so it would be nice to have a centralized rules
engine or rules managed by a single source,'' particularly for those developing
homegrown applications or integration pieces, or for those extending or
integrating legacy systems.
In addition to consolidating rules in one spot, some BREs auto-analyze rules
to ensure that they are logical and not contradictory. ''Corticon is a good
example,'' said AMR Research's Urban. ''It prevents you from doing contradictory
rules, and it makes you process things all the same way, which is a plus.''
Another distinction among rules engines is what Gartner's Sinur dubs
''outboard'' rules vs. inference capability. Sinur said outboard engines have a
simplistic forward-chaining approach, and are geared to business users who make
changes in English-language or decision-table fashion, in near real-time or
real-time. Inferencing capability means the engine can do both forward- and
backward-chaining, he said, as well as use deductive capabilities. Many
inferencing engines have outboard rules capability, Sinur explained, but not all
outboard rules engines have inferencing capability.
According to Gartner, rules engines vendors with inferencing capability
include Computer Associates International Inc., Elity Systems Inc., Entreon
Corp., Expert Solutions International (ESI), The Haley Enterprise, HNC Software,
Ilog, MindBox, Pegasystems and Sybase. Many BRE vendors, both with and without
inferencing capabilities, offer the BRE as one piece of a business process
management solution.
Business rules creation, automation and management have a natural symmetry
with other types of software solutions, particularly ERP, EAI and workflow
software. Gartner predicts that BRE solutions providers will partner with
vertical business sectors and EAI vendors. By 2006, projects Gartner, 80% of all
integration software will include either direct or indirect business-friendly
rules capability.
Some partnering has already taken place (see associated
story, AMS turns to
Versata
). Santa Clara, Calif.-based Savvion, for example, sells BizPulse -- a
rules engine based on Java J2EE and XML technologies -- directly to corporate IT
and OEMs it to independent software vendors like i2 and BroadVision, according
to M.A. Ketabchi, the company's CEO, president and chairman. Ilog also partners
with i2 Technologies Inc., Dallas, and enterprise software vendor HNC Software
acquired Blaze Advisor with its acquisition of Blaze Software.
BREs may not be a discrete market in the future, but they are meeting
critical business needs today, and the market is positioned for growth.
Gartner's Sinur said the BRE market is at about $300 million, but he expects to
see ''at least a 15% compound annual growth rate over the next three years.''
BREs are particularly appropriate for companies with complex business
processes and rules, companies that want to improve customer service, or for
those that want to ''differentiate vs. the competition. Anybody can do a loan,
but [a BRE] product allows companies to differentiate. You might create a unique
type of loan,'' using a business rules solution, said AMR Research's Urban.
For STW Fixed Income Management, Santa Barbara, Calif., which manages
fixed-income investments for its clients, the rules are quite complex. The
company must not only comply with each individual client's investment
guidelines, but with regulatory guidelines as well, according to Tony Plasil,
principal and CIO.
A problem in the industry, said Plasil, is that compliance checking, which
may take days, is typically done independently of trading activity. In the
meantime, he said, the market may change or some non-compliance identified may
force the trader to back out of the trade. In that case, the fixed-income
management company would assume any loss, not the client.
To address that issue at STW, Plasil said they began looking at rules
engines. They evaluated some packaged trading applications with rules
capability, but Plasil said they were too simplistic and not flexible enough for
the way STW wanted to define and maintain rules.
Plasil decided to go with the Corticon Decision Management Software. The
trading application has been in production for about six months. ''It took about
a year in development, with extensive testing,'' said Plasil. ''The implementation
is using Java EJB technology sitting on WebLogic. We're on the frontier of
everything; it's riskier but satisfying,'' he added.
Both technical and business advantages are already apparent, he added.
''There's a huge flexibility advantage; we can add or change rules fast, so we
have reduced costs by not changing code.'' And on the business side, ''one of our
clients' biggest concerns is how to maintain their guidelines. Now the portfolio
managers have that information at their fingertips. Client service is the bottom
line.''
While STW can point to improved customer service as a benefit of using a BRE,
the Westfield Group, a leading regional insurer located in Westfield Center,
Ohio, has already reaped the benefits of increased business without increasing
staff.
When the Westfield Group set out to develop a new policy processing system,
dubbed WesCom, in 1999, the IT team began building rules in Computer Associates'
Aion inference engine, which was mainframe-based. While the project was
underway, the company made the decision to move to a Java-centric enterprise
architecture. ''Aion didn't work well in a Java environment,'' said Paul Armborst,
application analyst. ''You couldn't see the data in Java classes; we would've had
to ship the data in and out in XML, and we were concerned about performance.''
(Now called CleverPath Aion Business Rules Expert, CA's BRE today embeds
business rules in Web-based applications by enabling an Aion knowledge base to
be called from a Java client.)
A consultant working with the Westfield Group recommended HNC Software's
Blaze Advisor, which fit the company's requirements; the project was then moved
to Blaze. The Westfield Group started with the Business Owners Policy (BOP)
line, creating edits and eligibility rules.
''Edits check for completeness in the data and any inconsistencies in data
entered. Eligibility checks to see if the risk proposed is eligible,'' said
Armborst. ''We execute database lookups with JDBC calls. The enterprise
architecture includes a reference data manager; it's a service-based
architecture. The security service is LDAP. The workflow service is iFlow from
Fujitsu, although workflow within the rules is handled using the rulesflow
product provided by Blaze.''
The new policy processing system enables the outside agents to handle most of
the insurance processing on their own, without requiring Westfield Group staff
to get involved. ''Before, we had the edits in the legacy system on the
mainframe, but no agent could interact with it directly. The underwriting rating
was automated somewhat for BOP, but not for other policies,'' said Armborst.
''We've gone from a situation where it took two weeks to issue a policy to
overnight. Agents are running 65% to 95% of the quotes, depending on the state.
Previously, all quote activity would have involved Westfield Group staff. Issued
policy volume has also increased, although it differs greatly by state.''
While volume has increased, the Westfield Group has not had to add any staff.
''The point was not to get rid of staff, but to grow our volume without adding
staff,'' said Armborst.
Because their business rules are complicated, neither the Westfield Group nor
STW Fixed Income Management is looking to business users to do the heavy lifting
yet. However, both companies say the creation and management of rules is much
easier than traditional programming.
Because of the complexity of STW's rules, ''deciding on all the data needed to
satisfy rules is beyond the user's ability,'' said the firm's Plasil. ''We have
rules defined in different levels. First is all the data required, then any
summarization, which requires a bunch of aggregations and is too complicated for
the user. The first levels are like the infrastructure, then the rules
themselves the users can define. Most business rules aren't complicated; all you
need to do is understand business -- the data itself -- and Corticon Studio
checks for ambiguities.''
At the Westfield Group, ''we are having developers create and change rules;
heavy-duty rules are more than your average end user wants to get involved
with,'' said Armborst. ''Even if they could write the rule, will they understand
how to test it? The specs were defined by the end users, but the coding is done
by IT people. There is a development environment with Blaze that you can create
rules in, which is point-and-click, but it needs to mature more; it's a little
slow and has quirks, so a lot of us do our coding with a different editor rather
than the development environment. You write in a structured rules language for
Advisor.''
Blaze Advisor does make it easier to keep up with changing business rules,
which is a plus, said Armborst. ''It's not only easier than in a traditional
environment to change rules, but it's easier to write new rules. The value of
the inference engine is that it matches rules up with data it wants to process
against it. If you write the rule and don't make it dependent on another rule,
you can write it and put it in with minimum testing and know you haven't
impacted anything else. This is different from a typical IT environment, in
which usually the slightest change causes problems.''
Web services
While Armborst and others are getting their
arms around BRE technology, the BRE vendors are positioning themselves for a
role in the emerging Web services arena, and industry observers agree that makes
sense. ''Web services is all about rules and flows, which is why these engines
will come in handy,'' said Gartner's Sinur. ''Pegasystems, for example, has both
rules and workflow, which is a good combination; Versata [Inc., Oakland,
Calif.,] has a workflow engine and a rules engine.''
AMR Research's Urban said AMR recommended a Web services approach to MindBox,
Greenbrae, Calif. ''You build a powerful engine, you might want to sell that
service; say it's an engine for loans. You could create a Web service that would
allow people to enter information and send that information in an XML message to
the engine via a SOAP call. The engine runs through the calculations and then
sends back an XML message. With Web services, you're making a call to another
database; there are no people involved.''
According to Savvion's Ketabchi, ''Our rules engine itself is a Web service;
to embed it into a product, you don't need an API.'' In addition, support for Web
services has been built into the latest versions of the Savvion BusinessManager
business process management system and the new BizIntegrator B2B process
integration component. According to representatives at the company, Savvion
customers will be able to add automated Web service processes, such as order and
quotation services, to their existing business process applications as well as
publish their own business processes as services in the UDDI directory.
HNC Software, too, is among the vendors positioning in this area. ''The 4.1
release [of Blaze Advisor] has significant additions in the area of Web
services,'' said the firm's Molay. For example, ''you can create a decision
process as a published Web service itself,'' he added. Blaze Advisor can be used
in both the Java environment and Microsoft's .NET environment.
Users are also following the market trends and looking ahead to Web services.
At STW Fixed Income Management, for example, Plasil said they are working on a
second application, this time an accounting system, using Corticon. ''The
advantage is that we can talk across platforms. We don't have to be hooked onto
WebLogic; we can have remote users working on it. Corticon doesn't have to
operate as a session bean; it's very flexible -- you can plug other things to
it, or break it down into components and put them on different machines, all
talking to one another through a bus structure, all through Web services. It's
clearly the direction the industry is going.''
Leveille at Providence Washington Insurance categorizes what he is doing with
Ilog as a Web service, albeit a rudimentary one. ''I will be enhancing our
product as I broaden Web services, not only for what services I offer, but for
what services I have appetite for. In my vision, we will be doing the entire
insurance process electronically. Ilog will help to automate the process by
taking humans out of it; it's a big component in my drive to Web services.''