≡ Menu

Business Rules Forum 2009 – Day 2 #brf


The second full day of the Business Rules Forum/Enterprise Decision Management Summit is over and once again I have been taking notes rather than blogging live. Once again there were some great sessions – today I heard Steve Hendrick of IDC, Sandeep Gupta of Equifax, Chaitan Sharma of DAASL, Zach Springborn of OneData and Mo Masud of Deloitte. I have to confess to skipping Eric Siegel’s session as I have heard it recently (check out 5 ways to reduce cost with predictive analytics or watch the webinar Eric and I did on Optimizing business decisions with predictive analytics).

Keynote: BRMS at a Cross Roads: The Next Five Years

Steve Hendrick thinks that the business rules market is at a crossroad thanks to increased visibility by virtue of the big players joining the market, the growing acceptance of decision services as an implementation approach, the availability of open source alternatives and more. Nevertheless the market is currently small – just $285M for the BRMS market itself out of a total software market of more than $280B. The market is continuing to grow but from a fairly small base. Steve sees no reason for this growth to stop and identified market momentum, growing awareness and companies’ desire to improve Governance Risk and Compliance or GRC as reasons for this growth to continue.

Steve, like me, sees the future not in terms of BRMS products as a standalone market, however. Instead he is talking about a Decision Management Platform and his description of it would resonate with any regular reader of this blog – business rules, predictive analytics and event processing. BRMS is good for categorical or data-driven decisioning while analytics can improve rule relevance and quality, make the data more readily consumable and refine the outcomes of rules-based decisioning to find optimal solutions to decisioning problems. The combination is thus particularly powerful.

Steve adds event processing and the handling of state because he believes that an always-on decisioning platform that listens for events, analyzes the information available when the event happens and then decides what actions/processes are appropriate is the best way to handle decisioning in an increasingly real-time world. He sees a decision management platform as enabling a move away from process-centric to information-centric decision making.

Steve advised vendors to focus on analytics as a complement to business rules, to think about how to migrate to the decision management platform he outlined and to track developments in CEP, BRMS, Analytics and event-driven architecture more generally. For customers he again advised a focus on analytics, especially segmentation (like me he thinks this is the best way for companies using rules to get started with analytics), and urged them to think about real-time decisioning going forward.

Maturing with Business Rules and Business Intelligence – The Combined Power

Sandeep gave an interesting presentation on the use of rules with business intelligence. He was talking not about using BI in the context of a transactional decision, the major use of business rules, but the possibility of using business rules as part of the strategic or management decisioning that relies on BI. Equifax is a user of business rules in a number of its solutions, but these use rules to handle decisions at a micro level – is this transaction fraudulent, is this customer a good risk etc. The work he was discussing in this presentation was attempting to bring the power of rules to increase transparency and business control as well as the action-orientation, to a BI environment.

Clearly the two environments have a significantly different focus – transactions on one hand, summaries and analysis on the other. Yet both are built on a business object model and in both cases there is a move to empower business people to write their own rules or build their own dashboards. If the business object model could be fused between the two environments, perhaps the rules approach could be used to make the dashboards and reports more actionable. This would require that the rules could have conditions that took advantage of the statistical analysis and summary capabilities of the BI environment as trying to handle this kind of data crunching in a rules engine would perform poorly.

Equifax has been conducting some research work to allow a user to specify the query, the WHERE clause if you like, in the condition of a rule using the business-friendly and English-like syntax of their rule engine (they use IBM ILOG). The rule can then be executed in the rule engine and can reach into the BI environment to get the data analysis it needs. Actions are taken using the rules engine’s usual ability to execute the THEN side of a rule. While this research is interesting, there are challenges with the impedance mismatch of the two environments as well as the difficulty of turning intuition (based on this graph I think I should do this) into an explicit rule.

An interesting topic – almost the reverse of the usual rules+analytics story in that it is bring rules to an analytic summary rather than analytics to a rules-based transactional decision.

Decision Automation – Implementation Challenges and Productivity Improvements

Chaitan presented on his experience working with ICICI Lombard, India’s largest general insurer, on a legacy modernization effort using FICO Blaze Advisor. The driver to modernize these applications was a change from fixed to variable pricing in the insurance market along with an influx of new customers about which very little is known – the information business is less developed in India so many of these customers must be taken on with an expectation that a different approach will be required once their behavior becomes known. It was a classic legacy modernization challenge in that SMEs were in short supply, the code was not documented and the system was in use while it was being reengineered.

To manage the process, DAASL focused on a couple of things. They classified business logic so they could prioritize the use of business rules based on the likely degree of change and the degree of reuse. They built a pilot by observing the current system so they could show business users how it would look. They did code walkthroughs to draw out rules and then matched these up with the output of business user interviews to see where the inconsistencies were. And they delayed integration/governance functionality until the business had experience with several decisions and with several iterations so that the requirements for these elements would be grounded in actual experience with the BRMS. All good advice for anyone approaching rules for the first time.

Besides the general improvement in agility and accuracy from using business rules, the system also gave the business the power to generate explanations and reason codes for the decisions as they were made and to log exactly how each decision was made. These were important as it enabled some real analysis on the part of the business into what works and what does not.

One final piece of advice from the team was to keep on top of the technology/platform deployment issues. Incompatible product versions, sudden changes in the platform and other technology-centric problems could have derailed the project even though the rules piece was going well. Keeping the project on track meant dealing with these issues and not allowing them to derail the main effort to extract, document and implement the business rules.

Decisioning: Managing Counterparty Risk for Financial Markets

Zach gave an interesting presentation on the use of business rules and extract-transform-load tools in delivering better counterparty risk management. OneData helps companies manage risk by providing information to them and, in this case, the focus was on counterparties. In particular the challenge of managing the final counterparty – who owns the company that owns the company that owns the company that you are considering taking a position in? And what, therefore, is your exposure to a risk that the parent company gets into trouble. He gave a great illustration of how complex companies like ABN AMRO or AIG get with a wonderful snowflake diagram with ABN AMRO at the center and all its layers of subsidiaries fanning out.

Anyway, the project combined TALEND (open source ETL tool) with IDIOM’s Decision Manager to bring multiple data sources together and largely automate the process of matching securities that were being bought and sold with the ultimate counterparty. About 90% of the securities could be matched using the decisioning engine and the remainder had between 2 and 15 options that had to be manually considered. This is complicated by the fact that different customers might have different rules for who counts as a counterparty – some consider 25% ownership enough, for instance, while others consider 51%.

The end result was a counterparty hierarchy matched to the individual securities that was fed into the Algorithmics risk management platform so that companies could consider the risk of their portfolio in terms of the ultimate counterparties involved in their positions. Traders and desk managers could see what their total exposure was to a particular parent company or what the impact of a particular trade would be on their risk exposure.

A somewhat unusual use of rules but a fun example of how two technologies, ETL and business rules, can be used together to deliver real business value.

Integrating Predictive Analytics and Business Rules Management to Enhance Insurance Marketing Strategies

Last session for me today was Mo Masud talking about the role of predictive analytics and business rules in commercial insurance. Deloitte has been doing a lot of work in this area, focusing on expanding the use of predictive analytics beyond its usual home in underwriting and pricing to marketing, agent selection, customer service and claims. This is important as the insurance industry has been losing money, as an industry, for a most of the last 10 years. Combined ratios over 100 (meaning they spend more on claims and admin than they take in in premiums), poor investment returns and some bad years for catastrophes have all contributed. But the net effect is that insurance companies must get better at their core business so they can drive down their combined ratio – and rules/analytics/decisioning has been shown to do just that.

The use of rules and analytics to automate decisions and so drive out subjectivity, identify profitable opportunities and improve efficiency is critical. In particular, an enterprise view of the use of these approaches so they are not applied only in underwriting, say, but broadly and systematically works wonders for the bottom line.

Deloitte has found that companies need to focus on data and data sources because these drive the kinds of models they can develop and on business rules to make these models actionable. Mo gave a great example of using a predictive analytic model to rank-order those applying to be agents in terms of how likely they were to become profitable for the carrier. This model was combined with rules to allow many applicants to be accepted or rejected automatically while those in the middle were referred to marketing managers for further analysis. Rules considered things like geography and coverage as well as the likely profitability of the agent. Using this approach, and focusing on signing up the agents in the top 7 deciles of the model, one carrier shaved 2 points off its loss ratio and saved 10-15% of its marketing/administration fees for agency management.

A good illustration of the power of analytics and rules in areas of the business not traditionally associated with them.

Here are other posts on today’s sessions:


Comments on this entry are closed.