Making your systems smart (enough) - ITARC Atlanta

Making your systems smart (enough) James Taylor VP Enterprise Decision Management Fair Isaac

Presentation Outline

Smart (enough) systems defined

Focusing on decisions to make your systems smarter

An architecture for automating decisions

Decisions, SOA and BPM

Getting there from here

First, introductions

James Taylor

Author of “Smart (Enough) Systems”

Blogger and writer on decision management and decision management technology

Experienced developer, development manager, product manager and product marketer

Not a singer (my brother is though)

Fair Isaac

Historically in credit risk and fraud detection

50 years old

“ Decisions” company

Dumb Systems Are Everywhere… Report but don’t learn Built to last, not to change Wait rather than act CRUD

Why Do You Need Smarter Systems? Decision-Making Boundaries Well-Defined Cross Channels / Geography / Departments Timeliness of Decisions Days Real-time / Point of Contact Level of Objectives Simple Complex Trade-Offs Regulatory Constraints Static Dynamic Changes to Decision Strategy Every 3-5 Years Frequent Value Creation Efficiency Effectiveness Operational Volume Low High

So What IS A Smarter System

Operational

Real-Time

Rapidly evolving - agile

Learning

Customer-Centric

Extended-Enterprise Ready

Demonstrably Compliant

Smarter Systems Make More Decisions People Not Made Embedded People Embedded Not Made New Before After Larger boxes represent more decisions, by volume

More and more decisions can be automated Complexity Value Automated Decisions Expert Decisions Manual Decisions

Different kinds of decisions E C ONOMIC IM P A C T OF INDIVID U AL DECISION Low High High ECONOMIC IMPACT OF INDIVIDUAL DECISION Low DECISION VOLUME High-value, low-volume decisions Medium-value, medium-volume decisions Low-value, high-volume decisions

Applications - from monolithic… The App

…to… Data Process Logic User Interface

… composite Data Process Logic User Interface BPM Browser Enterprise Database

Next Up – Business Logic Logic Decision Services A self-contained, callable service with a view of all the conditions and actions that need to be considered to make an operational business decision What is a decision service? A service that answers a business question for other services

Enterprise Decision Management is an approach that automates, improves & connects decisions to enhance business performance

Four components

Decision Services

Adaptive Control

Business Rules

Predictive Analytics

Decision Services

A subset of all services

Transparent, agile, coherent

Separate out decision logic

Business Rules Management

Define and manage the logic of decisions

Automate operational decisions

Let business users manage their rules

Rules are not AI, not code

Declarative not procedural

Forward not backward chaining

Accessible not obscure

Verbose not abbreviated

Business Rules Are Everywhere Experienced Personnel Regulations Policy Manuals Legacy Systems Managed Business Rules Historical Data

Some Examples Smart (Enough) Systems, Prentice Hall June 2007. Fig 4.3 I f cus t ome r ’ s debt e x c eeds cus t ome r ’ s assets then set the cus t omer's applic a tion st a tus t o D ecline d . I f o r de r ’ s pu r chasedD a t e is ea r lier than Janua r y 1, 2004 then p r i n t ( “ Y our pu r chase is no longer eli g ible f or r etu r n ” ). I f ( v ehicl e ’ s age is be t w een 0 and 8 y ears) and (poli c yholde r ’ s age is be t w een 21 y ears and 60 y ears) and (poli c yholde r ’ s number of claims does not e x c eed 3) then set poli c yholde r ’ s case t o “ S T AN D AR D ” M a r tial S t a tus C r edit S t o r e A cc ou n t Balan c e A cc ou n t A ge Single M a rr ied S epa rat ed T r e a tme n t Close_ A cc ou n t < 600 600 -750 > 750 < USD 900 >= USD 900 < 18 >= 18 T r e a tme n t E x t end_ T e r m T r e a tme n t F o r g i v e_ D ebt T r e a tme n t Close_ A cc ou n t T r e a tme n t R edu c e_ D ebt T r e a tme n t R edu c e_ D ebt T r e a tme n t Close_ A cc ou n t I n c ome C ondition I n c ome Limit A c tion C r edit Limit A c tion C r edit Limit A c tion C a r d T ype C ondition S tude n t B r on z e S tude n t G old S tude n t P l a tinum 7,500 - 9,999 10,000 - 19,999 20,000 - 29,999 30,000 - 39,999 40,000 - 49,999 60,000 - 69,999 50,000 - 59,999 70,000 - 79,999 80,000 - 89,999 90,000 - 99,999 1,000 1,100 1,200 1,500 1,500 1,600 1,700 2,000 2,200 2,200 2,000 2,100 2,500 2,500 2,700 2,800 3,800 3,000 3,300 3,500 4,000 5,000 5,200 5,200 5,700 4,500 4,500 4,000 4,800 4,700

Empower the Business to Manage the Rules So you business-types want to be able to change your business rules? I want to relax my underwriting policy I want to be able to promote a new product combination I need to add the new regulations No…

Which Would Your Users Understand? Smart (Enough) Systems, Prentice Hall June 2007. Fig 6.8 public class Application { private Customer customers[]; private Customer goldCustomers[]; ... public void checkOrder() { for (int i = 0; i < numCustomers; i++) { Customer aCustomer = customers[i]; if (aCustomer.checkIfGold()) { numGoldCustomers++; goldCustomers[numGoldCustomers] = aCustomer; if (aCustomer.getCurrentOrder().getAmount() > 100000) aCustomer.setSpecialDiscount (0.05); } } } If customer is GoldCustomer and Home_Equity_Loan_Value is more than $100,000 then college_loan_discount = 0.5% College Loan Discounts Current Discount = Eligibility Gold Customer and Home Equity Loan more than $100,000 %

Predictive Analytics

Forecast future behavior and performance

Target based on risk and opportunity

Derive actionable insight from data

Analytics <> Graphs

Even graphics require interpretation

Not everyone can see the patterns

And code does not “see” at all

Smart (Enough) Systems, Prentice Hall June 2007. Fig 9.3

Descriptive Analytics - Improve Rules Use: Find the relationships between customers Example : Sort customers into groups with different buying profiles. Operation : Analysis is generally done offline, but the results can be used in automated decisions – such as offering a given product to a specific customer © Fair Isaac Corporation, reproduced with permission * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * Low-moderate income, young High Income High income, low-moderate education Moderate-high education low-moderate income High Moderate education, low income, middle-aged Low education, low income Education Age High

Predictive Analytics – Add Insight Use: Identify the odds that a customer will take a specified action Example : Will the customer pay me back on time? Will the customer respond to this offer? Operation : Models are called by a business rules engine to “score” an individual or transaction, often in real time © Fair Isaac Corporation, reproduced with permission 170 190 210 230 Good Customer “ Bad” Customer

Adaptive Control

Learn faster with “champion / challenger”

Find the ideal decision for every customer

Connect decisions for customer-centric growth

Impact May Take Time to Play Out

What Is Champion/Challenger Anyway? Unknown Optimal Approach Single Approach Decision Space Considered Unknown Optimal Approach Champion Challenger 1 Challenger 2 Decision Space Considered Unknown Optimal Approach New Approach New Modelled Decision Space New Learning Approaches / Extrapolations

Analytics To Optimize And Adapt Use: Design a ruleset that will deliver the right decisions to reach goals Example : Identify how much money to spend on each marketing channel to maximise sales in a given timeframe and budget Operation : Decision models are used offline to develop rules, which can then be deployed to operate in real time © Fair Isaac Corporation, reproduced with permission # Samples Response Score Margin Prescription Volume Total Calls Cost Cost of Goods Sold Distribution Cost Unit Sample Cost Total Sample Cost Net Revenue per Rx Specialty Previous Rx Census Data Region Cost for each call Call Plan Profit Input Action Reactions / Factors Objective

Putting The Pieces Together Smart (Enough) Systems, Prentice Hall June 2007. Fig 5.1 Business Rules Adaptive Control Production Application Enterprise IT Infrastructure Predictive Analytics Data Warehouse Decision Service Predictive Model Operational Data Store Rules Rules Predictive Model Rules

The Basic Process

Identify Decisions

Integrate Decision Services

Automate with business rules

Empower the business

Analytically improve

Add predictive insight

Optimize and adapt

Many Decisions Are Hidden

Data and Analytic Model Technology Smart (Enough) Systems, Prentice Hall June 2007. Fig 5.5 Data Preparation Tools Event Processing Engine Model Deployment and Execution Modeling Tools Automated Tuning Production Application BI/PM Search Engine Model Repository Content Management System Data Warehouse Operational Database Decision Service Events

Business Rules Management Technology Decision Service Smart (Enough) Systems, Prentice Hall June 2007. Fig 6.6 Design Tools Rule Management Applications Rule Engine Production Application Operational Database Rule Repository

Adaptive Control Technology Production Application Smart (Enough) Systems, Prentice Hall June 2007. Fig 7.8 Simulation Environment Random Data What-If Scenarios Historical Data Optimization Engine Decision Service Champion Strategy Results Production Data Business User Tools Analysis Strategy Design Modeling Tools © Fair Isaac Corporation, reproduced with permission Challenger Strategy A Challenger Strategy B Challenger Strategy C Model Repository

Benefits Of Decision Services

“ Built to change not built to last”

In the real world, nothing is set in stone

What do you do if a service must change a lot?

Managing “change-time”

Legacy Modernization

Externalize decisions hidden in legacy applications

A bridge between your mainframe and your SOA

Differentiation

Buy services based on best practices, standards

Buy services that don’t differentiate you

Build Decision Services for competitive differentiation

SOA Principles Extended After Thomas Erl, SOA Systems Inc. (www.soasystems.com) Services are outwardly descriptive Discoverabilty Services minimize retaining information specific to an activity Statelessness The business has control over the logic their services encapsulate Services have control over the logic they encapsulate. Autonomy Business rules allow services to be coordinated and assembled by business owners Collections of services can be coordinated and assembled to form composite services Composability Business rules allow this logic to be managed by the business but remain hidden from other services Beyond the service contract, services hide logic from the outside world Abstraction Services maintain a relationship that minimizes dependencies Loose Coupling Services adhere to a communications agreement Contract Business rules add a layer of granularity through the reuse of rules and rulesets Logic is divided into services with the intention of promoting reuse. Reusability

Process Management, Decision Management

Standardize process

How should a process be carried out?

Facilitate compliance and collaboration

Workflow definition and management

Work Queue Management

Integration

Standardize decisions

What should the decision be based on?

Facilitate decision automation

Decision definition and management

Straight Through Processing

Enhancement

Risk and Benefits

Process Management without Decisions

Rules an afterthought

Rules are re-buried in the “new process”

“ New process” becomes complex, burdened by buried decisions

Inconsistency of rules

Entangled processes and decisions

Process Management with Decisions

Simpler “New process”

Decisions are linked but not buried

Decision Services can be tied to multiple systems and processes

Consistent decisions

Independent process and decision change

Rules automate and simplify BPM tasks

Marine vessel underwriting (simplified)

Gather policy application data

Validate for completeness and correctness

Prequalification rules “kick out” bad applicants

Physical vessel inspection, certification, etc.

Make underwriting decision

Take action related to underwriting decision

Write loan

Decline loan

Flag exceptional circumstances and pass to underwriter for special handling

 Process Tasks  Decision Tasks

All Kinds of BPM Need Decisions Sense and Respond Systems People-Centric Automation Transaction-Centric Automation ESB CEP BAM Process Analytics Process Warehouse Decision Services Investigate Exceptions Automation After IDC Group “A View of BPM in 2007 and beyond”, Maureen Fleming

The Evolution Of A Retention Offer

Automate Decision

Apply rules

Segment customers

Predict risk, value

Optimize decision

Web http://www.f Email Call Center Mobile M a r tial S t a tus C r edit S t o r e A cc ou n t Balan c e A cc ou n t A ge Single M a rr ied S epa rat ed T r e a tme n t Close_ A cc ou n t < 600 600 -750 > 750 < USD 900 >= USD 900 < 18 >= 18 T r e a tme n t E x * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * # Samples Response Score Margin Prescription Volume Total Calls Cost Cost of Goods Sold Distribution Cost Unit Sample Cost Total Sample Cost Net Revenue per Rx Specialty Previous Rx Census Data Region Cost for each call Call Plan Profit

Taking Control Of Business Optimal Operation Key Performance Indicator e.g. profitability time Operational Limit

Manage Decisions’ Multiple Dimensions Reduce expenses needed to make decisions COST Execute decisions faster – even in real time SPEED Adapt dynamically to changing conditions AGILITY Ensure coherence across channels, business units and geographies CONSISTENCY Make more profitable and targeted decisions PRECISION

An overview of traversing the steps in EDM Smart (Enough) Systems, Prentice Hall June 2007. Fig 9.1 Phase 1 - Piecemeal Phase 3 - Expansion Readiness Assessment Phase 2 – Local Decision Management Rules and Analytics Champion Challenger Steady State EDM Enterprise Management Analytic Value Chain Optimization and What If Enterprise Backbone Broaden Analytic Base Improved Foundation Foundation First Rules Project First Analytic Project Manage Scenarios Overlapping and Adjacent Projects

Action Plan 1

Identify EDM applications you have

Identify your decisions

Decisions that matter to customers

Hidden decisions

Transactional decisions

Consider

Who takes them now

What drives changes in them

What the context is for them

Action Plan 2

Manage declarative business logic

Improve business/IT collaboration

Focus on decisions

From predictive reporting to executable analytics

Enterprise Decision Management

Smart (Enough) Systems

If your business needs to make quick, accurate decisions on an industrialized scale, you need to read this book. Thomas H. Davenport, Author of “Competing on Analytics”

James Taylor and Neil Raden are on to something important in this book – the tremendous value of improving the large number of routine decisions that are made in organizations every day. Dr. Hugh J. Watson, University of Georgia

This book shows how to use proven technology to make business processes smarter. It clearly makes the case that organizations need to optimize their operational decisions. It is a must-have reference for process professionals throughout your organization. Jim Sinur, Chief Strategy Officer, Global 360, Inc.

http:// www.smartenoughsystems.com

My Contact Information James Taylor [email_address] or [email_address] www.edmblog.com | www.smartenoughsystems.com Grab the decision by the throat and don’t let go!