Business Rules Management System
DecisionSmart is a modular scoring and decision engine built to offer rules management capability on an account or customer level across the entire customer lifecycle.
DecisionSmart offers rules management capability on an account or customer level across the entire customer lifecycle, including the following decision areas:
PROFITABLE DECISIONS ACROSS THE CUSTOMER LIFECYCLE
Batch scoring the target population: Risk, Response, Profitability, Lapse/Cancellation Risk;
Segmentation of leads: Never contacted, Previously declined, Previously excluded, Thin/thick file;
Run champion/challenger campaigns to determine best channel.
Allocate campaign codes, channels based on segments and scores.
Support complex models (e.g. agent-matching for outbound call campaigns across the product set).
Scoring (application, fraud, behavioural, etc.) can all be executed through DecisionSmart’s scoring metaphor.
Scores (e.g. bureau and application) can be combined either in a matrix or decision tree.
Policy rules can be managed through DecsionSmart’s logical expression builder and policy rule groups.
Calculations, for example affordability and limit calculations, insurance premium values or mobile contract tariffs can be conducted through calculated workfields.
Champion challenger strategies can be managed with a random number generator decision trees.
ACCOUNT MANAGEMENT: CREDIT LIMIT MANAGEMENT
For credit limit management, a batch process is normally set-up feeding existing customer information into DecisionSmart – thereafter through a combination of Scorecards, Matrices, Decision trees, and Outcome Table. A final % limit increase offer is then determined.
The final limit offer amount can be calculated based on limit, balance, percentage increase, maximum limit, minimum incremental increase, etc.
ACCOUNT MANAGEMENT: AUTHORISATIONS
DecisionSmart will use the scoring engine, decision trees, authorisation table and workfield calculations to determine the “fit” and “cushion” percentages.
This will be fed back to the POS to approve or decline the transaction.
For customer engagement, DecisionSmart is enabled with the following key features: Scoring, Segmenting, Clustering and Action-Table;
If transaction data is used, then more advanced engagements utilising “next-best-product” are also available.
Primarily it can be deployed to daily segment the delinquent population and then prescribe appropriate actions.
It has also been used to deploy right-time-to-call models enabling call-centres to improve their right-person-connect models.
Furthermore it can be used to deploy a external-debt-collector placement strategy.
Further opportunities lie in using DecisionSmart to calculate settlement amounts where an optimised settlement strategy has been built.
DecisionSmart can calculate, on a monthly basis, the expected losses across each book per account by running the following models: Probability of Default (PD), Loss Given Default (LGD), Credit-Conversion Factors (CCF) and Exposure at Default (EAD);
If Markov chains are used, then through decision trees these values can be applied to each account within a specific segment within a book. Furthermore the expected losses (EL) can be calculated incorporating any macro-economic adjustments, if required.
DecisionSmart can calculate, on an ad-hoc basis fair value of a particular segment of a recovery book. The software utilises: Scoring, Segmentation, Percentage allocation table, Calculation of account valuation based on balance and percentage allocated (using calculated workfields)
DecisionSmart manages all types of scorecards, such as application, behavioural, pre-delinquency, collection and balance-build scorecards.
The heart of the technology, the scoring engine processes data, calculates the score and returns results. Able to host multiple scorecards for different products or classifications.
Defines and configures all elements required for the scorecard to function. Enables users to set standard and derived characteristics and manage properties and scorecard ranges for each scorecard.
Scores can be calculated in batch mode by receiving or fetchingfiles received from the host system, typically used for behavior scorecards. Can be scheduled to run automatically during overnight batch processing.
REAL-TIME DECISIONING SERVICE
Application processing systems often require an application score be calculated in real-time. This is typically through a web service that invokes the decision engine for these ad-hoc scoring requests.
KEY PRODUCT FEATURES
- Rapidly deploy and manage new additive scorecards;
- Run simple to complex strategies using easy-to-build decision trees;
- Build and deploy simple or complex mathematical calculations, generate random numbers, build logical statements, calculate affordability, installment values in an instant;
- Different rule “metaphors” are supported to enable complex decisions to be deployed across multiple brands / product sets;
- Create a variety of customer groups to be used in marketing strategies using clustering;
- Combine two scores in a matrix to create a risk-grade;
- Fully configurable terms-of-business/decision tables to manage the decisions you wish to make;
- Program a sequence of complex decisions with rule flow;
- Single / Batch mode;
- Ability to integrate into a machine learning environment;
- Supports mathematical optimisation;
- Full audit trail of new decision metaphors;
- Configurable reporting through Sequel Server Reporting Studio
- Can run with pre-configured scores/strategies; and
- Can call other decision services (e.g. SQL, R, Python) via stored procedures