Building and Deploying World-class Predictive Models and Data-driven Strategies

Download the PDF

PREDIGY™: Integration and co-existence

The introduction of new technology into your organization can be daunting, especially when you must integrate that technology into key business processes and legacy applications. The modular architecture and open interfaces in the PREDIGY™ customer analytics and decision management platform, from Intelligent Results, keep disruption to a minimum. This whitepaper describes the interfaces and options that make integrating PREDIGY with your operational applications so much easier than the alternatives.

An Integrated Platform

PREDIGY is an integrated customer analytics and decision management platform that combines analysis, prediction, strategy and reporting capabilities. The integrated modules within PREDIGY speed the process of building and deploying world-class predictive models and data-driven strategies.

The PREDIGY platform consists of the following modules:

• IR Discover™ A framework for identifying and assessing insights hidden in structured, semi-structured and unstructured data; discovered text-based attributes can be imported into PREDIGY models and strategies.
• IR Modeler™ Powerful modeling techniques and algorithms, automated variable selection, virtual variable transformation capabilities, and rigorous built-in model validation.
• IR Strategy™ Simulation and forecasting capabilities that support real-time what-if analysis using any number of key business metrics; identification of a desired outcome or business measure to optimize.
• IR Report™ Operational reporting, analysis and monitoring of models and strategies running on the IR Production Engine.
• IR Production Engine™ Scoring of operational data against predictive models and decision trees; loading of results into a reporting database.

IR Modeler, IR Strategy, and parts of IR Report make up the PREDIGY Design Environment, a web-based tool for creating and modifying predictive models and decision trees.

Getting data into PREDIGY

In data analysis or production mode, PREDIGY is able to load data from a variety of sources, including text files, relational databases, and Web sites. Interfaces are provided for the following data sources:

• Comma-delimited (CSV) files
• IRM (proprietary text format, similar to CSV)
• Unformatted text files
• SQL databases, via JDBC
• XML
• Web sites, via HTML crawler

In some cases, you may need to pre-process data destined for one of the modeling applications (IR Modeler or IR Discover). These applications typically operate on historical data. Aggregation or summarization of individual data records may be required prior to loading.

Getting data out of PREDIGY

In data analysis mode, statisticians and business analysts interact with data in the PREDIGY Design Environment, reviewing graphical representations of distributions and results. Statistical information on variables and other data can be exported from the user interface to a CSV file.

In production mode, PREDIGY models and strategies are integrated into the overall workflow. The default output from the IR Production Engine is text files (CSV), facilitating integration with other applications. Additional Production Engine output formats are provided through extensions called Listeners, which are Java applications triggered by events in the Production Engine (such as record scoring). Records passing through the Production Engine are copied to the Listener, which can then format the output as required.

A default listener configuration file specifies logs where errors and statistics about scoring operations are written. An additional listener formats output for loading into an Oracle or SQL Server database.

Integrating with other applications

Use PREDIGY to analyze data from operational systems or to to operationalize the results of your analyses through model scores or strategy codes. Analysis collects historical, demographic, and communications records to build models that predict customer behavior. These models can then be used on operational data to select specific treatments for each customer.

The input data for analysis may come from operational systems, but is more likely to be found in a data warehouse or mart, because analysis typically requires data that has been aggregated or summarized. The data warehouse or mart contains summarized account information that can be input directly into PREDIGY without additional pre-processing.

On the operational side, account records are typically used as input into the IR Production Engine, processing the records and appending model scores and/or strategy codes that are then consumed by the downstream operational systems. Account records are typically in text format (CSV), as is the Production Engine output.

The Production Engine itself can be operated in batch mode via a command-line interface, or on a record-by-record basis via SOAP.

Collections/recovery

PREDIGY can predict customer behavior and to prescribe specific actions. On the modeling side, PREDIGY uses historical data is extracted from the collections management system to predict how customers will respond to a settlement offer, for example. You can combine model results with other data to produce an overall plan for a portfolio. When you feed account records through the plan, the Production Engine assigns a strategy code to each record. You close the loop by feeding account records and associated strategy codes back into the collection management system. Typically, you use the strategy codes to assign accounts to specific queues, which guide the activities of the collection agents.

Marketing

You can integrate PREDIGY with CRM applications in several ways. For example, you can analyze emails, call center notes, and other text data to gain insight into customer behavior and intent. You can then use the results of such analyses to segment customers into groups with similar characteristics, feeding the resulting segmentation back into the CRM system through one of the previously-described interfaces and directing promotions or treatments to specific segments by the CRM application.

You can use customer demographic data, along with response data to promotions and sales history, as input into models for predicting future behavior. Typically, models predict the likelihood of specific actions, such as response to a promotion or up-sell/cross-sell offer. The predictions are routed back into the CRM system as input into specific campaigns. After the campaign has completed, outcome data is routed back to the predictive models to improve their effectiveness.

Conclusion

Integrating analytics into key business processes and/or legacy applications has never been easier than with PREDIGY, an integrated customer analytics and decision management platform from Intelligent Results. PREDIGY accepts input data and supplies output data in a range of standard formats that are compatible with most systems. The modular architecture provides abundant integration points with operational systems and processes, during both data analysis and production.

Models and strategies that you build with the PREDIGY platform can be integrated with collection management and CRM applications to predict customer behavior. The IR Production Engine scores records and can create strategy codes for records from operational systems, based on model and strategy definitions created in IR Modeler and IR Strategy. These scores and strategy codes can then be used by the operational systems (CM, CRM) to process customer/account records. Finally, support for both batch and record-based scoring enables the Production Engine to adapt to a variety of workflows and processes.

To request a personalized demonstration of the PREDIGY platform, please click here.