Skip navigation

Hello again, Blaze users! If you ever stop and wonder what others are doing with decision rules, you’re not alone. FICO World 2018, our bi-annual conference that brings together FICO decision management customers and experts from across the globe, always features customers who have adopted Blaze Advisor as a critical part of their business decision engines.

 

With that in mind, and the conference just around the corner (sign up if you’re interested!), here is a preview of a couple of FICO clients who will be sharing their stories.

 

African Bank

  • Fast facts: Retail Bank in South Africa: Realigned their business (2016) with a renewed focus on transparency, efficiency, and cost-effectiveness
  • Mission: Revise and re-engineer the entire credit lifecycle
  • Decision Rules Focus: Implement standards-based decision modeling, decision implementation best practice, streamlined credit decisions and business logic
  • Business Impact: 30% faster to implement new strategies leading to 20 releases in 8 months after go-live; 25% cost reduction due to rapid implementation and reduced testing times

 

Worldpay (formerly Vantiv)

  • Fast facts: Global leader in payments processing; the recent merger between Vantiv and Worldpay creates the #1 global acquirer
  • Mission: Increase industry leadership over FinTechs and other disruptors with integrated, transparent on-boarding platform
  • Decision Rules Focus: Early adopter of FICO Decision Modeler (powered by Blaze Advisor) has developed a highly integrated, automated merchant on-boarding platform, which it is expanding to consumer underwriting
  • Business Impact: Happier clients (particularly with consumer underwriting capability), faster/smarter automation, able to leverage machine learning and streaming data through FICO Decision Management Platform

 

The complete FICO World agenda is available here.

Are you curious how decision rules and big data technology can enable a smaller financial services company to compete with enterprises many times its size? Specifically, one mid-sized EMEA bank sought to enable new business and revenue growth through more timely and relevant product offers that complied with risk management criteria.

 

The story of how they developed and deployed a direct marketing application using FICO® Blaze Advisor® DRMS – in just under two months – is detailed in this new white paper, authored by Enterprise Management Associates’ noted analyst Steve Hendrick.

The paper examines how the direct marketing solution, which is a single-batch application, is driven by two distinct decision rule services. The first decision service is eligibility-focused and based on credit risk, while the second focuses on offer management and considers relevance and a customer’s propensity to engage. When combined with the bank’s Apache HBase and Hadoop-based Big Data architecture, covering the need to examine 20 million prospects, Blaze was the obvious choice for the client, as the platform supports Hadoop as well as inference-based and sequential rules execution.steve headshot.png

 

The capabilities of the deployed solution include the ability to:

  • Proactively identify prospects for financial services and assess eligibility for inclusion in marketing campaigns.
  • For each eligible prospect, select the most relevant product offer and approach through optimal channel of direct marketing.
  • Process large volumes of data efficiently, extend timely enticements, and improve deal-close and performance metrics.

 

We look forward to your comments and questions about this new case study. Let us know what you think and get in on the discussion in the FICO® Blaze Advisor® User Forum.

 

Download the full white paper here.

Did you hear the news? You can now import a single-step or multi-step SAS program into Decision Modeler and reference it in your decision logic.

 

Here's a quick video that shows how easy it is to import a SAS program in Decision Modeler:

 

To see this new feature in action, get the latest trial version of Decision Modeler.

 

Attached to this blog is a decision service with a SAS program (created in FICO® Analytics Workbench™) that you can download and upload to your trial version.

I’m excited to share a new feature that was added in Decision Modeler 2.3. It’s one of those features that I use so frequently now that I wonder how I ever got along without it. There is a new search field on the Command Bar that lets you quickly locate one or more decision entities that match a simple text query. When a string is entered, the search returns a list of any decision entities that contain the string in its name or in the content.

 

This video shows the new search field.

 

Here’s just one example of why I especially like this feature. Last week I was finalizing work on a decision service when the individual who had requested the decision service informed me that some of the property names in the object model needed to be changed. Rather than spending my time trying to get out of making the changes, I decided to give the new search field a try. Within minutes, I knew exactly where the properties
were being used and how long it would take to make the requested changes. Within a half hour, I had updated the object model, modified the logic, and had a clean run for my test cases in Decision Testing.

 

Try it out, and and explore all the other new features with the latest trial version of Decision Modeler.

Can you explain a decision made just minutes ago in one of your production systems?

 

Decision Modeler 2.3 expands the capabilities of Decision Testing to include testing and analyzing data generated from an executed decision service in the Decision Management Platform (DMP). Data generated while executing a decision service via a batch job, SOAP or REST web services are automatically stored in the Analytic Datamart (ADM) and can be imported into Decision Testing.

 

To explain any decision, all you have to do is open the Decision Analyzer to trace the execution flow to see why a decision was made.

 

The results from the decision service execution are used as expected data in Decision Testing. This allows you to see how additional changes you make to the decision logic in Decision Modeler affect the decision service outcome.


The value of this feature is apparent when you need to run Champion and Challenger type strategies. Before actually introducing these changes to your strategy, you can assess their impact on the outcome of the decision service. After you have decided on your approach, you can simply deploy the compiled project to the appropriate staging environment.

 

See this feature in action in the video below.


 

Take the power of your decision testing to the next level and explore all the other new features with the latest trial version of Decision Modeler.

Have you ever wished that you could visualize data flowing through your decision trees so you can easily assess the impact of your changes?

 

By using Decision Tree Profiling, you can now upload a dataset to profile decision trees. A profiling run executes the data against the decision tree logic. This provides information about the path followed by each record, as well as aggregate information for all records that flowed through a particular node in the tree. A color is assigned to each action; these colors allow you to distinguish the proportion of records flowing through each node of the tree so you can identify what in being assigned to each particular action. These visual insights help you make changes to your decision tree logic to better align it with your business goals.

 

There are several tabular views that you can use to see the results, including: a profile of all nodes, a listing of leaf nodes, and a summary of actions. These tabular views are particularly helpful if you have a large decision tree with many different actions.

 

Here's a brief demo showing how to use the Decision Tree Profiling feature:

 

Get the latest trial version of Decision Modeler and see how easy it is to author decision logic.

In Decision Modeler 2.2, you can find where a decision entity (such as a function, ruleset, decision tree, decision table, or scorecard) is referenced in your decision logic by clicking the Show Usage icon on the Project Explorer or editor toolbars. References to a particular decision entity are displayed as links in the Usage pane. You can click a link in the Usage pane to preview or open the referenced decision entity in an editor.

 

showusage+1.png

 

Knowing where a decision entity is referenced can help minimize any errors that are introduced when modifying a decision entity or removing it from your project. Watch a brief demo below showing how to use this feature and navigate to any decision references in your project.

 

 

Comment here with any questions. The example shown in the video is included with the trial version of Decision Modeler. Get the latest trial version of Decision Modeler and see how easy it is to locate where decision entities are referenced.

Update! The Java Object Model support added as a beta feature in Decision Modeler 2.1 has been improved and officially launched in Decision Modeler 2.2. You can manage and import Java classes directly in Decision Modeler, no need to go through a Blaze Advisor project upload anymore.

 

Not only can you bring in Java classes as part of your business object model, but you can also use a Java class to define the entry point signature in your setup configuration. Now when you import a .jar file and use one of the Java classes, you see messages to help you manage dependencies in situations where you may need to modify your .jar file contents.

 

microblog_JavaBOM.png

Comment here with any questions. Get the latest trial version of Decision Modeler and see how easy it is to author decision logic.

You can now import a Tree Ensemble model as a PMML Mining Model in Decision Modeler 2.2! This new feature allows you to easily leverage your machine learning models in operational decision making. Decision Modeler supports gradient boosted trees and random forest trees, allowing you to easily use models to impact a decision.

 

Here's a quick video that shows how easy it is to import a PMML Mining Model into Decision Modeler:

 

 

Comment here with any questions or ideas. Get the latest trial version of Decision Modeler and see how easy it is to upload a PMML Mining Model.

With Decision Modeler 2.1, you can create a new scorecard and a reason code list without having to import a PMML model or upload a Blaze Advisor project. The Scorecard editor’s sleek new interface makes it easy to create a scorecard and to associate a reason code list.

 

scorecard.png

Enhancements to the Scorecard editor allow you to set options for reason codes such as the number of reason codes to return with an executed bin and the ability to select a method for how reason codes are returned.

Get the latest trial version of Decision Modeler to see how easy it is to create and edit scorecards.

 

Great news. With the release of Decision Modeler 2.1, you can create a decision table without uploading a Blaze Advisor project. Decision Tables remain simple and intuitive to edit. The Decision Table editor has the features you have come to depend upon, including:

  • Ability to import and export values using a .csv file. Import the values, modify them as needed, and export them to use in other decision tables.
  • Rule Profiling to verify that the decision logic is generating the expected results. Just attach a sample dataset and watch as the results are displayed as a percentage and/or a numeric count of the data that matches the conditions. You can even get real-time results while editing the values.
  • Filtering to control the decision logic that is displayed. Select or enter an expression in a filter and only the decision logic that contains the expression is displayed.

Get the latest trial version of Decision Modeler to easily create your own decision tables.

 

What's new with FICO Decision Modeler 2.1

  • Generic scorecards and decision tables: Customers can now use decision tables and scorecards in Decision Modeler without the need to create templates. Business users no longer require any assistance from technical users, putting the power into the business user's hands and allowing for greater control, increased productivity and efficiency.
  • Analytic Integration:  Business users can now inject prescriptive analytics into decision processes by importing either PMML (Predictive Model Markup Language) for scorecards or FSML (FICO Strategy Markup Language) for decision trees. Once the prescriptive models are imported, they can be viewed and edited by business users and then executed to be part of the decision service. Business users can also easily map the variables used in the prescriptive models to their operational data structure.
  • Scorecard Enhancements:  Scorecard has been redesigned to have an automatic bin creation which will detect gaps in overlaps. This feature is essentially a way to check scorecard integrity.
  • SOAP/RESTful Deployment:  Decision Modeler now allows for multiple ways to support decision logic including SOAP, RESTful and Batch. This allows customers to have more options to choose their desired deployment method for different applications.
  • Java BOM Management: Java BOM (Business Object Model) Management enables business and technical users alike to quickly upload and refresh their operational data structure, reducing their dependency on IT. This will result in improved processes and increased speed in the ability to update and refresh data definitions and provide custom data providers with their decision logic.
  • Decision Testing: Users can conduct decision testing and validation at the same time, as well as upload validation data. If a user uploads their validation data and runs the test, the test will be validated and can easily inform the user if the test cases have passed or failed, ensuring the quality of the decision.
  • Reason Code Editor: In addition to returning a score, you can configure a scorecard to return a reason code for each bin that is fired. Reason codes provide users with reasons as to why a particular score was assigned during evaluation.