I last got a briefing from Sapiens last year and I recently got an update on Sapiens DECISION release 3.2. Obviously Sapiens DECISION is focused on a robust and complete implementation of The Decision Model as described by KPI.

Governance has seen a major update in the recent release. Users create a Business Change Request to describe a change to a decision model. This describes the change from an external perspective – what do people “outside” the model think needs to be done. This is described (optionally with a document) and assigned to a person (with roles defined in the tool) who will act as the lead for the request. Communities can have additional custom properties that are filled out. Once assigned it shows in the user’s work queue for approval.

The change request can use text associated with the request and run text analysis to see what terms and synonyms defined in the model are recognized. The text is saved with the change request, can be marked up and additional synonyms can be added directly during the review. This analysis generates a list of terms and these are then fed into a query that returns model elements (decision views, fact types rule families) that might be impacted.

Individual change documents can be created for a request, allocating the work for the change request to the various communities. Tasks can be created and assigned to target releases (Releases are also a new feature to package up changes). Workflow is created from available, customizable, templates and then assigned to specific users (though group assignment is coming). Elements identified as needing to be changed can be assigned to the white board for the task at this time but in practice this happens when modelers start performing their assigned work. Either way, elements copied to the white board can then be edited using the tools.

Support for all the principles in The Decision Model has been added – so new ways to determine completeness, new contradictions between rules (like intersected rows) are defined. As the rule families are edited and extended these new principle checks are applied in-situ.

Testing has also been extended in this release. A quick test can do a single test while a Test Group can be defined that imports test data or the system can generate test data from fact type test data (valid and invalid), from the domain values for the fact types or from the values used in rule families. For instance all the persistent fact types can be identified and all the test values identified for these automatically. Additional ones can be defined and added. Of course the number of test cases typically explodes as the Cartesian product so the user can decide how many test cases to generate in absolute terms or percentage. For each generated case the user can say it expects it to resolve and pick an expected result. This then becomes the basis for the test cases. Test cases that fail to match are shown and a view of the Decision Model is used to display the conclusions drawn in the test case for rapid understanding of the test though the user can drill into complete execution detail. These tests can accommodate functional testing as well as regression testing.

Once all the work is complete it runs through the approval process and approved elements are associated automatically with the release that was described in the original request. These releases can be deployed and this requires a deployment descriptor that has effectively dates for instance. Different deployment environments can be selected from the installed adaptors (currently Sapiens eMerge, Drools, ILOG, Blaze Advisor, Java, SQL, XML). These generate executable code or rules from the Decision Model (they are not intended to be managed in the target environment, just executed, with all management taking place in Sapiens DECISION). This ability to deploy business logic to various business rule engines allows enterprises to use multiple business rules engines while the ability to generate XML allows for integration with home-grown execution engines.

Other features:

• New in this release is a formal mapping of fact types to target data models (such as ACORD or MISMO standard models). Values can also be mapped (Yes to 1, No to 0 for instance).
• Communities have been added also with support for each community to have its own language or vocabulary as well as settings and permissions, user roles, workflow templates etc.

Sapiens is one of the vendors listed in our Decision Management Systems Platform Technologies report and continues to develop new capabilities.

It’s been over a year since I last got updated on Zementis (their 3.5 release of ADAPA and PMML Universal Plug-in). Since then they have been busy adding support for new platforms and signing new partnerships. Most recently they have announced a strategic alliance between Zementis and FICO, something Zementis sees as a big win for open standards (see this post on the new FICO cloud). The partnership means that Zementis’ PMML-based solutions will run on the new FICO Analytic Cloud platform. The FICO Decision Management Platform will tightly embed the Zementis engine so that any PMML model can be loaded up into it and executed as part of a broader decision service built on the FICO DMP.

The Zementis Universal PMML Plug-in itself has evolved with new partnerships. Zementis has partnered with IBM for Netezza, with Teradata for both Teradata (coming soon) and Teradata Aster, and now supports Hive on top of Hadoop. All of these relationships are in addition to already announced relationships with SAP Sybase IQ, EMC Greenplum and Datameer. Common to all these deployments is massively parallel execution of PMML models on the platform you already own. In addition to platform-independent deployment, of course, the use of PMML allows the model development to be very vendor neutral. Lots of modeling tools are being used to produce PMML models which can then be consumed in various MPP platforms using the PMML Plug-in. The ADAPA platform is also broadening its support with the original Amazon cloud deployment being extended with support for the IBM Smart Cloud Enterprise and now the FICO Analytic Cloud.

At the end of the day Zementis envisions their platform giving companies the maximum flexibility for their analytic expertise – use any data mining tool to build models with the ability to execute those models on any platform. They see an increasing number of customers, major leaders across all various industries, that are using PMML as the primary mechanism to move from analytic development to operational production. Big Data is also driving complexity in analytic model development, making companies want to use a wider range of more specialized tools. At the same time hiring of data scientists is hard, so companies don’t want to restrict themselves to people who know a specific tool. This is reflected on the data mining tool side also with tool vendors improving their support for PMML , plus constantly expanding R support for PMML. More and more features in PMML –  e.g., 4.1 has expanded pre- and post-processing as well as ensemble models – is combining with better support in modeling environments to allow more advanced, more sophisticated models to be delivered as PMML. Momentum is really building they say.

Finally partnerships are important to Zementis and they have added a new Japanese partner, Infocom, to add to Dymatrix in Europe and Purple Analytics for other parts of Asia. These partners are focused on predictive analytics and act as Zementis implementation partners in those countries.

You can get more information on Zementis here and Zementis is one of the vendors in our Decision Management Systems Platform Technologies Report.

A new week and another new white paper. This week it is Beyond Business Agility: Becoming Adaptive and Analytic

Business rules are a powerful tool for developing agile systems. Moving forward, becoming analytic and adaptive as well as agile will be a source of competitive advantage. This white paper is based on my keynote from the Building Business Capability conference.

Contents:

• Introducing Decision Management Systems
• Becoming Analytic
• Becoming Adaptive
• Next Steps

You can find this white paper, and many others, on our white paper resources page.

While at IBM IMPACT this year I was interviewed about Decision Management, Mobile and more. Check out the video below:

If you want to access the white papers I mention you can find them on decisionmanagementsolutions.com/white-papers.

In-memory databases and in-memory analytics are interesting technologies when it comes to Decision Management Systems. Memory is thousands of times faster than disk and the amount of memory on a node is increasing rapidly. Putting data is put in memory and executing analytics against it, especially if the analytics engine has plenty of memory too, can dramatically increase the speed of producing complex analytic models. This allows companies to better cope with increased data complexity (which is driving more use of multi-part ensemble analytic models for instance) and still decrease the time for each iteration in modeling, making data miners and data scientists more productive.

I have written about  few in-memory technologies and approaches in recent months (including IBM DB2 BLU, SAS Overall in-memory strategy and SAP Predictive Analytics with its support for HANA in-memory) and into this space has come Teradata with its new Intelligent Memory. I would summarize the idea behind the Teradata Intelligent Memory product with three points:

1. While much faster than disk, and much cheaper than it used to be, memory is still 80x more expensive
2. Data volumes are exploding even faster than memory availability so putting all your data in memory is impractical
3. Why ask your DBA to do manually what your system can do automatically?

Teradata has taken its prior work on how “hot” data is – how important a given piece of data is in the IO of a database – and extended this so that very hot data is put into memory. This is exactly what you would do manually – put the data that makes the biggest performance data into memory – but does it automatically and dynamically, constantly updating what’s in memory v just on disk to continually optimize performance. Nice.

Anyway, I am going to be writing some though leadership stuff on in-memory processing as it relates to Decision Management Systems so look for that in the coming months.

As part of the ongoing blog series on Marketing Decision Management Solutions I got an update from SAP on Real-Time Offer Management. This is based on an acquisition made at the end of 2006 of a real time recommendation technology developed for the CRM market, with early adoption mainly by the competitive service industries – Telco and FSI. With the emergence of loyalty and mobile more broadly, this technology has evolved into SAP Real-Time Offer Management and is being adopted by industries such as Retail, Consumer Products and Hi-Tech who want to engage their customers over multiple digital and social channels.

The intent behind SAP Real-Time Offer Management is to generate offers that are targeted, contextual, customized and optimized. One can analyze a lot of historical data to become intimate with customers, finding out their intentions using consumption, purchasing behavior etc. At the same time context is the core for real-time analysis, finding relevance by considering what is going on right now. Mass promotions that are neither contextual nor targeted, contextual but not targeted offers or targeted offers that are not contextual are all equally flawed – what is required is offers that are both optimized and customized to be targeted and contextual.

SAP Real-Time Offer Management solution contains several elements:

• Multi-channel contextual recommendation engine
  Available either stand alone or as part of SAP Precision Retailing as an on-demand solution
• Automatic learning engine that uses responses to drive next best action or offer
• Integrated tools for end to end offer management i.e. offer design, simulation, analysis, management, integration etc.
• Specific industry solutions and connectors to different SAP solutions, such as CRM, ERP, eBanking, etc.

SAP Real-Time Offer Management supports the classic marketing flow – define offers, feed these into the system, and then connect a customer to the optimal recommendation by combining a trigger containing real-time contextual information (agent, basket, location etc) with real-time information fetched from customer profiles and history. These could be cross-sell, up-sell, loyalty offers, next best action etc. Responses are captured and fed back into the system for reporting, self-learning and optimizing future interactions both with the specific customer and with similar customers in similar contexts.

Like all such systems this combines some offline and online activities. Offline predictive analytics are used to process large amounts of data into predictive analytic models(e.g. scan all shopping baskets to identify items that are likely to be bought together). Online context is combined with data and online predictive analytics to come up with an optimal offer or action while also capturing the response data for learning over time. SAP Real-Time Offer Management uses its own online analytic engine that learns as offers are presented and interacted with. It also allows all the constraints to be defined and other analytics to be used, either stored in the database or by calling external scoring service.

The newest version can use HANA as its database as well as pushing its self-learning processes into HANA for increased performance. It also has access to the HANA Predictive Analytics Library (reviewed here) for more analytic algorithms. This use of HANA allows more data to be accessed in real-time. Over time more of the solution will run in HANA and take advantage of the HANA infrastructure and real time performance.

SAP Real-Time Offer Management also supports offers with location – a place attribute – and this location/range can be used as part of selecting offers. This can be used, for instance, in different mobile scenarios such as on consumers’ mobile device or in mobile terminals embedded in cars and GPS devices. Finally it can also be integrated both with batch campaign allocations AND with a real-time offer approach and can balance the two for a specific moment of interaction.

For more information please consult the solution description at sap.com

A new week and a new white paper. This week’s is Operationalizing Analytics

The most powerful examples of analytic success use Decision Management to deploy analytic insight into day to day operations helping organizations make more profitable operational decisions. This paper sponsored by SAS discusses both how to use a focus on decisions to ensure the right problem gets solved and what such an analytic technology platform looks like.

You can find this white paper, and many others, on our white paper resources page.

While at IBM’s IMPACT I recorded an Ignite presentation. For those of you who don’t know, the idea behind Ignite presentations is that you get 5 minutes (exactly) and 20 slides that move every 15 seconds to make one point. I presented on Why Model Decisions? and the result is shown below. Decision requirements modeling is great whether you are developing analytics or implementing business rules (or doing both together).

If you are interested in modeling decision requirements, check out decisionsfirst.com too.

Some of the slides got a little distorted but I posted them to slideshare too:

As some of you know I am an IBM Champion for my work on Decision Management. As part of this year’s IMPACT event they interviewed a number of us and made this cute composite video. Enjoy

Last session for me at FICO World is SulAmerica (part of ING) talking about their work with FICO moving decision management applications to the cloud. Cloud is clearly a growth area for businesses, with much of companies’ IT  investment being pushed to the cloud. This move to the cloud turns applications to SaaS, middleware into PaaS and hardware to IaaS – a new cloud stack. This new stack changes the dynamics of solution development: Usage-based models lower barriers to end, no install and no hardware means faster time to value and rapid scaling – all without needing IT and with less complexity.  Adoption of this new stack is rapid and accelerating. So what does this mean for decision making? Well decision management is ideally suited to the cloud – once we decide to manage decisions separately and we focus on delivering decisions to multiple systems then delivering decisions as a service is both powerful and clearly valuable (see my research on Predictive Analytics in the Cloud, for instance).

SulAmerica is the largest independent insurance group in Brazil with over 30,000 independent agents, $5B in revenue and nearly 7M clients. Insurance in Brazil is sold through independent agents or brokers and customers only contact the insurance company directly when making a claim or otherwise needing help. In 2005 SulAmerica licensed Blaze Advisor and developed an auto quote solution to replace 7 existing systems written in 3 different languages that had to be updated every month. By 2008 they had expanded from auto (one of their smaller lines of business) and had developed a centralized decisioning engine for multiple lines. More consistent decisions, increased agility and reduce maintenance costs.

SulAmerica is looking at cloud in three scenarios:

• When  a service needs to be turned on and off periodically
• When there is a need to add new capacity over time
• When there are occaisional sudden peaks in demand for additional capacity
• When there are predictable peaks and troughs in the demand for a service

All of these show clear value propositions from the scalability and flexibility of cloud provisioning. SulAmerica moved to SaaS in 2011,Paas in 2012 and IaaS in 2013. They deployed their health and dental quoting application to the Google App Engine (PaaS) and used Azure (IaaS) to provide a web downloadable auto quoting engine (internet connectivity is such that being able to download something locally can be important).

The health and dental insurance underwriting application was written using Blaze Advisor, targeted at small and medium businesses and delivered on the Google App Engine. The decision service replaced an Excel-based approach that was horribly hard to maintain and change as regulations and policies changed. The new application as able to match the on-premise performance and be delivered with the elasticity and scalability you would expect from a cloud solution. From a user interface perspective the business users continue to use Excel for maintaining the business rules. These spreadsheets are uploaded to update the business rules, minimizing the process change within SulAmerica. The brokers come to a web portal or a mobile application to get quotes.

The new system allowed for more rapid updates and helped the business grow 30% while reducing $2M in IT costs. The cloud-based performance and availability was actually better than for the on-premise version, critical because the agents are asking for several quotes and the faster, more reliable a response the better for the insurance company.

That’s a wrap.