Instrumenting Your Enterprise for Maximum Predictive Power

By James Kobielus

Business is all about placing bets and knowing if the odds are in your favor.

As I noted in my most recent Forrester report, business success depends on your company being able to visualize likely futures and take appropriate actions as soon as possible. You must be able to predict future scenarios well enough to prepare plans and deploy resources so that you can seize opportunities, neutralize threats, and mitigate risks.

Clearly, predictive analytics can play a pivotal role in the day-to-day operation of your business. It can help you focus strategy and continually tweak plans based on actual performance and likely future scenarios. And, as I noted in a recent Forrester blog post, the technology can sit at the core of your service-oriented architecture (SOA) strategy as you embed predictive logic deeply into data warehouses, business process management platforms, complex event processing streams, and operational applications.

The grand promise of predictive analytics—still largely unrealized in most companies—is that it will become ubiquitous, guiding all decisions, transactions, and applications. For the technology to rise to that challenge, organizations must move toward a comprehensive advanced analytics strategy that integrates data mining, content analytics, and in-database analytics. Already, we’ve sketched out a vision of “Service-Oriented Analytics,” under which you break down silos among data mining and content analytics initiatives and leverage these pooled resources across all business processes.

You may agree that this is the right vision but have doubt about whether there is a practical, incremental roadmap for taking your company in that direction.  In fact there is, and it starts with re-assessing the core of most companies’ predictive analytics capability: your data mining tools. As you plan your predictive analytics initiatives, you should avoid the traditional approach of focusing on tactical, bottom-up project-specific requirements. You should also try not to shoehorn your requirements into the limited feature set of whatever modeling tool you currently happen to use.

To become a fully predictive enterprise, you will need to take both a top-down and bottom-up approach to your data mining initiatives. From the top-down, it’s all about building and integrating alternate models of how your business environment is likely to evolve internally and externally. In our recent report on advanced analytics, Boris Evelson, Leslie Owens, and I sketched out the many business processes that can be enriched by predictive analytics.

So how do you instrument your company to become more predictive? For starters, assess whether your analytics tools support the following capabilities for developing, validating, and deploying predictive models:

• Model multiple business scenarios: You should be able to build complex models of multiple, linked business scenarios across different business, process, and subject-area domains, using such key features as strategy maps, ensemble modeling , and champion-challenger modeling.
• Incorporate multiple information types into models: You should be able to develop models against multiple information types, including unstructured content and real-time event streams, while leveraging state-of-the-art algorithm in sentiment analysis and social network analysis.
• Leverage multiple statistical algorithms and approaches in models: You should be able to develop models using the widest, most sophisticated range of statistical and mathematical algorithms and approaches, including regression, constraint-based optimization, neural networks, genetic algorithms, and support vector machines.
• Apply multiple metrics of model quality and fitness: You should be able to score and validate model quality using multiple metrics and approaches, including quality scores, lift charts, goodness-of-fit charts, comparative model evaluation, and auto best-model selection.
• Employ multiple variable discovery and assessment approaches: You should be able to build and validate models using various approaches for variable discovery, profiling, and selection, including decision trees, feature selection, clustering, association rules, affinity analysis, and outlier analysis.

How is this different from predictive analytics as usual? Traditionally, most predictive modeling specialists focus on the latter three capabilities: statistical algorithms and approaches, model quality and fitness, and variable discovery assessment. Most models are built in narrowly scoped business or subject domains—such as customer analytics for marketing campaign management—and only against structured data sources (such as relational tables). Traditionally, few predictive analytics projects have entailed modeling of multiple business scenarios across diverse domains--such as sales, marketing, customer service, manufacturing, and supply chain-- though in the real world these business processes are often quite interconnected. Also, many data mining initiatives fail to incorporate information from unstructured sources—such as text in call-center logs—though this content may be as important as what comes relational databases and other structured sources.

It’s very important to build multi-scenario predictive models against complex information sets, but becoming a fully predictive enterprise demands much more. To instrument your organization for maximum predictive power, you should also tool your advanced analytics to support the following capabilities:

·         DW-integrated data preparation: To speed up and standardize the most time-consuming predictive modeling project tasks, you should be able to leverage your existing data warehouse, extract transform load, data quality, and metadata tools to support a full range of data preparation features. These features include the ability to discover, acquire, capture, profile, sample, collect, collate, aggregate, deduplicate, transform, correct, augment, and load analytical data sets.

·         Deep application and middleware integration: To deliver models deeply into whatever heterogeneous SOA-enabled platform you happen to use, your predictive analytics tool should deploy on and/or integrate with a wide range of enterprise applications, middleware, operating platforms, and hardware substrate. You should be able to deploy models seamlessly into your data warehouse, business intelligence, online analytical processing, data integration, complex event processing, data quality, master data management, and business process management environments. And to play well in your SOA, your predictive modeling tools should support application programming interfaces, languages, tools, and approaches such as Web services, Java, C++, and Visual Studio, as well as emerging languages such as SQL-MapReduce and R.

·         Consistent cross-domain model governance: To avoid fostering an unmanageable glut of myriad models, your predictive analytics solution should support a wide range of tools, features, and interfaces to support life-cycle governance of models created in diverse tools. At the very least, your tools should enable model check in/check-out, change tracking, version control, and collaborative development and validation of models. To realize this promise, it should support a full range of tools, standards, and interfaces for import and embedding of models from other tools, as well as export and sharing of models to other environments.

·         Flexible model deployment: To execute modeling functions--such as data preparation, regression, and scoring—on the widest range of data warehouses and other platforms, your tools should support in-database or embedded analytics. And to scale to the max, your predictive analytics tools should deploy models to massively parallel data warehouses, software-as-a-service environments, and cloud computing fabrics.  Your advanced analytics tools should also support development of application logic in open frameworks—such as MapReduce and Hadoop—to enable convergence of data mining and content analytics in the cloud.

·         Rich interactive visualization: To deliver their precious payload—actionable intelligence—your advanced analytics tools should support interactive visualization of models, data, and results. Ideally, you should be able to visualize all of this in your preferred business intelligence tool, or in the predictive modeling vendor’s integrated visualization layer. Of course, you have every right to expect the full range of visualization techniques--histograms, box plots, heat maps, etc.—regardless of who provides the visualization layer.

As you can see, this goes well beyond data mining as usual. Forrester has a slightly different perspective on the development of the predictive analytics market than you’re likely to get from other sources. We see a robust, flexible, SOA-enabled data mining tools as the centerpiece of advanced analytics for fully predictive enterprises. The competitive stakes are too great for businesses to take the traditional silo-mired approach when implementing this mission-critical technology.

What do you think?

Service-Oriented Analytics: Tapping into the Predictive Smarts of Your Entire Organization

By James Kobielus

Advanced analytics is a key competitive weapon of companies everywhere. Visionary organizations are those that take a future-facing, analysis-driven perspective on new challenges. They do this by grounding management forecasts in solid historical information sets, leveraging and extending companies’ existing investments in data mining and predictive modeling.

To become truly future-focused, organizations must build out their predictive muscles through deepening commitment to these and other advanced analytics technologies, which also include interactive visualization, multivariate statistical analysis, text analytics, and massively parallel enterprise data warehousing. However, enterprises must be careful not to adhere to the common practice of implementing advanced analytics tools in tactical, application-specific silos. One of the downsides of this traditional practice is that diverse predictive modeling teams can find it difficult to share their deep domain expertise, best statistical approaches, and most powerful data exploration and visualization features.

How can companies realize the transformative potential of predictive analytics for business success? For starters, you must get rid of siloes that fragment your data mining initiatives into separate camps. You must also build a bridge between your data mining operations and the teams that manage your text analytics, business intelligence, complex event processing, and business process management efforts. And the key approach for silo-smashing is service-oriented architecture (SOA).

At first glance, SOA may seem like a foreign topic to many analytics professionals, but it shouldn’t be. In the broadest perspective, SOA refers to best practices for encouraging greater reuse, sharing, and cross-platform interoperability among key business resources. Typically, one associates SOA with reuse of one specific type of resource: application functionality that is distributed across heterogeneous, networked platforms. Nevertheless, key SOA principles—such as standards-based service virtualization, reuse, brokering, and governance—are as applicable to predictive models as to any other resource that lives online.

Predictive models empower your product managers, marketing specialists, risk managers, process analysts, senior executives, and other personnel with access to sophisticated forecasting, time-series analysis, and scenario-testing tools. Each model is a statistical encapsulation of your business’ current view of the future in some specific application, subject, or decision-support area. Incorporating the expertise of subject-matter experts, these models allow organizations to gauge the potential impact of future projects, campaigns, and other initiatives, and also to adjust execution of these initiatives in mid-stream.

Predictive analytics need not be a purely blue-sky planning tool. This technology can sit at the core of your SOA strategy, and leading-edge enterprises are in fact doing that. Companies in such verticals as financial and telecommunications are embedding predictive logic deeply into data warehouses, business process management (BPM) platforms, complex event processing (CEP) streams, and operational applications.

What will it take for your company to fully align your advanced analytics efforts with your SOA strategy? Most important, Service-Oriented Analytics requires an executive-level commitment to becoming a predictive enterprise on all levels. From the perspective of your existing predictive modeling teams, this will require an ongoing focus on collaboration across business, function, and subject domains. You should create a culture and offer incentives that encourage modeling professionals to reuse each other’s expertise on problems that cross multiple domains.

Reuse is everything. Here are some high-level guidelines for establishing a reuse-friendly Service-Oriented Analytics practice in your company:

• Reuse modeling best practices:  Start by consulting Forrester’s report on setting up a Business Intelligence Solution Center (BISC), which we defined as “an institutional steward, protector, and forum for BI best practices.” Given that predictive analytics is a key segment of BI, you will find it necessary to incorporate this technology into your BISC’s scope.
• Reuse modelers: Cultivate a professional cadre of predictive modeling experts who are more than just wizards in advanced statistics and mathematics. Encourage subject-matter experts in all business areas to undertake predictive modeling projects and to team with modeling experts in other projects, applications, and business units. Provide incentives for modelers to regularly move between business units and subject domains, thereby spreading their expertise throughout the enterprise.
• Reuse models: You should begin to investigate predictive-model governance tools, which support version control, check-in/check-out, and other controls over models created in and imported from diverse tools. Model governance tools—from vendors such as SAS, SPSS, and KXEN--facilitate reuse, consolidation, combination, and cross-synthesis among disparate models. And you should also investigate options for embedding predictive models within BI, BPM, and other operational applications, thereby leveraging your growing analytic asset into new deployments.

As you deploy predictive models into operational applications, you should provide other applications with SOA-based access to them through Web services, Web 2.0, and other standardized interfaces. In that way, you will be creating a critical bridge to your application development teams and be creating a thoroughly Service-Oriented Analytics environment.

Your first steps down this road are clear. Forrester will be happy to assist you in developing a detailed enterprise roadmap for Service-Oriented Analytics. In so doing, you’ll be better able to tap into the future-facing analytic expertise that exists throughout your business.

IBM Goes Deeply Predictive, Announces Acquisition of SPSS

By James Kobielus

IBM dropped a big bombshell at the start of any already action-packed day for the analyst community. At this moment, I’m sitting, along with several dozen of my peers from Forrester and other firms, at the IBM Smart Analytics System launch event in Hawthorne NY. I’ll blog on IBM’s other announcements in a separate items.

The bombshell was IBM’s announcement that it’s acquiring SPSS, a long-established, leading provider of predictive analytics (PA), data mining (DM), statistical analysis, and text mining tools. The acquisition, subject to the usual shareholder approvals and regulatory reviews, is expected to close later this year. But, even in advance of that near-certain consummation, IBM’s bold move has already sent shockwaves throughout the analytics market.

Most important, IBM has acquired the second largest vendor of PA/DM solutions, dwarfed only by privately held SAS Institute. In this segment, IBM’s proposed acquisition is having the same impact that its Cognos buy had on the business intelligence (BI) market two years ago. In many discussions with Forrester customers, SPSS is often mentioned as a key solution provider for predictive modeling and statistical analysis against structured, semi-structured, and unstructured content.

For IBM’s competitive standing in the data management market, this acquisition represents one of the last missing pieces of its Information On Demand (IOD) portfolio. By acquiring SPSS, IBM has acquired a substantial PA/DM brand with a very loyal set of longtime customers who have build their customer churn, supply chain optimization, and other predictive models on its best-of-breed platform. SPSS recently underlined its feature-comprehensive value proposition through re-branding around the “Predictive Analytics Software” (PASW) family name. But longtime customers didn’t need to be reminded, of course, that what used to be known as “Clementine” defines a functional high-bar in this solution segment.

IBM would probably be the first to admit that it took its focus off the PA/DM market over the past several years as it build out the BI, data warehousing (DW), and other pieces of its IOD portfolio. IBM had never really exited the PA/DM market, but casual observers might have thought otherwise. However, the vendor three years ago chose to de-emphasize its Intelligent Miner tools--which support mining of structured data--as stand-alone offerings. It essentially buried these solutions, moving them into its InfoSphere Balanced Warehouse family, where they are now offered as features of its Enterprise Edition DW software, rather than as stand-alone tools that would be enhanced and evolved independently.

IBM and SPSS’s respective customer bases should rest assured that overlaps among their respective product portfolios are not extensive. Once the acquisition closes, IBM is almost certain to build out its SPSS brand and, over the coming 1-2 years, phase out the Intelligent Miner technology within its InfoSphere portfolio. One tricky issue is which text analytics solution family--SPSS’ or IBM’s OmniFind solutions--will prevail as the parent converges these offerings in its IOD portfolio. Another issue is how IBM will integrate the SPSS offerings into its still-evolving in-database analytics roadmap for InfoSphere Balanced Warehouse. Hopefully, IBM will maintain and extend SPSS’ already extensive in-database analytics integration with a broad range of vendor DWs, including such Big Blue rivals as Oracle, Microsoft, Sybase, and Teradata.

Who loses from IBM’s acquisition of SPSS? Fundamentally, one can’t help think that SAP missed the boat by not seizing the opportunity to acquire partner SPSS, whose Clementine technology it OEMs, has integrated with its BI technology, and sells as SAP BusinessObjects Predictive Workbench. PA/DM is an increasingly key component of a full-fledged BI solution stack. However, the remaining field of vendors with stand-alone, horizontally applicable PA/DM vendors consists primarily of vendors who are large but proudly and stubbornly independent (especially, SAS Institute); high-quality but much less widely adopted (e.g., KXEN, ThinkAnalytics); or specialized on customer, financial, scientific, or other specialized analytics (e.g., Unica, Fair Isaac, Accelrys).

Some IBM rivals in the BI space already have strong PA/DM tools, most notably Oracle (Oracle Data Miner) and TIBCO/Spotfire (the Insightful tools). Among BI vendors, Microsoft, MicroStrategy, and Information Builders have PA/DM capabilities, but they are not to a SAS or SPSS level of sophistication. These and other BI vendors should also be scouting for strategic acquisitions.

What do you think? Will IBM’s acquisition of SPSS lead to further merger and acquisition activity in this space as other leading BI players strengthen their PA/DM solutions?

By James Kobielus

IBM dropped a big bombshell at the start of any already action-packed day for the analyst community. At this moment, I’m sitting, along with several dozen of my peers from Forrester and other firms, at the IBM Smart Analytics System launch event in Hawthorne NY. I’ll blog on IBM’s other announcements in a separate items.

The bombshell was IBM’s announcement that it’s acquiring SPSS, a long-established, leading provider of predictive analytics (PA), data mining (DM), statistical analysis, and text mining tools. The acquisition, subject to the usual shareholder approvals and regulatory reviews, is expected to close later this year. But, even in advance of that near-certain consummation, IBM’s bold move has already sent shockwaves throughout the analytics market.

Most important, IBM has acquired the second largest vendor of PA/DM solutions, dwarfed only by privately held SAS Institute. In this segment, IBM’s proposed acquisition is having the same impact that its Cognos buy had on the business intelligence (BI) market two years ago. In many discussions with Forrester customers, SPSS is often mentioned as a key solution provider for predictive modeling and statistical analysis against structured, semi-structured, and unstructured content.

For IBM’s competitive standing in the data management market, this acquisition represents one of the last missing pieces of its Information On Demand (IOD) portfolio. By acquiring SPSS, IBM has acquired a substantial PA/DM brand with a very loyal set of longtime customers who have build their customer churn, supply chain optimization, and other predictive models on its best-of-breed platform. SPSS recently underlined its feature-comprehensive value proposition through re-branding around the “Predictive Analytics Software” (PASW) family name. But longtime customers didn’t need to be reminded, of course, that what used to be known as “Clementine” defines a functional high-bar in this solution segment.

IBM would probably be the first to admit that it took its focus off the PA/DM market over the past several years as it build out the BI, data warehousing (DW), and other pieces of its IOD portfolio. IBM had never really exited the PA/DM market, but casual observers might have thought otherwise. However, the vendor three years ago chose to de-emphasize its Intelligent Miner tools--which support mining of structured data--as stand-alone offerings. It essentially buried these solutions, moving them into its InfoSphere Balanced Warehouse family, where they are now offered as features of its Enterprise Edition DW software, rather than as stand-alone tools that would be enhanced and evolved independently.

IBM and SPSS’s respective customer bases should rest assured that overlaps among their respective product portfolios are not extensive. Once the acquisition closes, IBM is almost certain to build out its SPSS brand and, over the coming 1-2 years, phase out the Intelligent Miner technology within its InfoSphere portfolio. One tricky issue is which text analytics solution family--SPSS’ or IBM’s OmniFind solutions--will prevail as the parent converges these offerings in its IOD portfolio. Another issue is how IBM will integrate the SPSS offerings into its still-evolving in-database analytics roadmap for InfoSphere Balanced Warehouse. Hopefully, IBM will maintain and extend SPSS’ already extensive in-database analytics integration with a broad range of vendor DWs, including such Big Blue rivals as Oracle, Microsoft, Sybase, and Teradata.

Who loses from IBM’s acquisition of SPSS? Fundamentally, one can’t help think that SAP missed the boat by not seizing the opportunity to acquire partner SPSS, whose Clementine technology it OEMs, has integrated with its BI technology, and sells as SAP BusinessObjects Predictive Workbench. PA/DM is an increasingly key component of a full-fledged BI solution stack. However, the remaining field of vendors with stand-alone, horizontally applicable PA/DM vendors consists primarily of vendors who are large but proudly and stubbornly independent (especially, SAS Institute); high-quality but much less widely adopted (e.g., KXEN, ThinkAnalytics); or specialized on customer, financial, scientific, or other specialized analytics (e.g., Unica, Fair Isaac, Accelrys).

Some IBM rivals in the BI space already have strong PA/DM tools, most notably Oracle (Oracle Data Miner) and TIBCO/Spotfire (the Insightful tools). Among BI vendors, Microsoft, MicroStrategy, and Information Builders have PA/DM capabilities, but they are not to a SAS or SPSS level of sophistication. These and other BI vendors should also be scouting for strategic acquisitions.

What do you think? Will IBM’s acquisition of SPSS lead to further merger and acquisition activity in this space as other leading BI players strengthen their PA/DM solutions?

BI Mashup Maturity Model? Oxymoron? Au Contraire Mon Frère!

By James Kobielus

In one of my recent tweets, I commented that Forrester has developed a maturity model for enterprise adoption of mashup-style, self-service development of business intelligence (BI) applications. Indeed, we have, and it will appear in my forthcoming Forrester report, “Mighty Mashups: Do-It-Yourself Business Intelligence for the New Economy.”

Another tweeter--an astute, but sadly, non-Forrester BI analyst--scoffed that “BI mashup maturity model” is an oxymoron. Respectfully, I must disagree. Enterprises are adopting self-service BI approaches for many reasons--principally, to cut costs in a tight economy, to unclog the development backlog, and to speed delivery of actionable, targeted intelligence to decision makers. Also, companies are providing users with BI tools to do interactive, deeply dimensional exploration of information pulled from enterprise data warehouses (EDW), marts, cubes, transactional applications, and other systems. Furthermore, organizations everywhere have adopted browser-oriented BI environments that leverage the full Web 2.0 interactivity and collaboration.

Sitting at the convergence of those trends is BI mashup, which Forrester sees as the new paradigm for truly pervasive decision-support systems. What throws off some people is the term “mashup,” which sometimes gets pigeonholed as simply referring to using, say, Google Maps to display geocoded performance metrics and sundry Internet-sourced data in a browser-based dashboard. Yes, BI mashup encompasses that approach to presenting and integrating diverse data, but its application is much broader.

Just as important, BI mashup is not bleeding-edge. Rather, BI mashup leverages the in-memory BI clients, semantic virtualization layers, data federation middleware, automated data discovery, and other next-generation BI tools and platforms.

No one vendor or user has yet put together an end-to-end BI environment that is entirely focused on mashup-style self-service development. However, Forrester sees the BI industry converging toward as mashup-oriented architecture over the coming 2-3 years. With that in mind, we sketched out a BI maturity model that encompasses the following four levels (the first 3 of which are represented in case studies in the upcoming report):

• Level 1: Lightweight presentation mashup against transactional applications: This basic maturity level is for companies that have no prior BI or EDW; have little in-house BI expertise; and are comfortable with allowing casual users to use their browsers to customize parameterized reports from data from packaged business applications.                                                                
• Level 2: Deep presentation mashup against EDW: This level is for organization that do have prior BI and centralized EDWs, but have an understaffed BI development group and/or  power users and data modelers urgently require the ability to mashup and explore historical and current data within sophisticated BI workspaces.
• Level 3: Full BI mashup in federated environment: This level is for organizations that have decentralized, dynamic data management environments, and have the expertise to design reusable, composite data services to seamlessly mashup internal and external information.
• Level 4: Full collaborative mashup with IT governance: This level is for organizations that want to encourage subject  matter experts and operational users to collaborate on analytics created through mashup, but who are also concerned that all mashups be controlled, governed, and monitored in accordance with enterprise policies and best practices.

As I said, it will take a few years before we see a substantial number of enterprise case studies that implement the pinnacle of collaborative mashup with tight governance. Nevertheless, when you follow the evolution of next-generation solution portfolios from leading BI vendors such as SAP, IBM, Microsoft, and others, it’s clear that self-service user-centric mashup, to varying degrees, is a core theme.

BI mashup has such a strong business case that we’re confident it’s more than simply a “down economy” theme. It will almost certainly grow in importance for information and knowledge management professionals as the economy improves.

Database Religions Dissolve Into The Big Billowing Virtual Data Cloud

By James Kobielus

Virtualization is a venerable old computing concept that has achieved new life in recent years.

Virtualization brings to life a new world of more flexible service provisioning while cleverly emulating the old world that is being replaced. Virtualization refers to any approach that abstracts the external interface from the internal implementation of some service, functionality, or other resource.

The promise of virtualization is that, no matter how scattered and diverse, all pooled resources behave as if they were a single unified resource, both for usage and administration. In a sense, this is the practical magic that Arthur C. Clarke identified with advanced technology. The external interface may conceal various facts about the implementations of the underlying resources. The virtualized resources may:

•    run on diverse operating and application platforms;
•    have been deployed on nodes in diverse locations;
•    have been aggregated across diverse hosting platforms (or partitioned within a single hosting platform, either through virtual machine software, separate CPUs, or separate blade servers); and have been provisioned dynamically in response to a client request.

When Noel Yuhanna and I presented on enterprise database virtualization last week at Forrester IT Forum, we took pains to point out that is not a radically new paradigm. In fact, database administrators (DBAs) have been doing virtualization for a long time and not realizing it. We’re all familiar with such database virtualization approaches as policy-based server clustering, massive parallel processing database grids, and enterprise information integration. In these environments, you can identify the virtualization layer as “single system image,” “semantic abstraction,” or some other approach.

What all these approaches share is that they make two or more repositories behave as if they were a single database for unified access, query, reporting, predictive analytics, and other applications. If you wish, I could drill down further into the layers of database virtualization—data virtualization, transaction virtualization, and platform virtualization—but that would be too much for a mere blog post.

One twist that I didn’t have time to explore in depth last week is the notion that the traditional hub-and-spoke enterprise data warehousing (EDW) architecture is itself a form of database virtualization. The hub-and-spoke model transforms analytic data to a common “spoke-side” semantic access model, such as star schema or columnar. As such, this approach abstracts from the data models (usually 3NF relational) implemented at the EDW hub tier, the staging tier (perhaps file-based), and OLTP sources (perhaps hierarchical, XML, or what have you).

When you realize that each data-persistence approach has its optimal deployment sphere, you’re thinking database virtualization. At that point, you start to realize that the various database religions—relational is supreme, columnar is king, and so forth—are not absolute truths. They’re simply sectarian texts in a tradition of longer vintage: the evolution of truly all-encompassing data virtualization clouds.

Yes, I’m using “cloud” in this context because it best describes this new paradigm. Cloud-based virtualization is beginning to seep into analytic infrastructures. To support flexible mixed-workload analytics, the EDW, over the coming five to 10 years, will evolve into a virtualized, cloud-based, and supremely scalable distributed platform.

What are the outlines of this new paradigm? The virtualized EDW will allow data to be transparently persisted in diverse physical and logical formats to an abstract, seamless grid of interconnected memory and disk resources and to be delivered with sub-second delay to consuming applications. EDW application service levels will be ensured through an end-to-end, policy-driven, latency-agile, distributed-caching and dynamic query-optimization memory grid, within an information-as-a-service (IaaS) environment. Analytic applications will migrate to the EDW platform and leverage its full parallel-processing, partitioning, scalability, and optimization functionality. At the same time, DBAs will need to make sure that cloud-based DW offerings meet their organizations’ most stringent security, performance, availability, and other service-level requirements.

I won’t opine here and now on how much enterprise data will be persisted in public clouds vs. private environments that incorporate many of the same platform virtualization technologies. I’ll save that discussion for the upcoming Forrester reports that Noel and I are developing in virtualization of transactional and analytic databases, respectively.

Expect those in Q3 or thereabouts. Thanks everybody who attended our preso last week in Vegas!

Information Post-Discovery - Latest BI Trend

By Boris Evelson

I just came back from an exciting week in Orlando, FL, shuttling between SAP SAPPHIRE and IBM Cognos Forum conferences. Thank you, my friends at SAP and IBM for putting the two conferences right next to each other (time- and location-wise), and for saving me an extra trip!

Both conferences showed new and exciting products and both vendors are making great progress towards my vision of “next generation BI”: automated, pervasive, unified and limitless.  I track about 20 different trends under these four categories, but there’s a particular one that is especially catching my attention these days. It went largely under covers at both conferences, and I was struggling with how to verbalize it, until my good friend and peer, Mark Albala, of http://www.info-sight-partners.com, put it in excellent terms for me in an email earlier today: it’s all about “pre-discovery” vs. “post-discovery” of data.

We can debate endlessly the pros and cons of traditional row oriented RDBMS vs. newer DBMS architectures specifically designed for BI and OLAP (like columnar – Sybase IQ, Vertica, Paraccel,  inverted index – Microsoft FAST Search, Endeca, Attivio, tokenized – illuminate, and in-memory analytical DBMS – TIBCO Spotfire, QlikTech) – and I had lots of fun doing that on the DM Radio show last Thursday!  One thing, however, remains undisputed: traditional RDBMS and OLAP architectures require pre-discovery of data, aka data integration and data modeling. No matter how much flexibility and richness we think we built into out relational or multidimensional data models, they are still only as good as our initial design. If we did not anticipate the types of questions that would be asked of our application in the future, no fixed relational or multidimensional data models will be able to help us.

But the world moves way too fast on us. For example, the methodology behind economic capital calculation in the financial services industry, according to Basel II requirements, may change on a weekly, sometimes even on a daily, basis due to regulatory and competitive pressures. No traditional data models and BI tools can keep up with such furiously quickly changing requirements. As a result, one of our recent surveys found that more than half of the respondents did not have most of the information they were looking for in their BI applications, and close to two thirds relied on IT for new BI requests.

What’s the answer? There are many, but one partial answer is post-discovery, rather than pre-discovery of data. For example, an inverted index DBMS from Attivio, or a tokenized data store from illuminate, or in-memory models from TIBCO Spotfire and QlikTech just need you to index or loading data "as is", not really requiring any modeling up front. And because all of these technologies can indeed cross reference every attribute with every other attribute (it's an index!), a virtual data model is created on the fly simply by virtue of asking it a question. Gone are the days of having to analyze your requirements, document your requests, work with IT to make it happen - a process that often takes weeks or months.

Sounds good? It does, but this is obviously not a BI panacea. Yet. I do not think we will see a mass conversion to these analytical engines that allow for post-discovery of information and data models in the short term. Why?

• Many challenges still remain with these technologies such as lack of operational BI (you typically need to reload entire model or rebuild entire index to make updates), administration (partitioning, and modular backup and restore are not easy tasks with index DBMS today), and other mission critical, production features of large enterprise DBMS.
• All these new technologies still rely on someone else doing all of the upfront leg work to integrate, reconcile and cleanse the data. There is no magical work around the hard work of planning, designing and implementing data quality and master data management processes and applications.

How’s all this related to SAP and IBM, you ask? Simple.

• SAP is heavily promoting its guided searchengine – Explorer – previously known as Polestar. It’s a similar index to Endeca and Microsoft FAST Search, capable of allowing BI users discovering answers to previously unplanned questions. Unlike Attivio, though, Explorer still requires an underlying data model, either in the form of Business Objects Universe or SAP BW, but because under the covers it uses the SAP TRex index engine, the right technology is there, and it’s a huge step in the right direction for SAP.
• IBM is also leading the market here. While its Cognos GoSearch product allows for some guided exploration too, it’s the acquisition of Exeros a couple of weeks ago that gives IBM unique capabilities (with some competition from Composite Software and Sypherlink) to post-discoverdata, data relationships and continuously update data models and data content based on the newly discovered information. I can't wait till IBM comes up with a fully automated way to update the data model, Cognos Framework Manager, and Cognos reports and dashboards based on a source system change!

This is yet another proof – and I’ll never get tired of saying this – that BI market is as vibrant, exciting and far from commoditization as ever!

Oracle’s Sun Acquisition Accelerates Push Into Data Warehousing Appliances

By James Kobielus

Last fall, Oracle CEO Larry Ellison announced that his company was getting into the hardware business, but I think he misspoke. At that time, he was referring to the new HP Oracle Database Machine with Exadata Storage, a high-end data warehousing (DW) appliance that incorporated hardware from his partner, as well as intelligent storage software technology from that partner--and even had the partner’s name first in the product name. If that was the criterion for “getting into the hardware business”--i.e., running on someone else’s hardware--then every software vendor on earth is in the hardware business, by my reckoning.

But today’s Oracle announcement is the real deal. Oracle is acquiring longtime partner Sun Microsystems, putting the software powerhouse fully into the hardware business--and hitting the DW industry like an earthquake. I’ll let my Forrester colleagues blog on the other implications of this deal--for the open source, Java, middleware, SOA, and other markets that Sun is in--and give you a few quick thoughts on the deal’s implications for the DW market.

For starters, this deal will give Oracle the ability to engineer a completely integrated DW appliance composed of all Oracle components, including hardware and software. Now Oracle will be able to take on Teradata and IBM--both of which have long offered their own integrated solutions--more aggressively with high-performance DW offerings. Just as important, Oracle will be able to leverage Sun’s manufacturing scale economies to bring its all-Oracle DW appliances below the $25K-per-terabyte threshold needed for penetration into the midmarket.

Also, Oracle will now have another widely adopted transactional database, the open-source MySQL, that it can--and should--consider tweaking and packaging on an DW appliance. To the extent that Oracle gives customers a choice of DBMSs on a DW appliance platform, it can gain a differentiator that Teradata, IBM, Microsoft, Sybase, and Netezza lack (you have to go to a startup such as Dataupia for multi-DBMS choice on an appliance). Many information managers prefer to stick with their existing DBMSs when building a DW, and prefer to implement that DW on an appliance to take advantage of its out-of-box balanced configuration of CPU, memory, storage, and I/O.

Furthermore, Oracle is acquiring a hardware and operating system vendor that has long been one of the primary platforms on which its own DW/DBMSs, middleware, and tools have been deployed. This acquisition can only be welcome news for joint Oracle-Sun DW customers who have worried about Sun’s solvency for some time now and began to sweat serious bullets when IBM failed to emerge as a white knight. For many Sun customers, an Oracle-powered DW platform will now look like a safer bet than ever.

Of course, there are clear risks in this pending acquisition.

First, a combined Oracle/Sun sows uncertainty among the DW appliance vendors--such as Greenplum and ParAccel--who have partnered with Sun and now find themselves in earnest “co-opetition” with full-competitor (and then some) Oracle.

Second, Oracle’s other DW appliance hardware partners--including HP, IBM, and EMC/Dell--must be concerned that Oracle will now shift focus away from their respective appliance products in favor of those it builds with its own Sun hardware group.

And finally, Oracle’s acquisition of Sun--and possible future development of a MySQL DW appliance--may discourage customers from considering third-party DW appliances, such as from Kickfire--that build on MySQL. If that happens, and a market for non-Oracle-branded MySQL DW appliances never takes root, Oracle will be denying its MySQL customers the choice that Oracle Database customers already enjoy. Currently, Oracle Optimzed Warehouse customers can deploy that enterprise DBMS as a DW on their choice of Sun, HP, IBM, and EMC/Dell platforms.

Let’s hope that Oracle makes the most of its pending Sun acquisition. Ellison either misspoke last fall, or was speaking prophecy. Like most DW vendors, Oracle’s destiny is to grow ever more hardware-dependent for its long-term scalability, performance, and optimization story.

Inmon’s Vitriolic Slap At “Virtual Data Warehousing” Does Not Withstand Scrutiny

By James Kobielus

In a recent article, Bill Inmon incinerates a straw man concept that he refers to as “virtual data warehousing (DW).” For those unfamiliar with Inmon, he is generally considered the founder of DW as a data management discipline, has been at it since the 70s, and has more published books and articles to his name than most mortals. So he clearly may be considered an authority on the topic of DW.

But methinks Mr. Inmon doth protest too much on this “virtual DW” bugaboo, however defined (we’ll get to that in a moment). Also, he attacks this concocted notion with such emotional vehemence that it’s clear he considers it a threat to the centralized EDW paradigm upon which he has built his career and reputation.

For starters, his definition of this concept is oddly vague and questionably narrow: “a virtual data warehouse occurs when a query runs around to a lot of databases and does a distributed query.” Essentially, Inmon defines “virtual DW” as the ability to a) farm out a query to be serviced in parallel by two or more distributed databases, b) aggregate and join results from those databases, and c) deliver a unified result set to the requester.

That’s an important query pattern, but not the only one that should be supported under (pick your quasi-synonym) data federation, data virtualization, or enterprise information integration (EII) architectures. Inmon’s definition excludes the many federated queries that may only hit on a single database, with no joins and results aggregation, and with the EII fabric handling the necessary on-demand transformation from that source’s schema to an abstract semantic model.

Per my data federation report from last fall, Forrester has a broader perspective on the topic than does Mr. Inmon. Data federation is any on-demand approach that queries information objects from one or more sources; applies various integration functions to the results; maps the results to a source-agnostic semantic-abstraction model; and delivers the results to requesters. Nothing in the scoping of data federation necessarily requires the multi-source aggregation and joining that Inmon puts at the heart of “virtual DW.”

Putting Inmon’s narrow scoping of “virtual DW” behind us for the moment, let’s consider his chief objections to this approach. First, it requires the “analyst to integrate data” (as if that’s something analysts are ill-suited for or regard as some inordinate burden). Second, it consumes resources, experiences suboptimal performance, and “shuffles a lot of data around the system that otherwise would not need to be moved” (as if centralized DWs don’t consume resources, experience performance bottlenecks, and move data). Third, it is “limited to the [historical] data found in the [source] databases.” Fourth, it suffers from “no reconcilability of data...[hence] no single version of the truth for the corporation.”

It’s a fairly straightforward matter to dispatch these objections:

First, data integration--through ETL, EII, and other approaches--is a core job function for DW professionals, not some alien function outside their core competency.

Second, data federation is often the optimal approach for low-latency BI (just check out the case studies in my data federation and really urgent analytics reports). Federated environments can be tuned to provide top-notch performance and minimize source-system impacts when “shuffling” data around in a decentralized fabric.

Third, the source databases in a federation environment often include DWs, which, per their core function, usually manage a considerable amount of historical data. Once again, see my data federation report with discussion of case studies for a) Federation of Local DWs via Centralized EII Infrastructure and b) Federation of Dispersed EDW and ODS Data Into Siloed BI Environments.

Fourth, data federation is not totally incompatible with data reconciliation. In fact, federation environments can be architected for single version of the truth, data governance, and master data management. However, it can indeed be tricky to manage data quality in federated environments (see Rob Karel’s coverage of MDM and DQ for a deep dive on that issue).

My basic objection to Inmon’s line of discussion is that he treats data federation as mutually exclusive from the enterprise DW (EDW), when in fact they are highly complementary approaches, not just in theory but in real-world deployments. Yes, data federation can be deployed as an alternative to traditional EDWs, providing direct interactive access to online transactional processing (OLTP) data stores. However, data federation can also coexist with, extend, virtualize, and enrich EDWs, as well as other data-persistence nodes such operational data stores (ODS) and online analytical processing (OLAP) data marts. The case studies in the cited reports bear that out.

Inmon’s arguments are worth consideration. The centralized EDW model he touts is useful for illuminating some traditional best practices. But by no means can it do justice to the stubbornly heterogeneous, distributed, mixed-latency BI and DW requirements of most enterprises.

After So Many Years Of Ballyhoo, Semantic Web Still Searching For Killer App

By James Kobielus

Cynics might call Semantic Web a technology looking for a solution. And they might have a point.

Semantic Web refers to a long-running World Wide Web Consortium (W3C) initiative that is working toward an ambitious--some might say hopelessly Utopian--goal. At heart, it is a vision for how the World Wide Web should evolve to realize its full interoperability potential.

People vary widely in how they interpret the scope of the Semantic Web initiative. The tech industries are swarming with a wide range of projects, products, and tools that implement different variants of this vision. What vision is that? In the broadest sense, Semantic Web refers to an all-encompassing metadata, description, and policy layer that can, potentially, support universal, automatic, comprehensive end-to-end interoperability across every macro or micro entity—including data, components, services, applications, and services—on every conceivable level.

Whew!!! If that’s not the working definition of “pie in the sky” or “boil the ocean” (pick your metaphor), I don’t know what is. In fact, I’m hard-pressed to refer to Semantic Web as a definable market or solution segment. However, it’s not entirely vacuous.

For starters, organizations can implement W3C-developed semantic description standards—such as Resource Description Framework (RDF) and Web Ontology Language (OWL)--to make the meaning of content unambiguously comprehensible to services, applications, bots, and other automated components. Second, there is a reasonably robust market for “ontology” tools to support RDF/OWL-based modeling of application semantics. Finally, there is some incremental adoption of these tools and concepts in established IT segments, such as:

• Enterprise content management (ECM): Semantic approaches can support more powerful discovery, indexing, search, classification, commentary, and navigation across heterogeneous stores of unstructured and semi-structured content. Semantic search—driven by concepts, not mere text strings--is regarded by some as a primary Semantic Web application. Indeed, many Semantic Web vendors are primarily implementing the technology in search engines that leverage ontology-based concepts to improve search accuracy and reduce spurious hits.

• Enterprise information integration (EII):Semantic approaches enable consolidated viewing, query, and update of structured data that has been retrieved from diverse sources. Indeed, most commercial EII environments present an abstract semantic layer that mediates access to heterogeneous data, such as enterprise resource planning and customer relationship management applications, converging it all to a common presentation-side schema. A handful of those EII vendors have begun to support Semantic Web standards, primarily through third-party software plug-ins

• Enterprise service bus (ESB):Semantic approaches can facilitate multilayered application, process, and service interoperability across disparate environments. To date, there has been little production implementation of Semantic Web standards in the ESB arena, though some vendors have adopted semantics, ontologies, and RDF to describe the conceptual models implemented by application endpoints, agents, and intermediary nodes within ESB-like middleware approaches such as event stream processing.

But Semantic Web approaches are still on the periphery of these markets. 10+ years into its inception, Semantic Web still has no clear killer app. It’s not clear if or when that app will emerge.

Lean Information Management Strategies for Lean Times

By James Kobielus

When the going gets tough, the tough get lean, focused, and flexible. To help organizations survive the bad times and thrive in all climates, their information management initiatives must remain agile and adaptable.

If you feel your information management strategy is anything but lean, you’re not alone. Many organizations struggle to gain control over information infrastructures that have become too bloated, rigid, and slow to realign with new business drivers.

Lean information management practices are essential for corporate survival. They are far more than belt-tightening exercises. They also help you build analytic muscle for excelling in any business environment. Here are some basic pointers for keeping your information management strategy lean:

• Trim your information infrastructure of excess cost. Lean means you should cut excessive, budget-busting overhead from your information management environment. Careful cuts are best, because they optimize your existing operations without gutting the core information, analytics, and applications that underpin your core competencies. Silo, server, database, and application consolidation should be your principal approaches. Also, you should re-evaluate vendor-sourcing strategies and renegotiate licenses at more favorable terms. And you should investigate lower-cost alternatives, such as software-as-a-service, to address business intelligence, business performance solutions, enterprise data warehousing, master data management, enterprise content management, and other information management requirements.
• Fit information initiatives to key business imperatives. Lean also means you fit, focus, and fully align your information management initiatives to mission-critical business imperatives. Strategic alignment ensures that you leverage information assets across diverse application domains and business processes, rather than allow that intelligence to languish underutilized in silos. To sustain this approach, you should establish an information management framework, such as a Business Intelligence Solution Center, that enables ongoing collaboration between business and IT stakeholders. You should engage all key business and technical groups in information management planning discussions.
• Flex information architectures to changing circumstances. Finally, lean means maintaining an approach that is flexible and adaptable, able to shift course as your needs and environment change. In yoga terms, lean is all about building, toning, and stretching analytical muscle to keep it from tearing when you need to transition rapidly from one strategic alignment to the next. You need the flexibility to swing between centralized information management infrastructures and decentralized or federated environments. For end-to-end data management environments, Forrester has developed an architecture decision support tool that helps information managers to determine which of several topologies is best suited to their needs: centralized enterprise data warehouse, hub-and-spoke, independent data marts, data federation, and information-as-a-service.

Considered as a comprehensive strategy, these lean practices are true bloat-busters and recession-beaters. They allow organizations to deliver practical insights that address all pain points, even--especially!!!--within strict budgets.