RESEARCH & RESOURCES

Analysis: Watson Analytics Advances New Paradigm of User Assistance

Does IBM's new Watson Analytics offering herald the advent of a new class of quasi-intelligent machines capable of beating the best of human analysts at their own game?

Does IBM Corp.'s new Watson Analytics offering herald the advent of a new class of quasi-intelligent machines (in the case of the original Watson challenging Jeopardy! champion Ken Jennings) capable of beating the best of human analysts at their own game?

The simple answer to this question is: not exactly. Putting aside the philosophical question of whether Watson is actually capable of thinking (no less a luminary than American analytic philosopher John Searle rejects this claim), both the original Watson and IBM's new Watson Analytics cloud service are designed for altogether different use cases.

The original Watson was designed for a very specific application -- to compete in a television game show that has a well-defined set of parameters. Watson Analytics, on the other hand, is designed for use in business, research, and other contexts by for-profit, public sector, and non-governmental organizations (NGO). In its initial incarnation, it's a cross-use case and cross-vertical offering, too: IBM isn't (yet) marketing a "Watson Analytics for Fraud Detection in Microlending," for example.

Therefore, Watson Analytics must be appropriate for a range of applications and contexts, each with a distinctly different set of parameters. The original Watson was an antagonistic technology, designed primarily for competition, and reliably preempted, frustrated, and occasionally confounded human opponents Ken Jennings and Brad Rutter. Watson Analytics is an assisting technology: it's there to help, to guide, to enrich, and -- above all -- not to get in the way.

"It is truly combining the cognitive capability [of the original Watson], even some of the natural language capabilities, with the visual [self-service] capabilities," says Nancy Kopp-Hensley, director of strategy and marketing for IBM's database systems. "Watson Analytics is really made to guide your analysis. It offers a completely new kind of 'guidance' -- it's [the guidance] you get with self-service but with the natural language and the cognitive features, too. No one else is doing this."

Many analytics solutions are happiest when consuming or working against data from relational or "structured" (tabular, name-value pair, etc.) sources. Watson Analytics bundles text analytic technologies and a host of algorithmic functions that (in combination with its NLP capabilities) permit it to meaningfully manage, analyze, integrate, and present (i.e., discover and synthesize) information from plaintext, XML, files, and other non-traditional sources. It isn't so much that the ceiling is somehow higher with Watson Analytics -- it's that (when it comes to getting data out of non-traditional sources) IBM already built a significant part of the scaffolding.

On the other hand, IBM may be hard-pressed to match the intelligence and discrimination of the user experience of a visual analytics tool such as Tableau, which, as Cindi Howson, a principal with ASK Consulting and a recognized (hand's-on) BI tools expert, noted "combine[s] entertainment with expertise in visual perception." There's a sense, too, in which Watson's combination of visual data discovery with guided self-service features -- or, for that matter, with NLP -- isn't entirely new. Several of IBM's competitors (such as Oracle Corp. and SAP AG) tout similar capabilities, and specialty vendors (such as Information Builders Inc. and Neutrino Concepts Ltd.) market NLP-based search and discovery offerings.

This is to say nothing of the glut of visual data discovery tools -- from IBM's own Cognos Insight to Microsoft Corp.'s PowerView to MicroStrategy Corp.'s Visual Intelligence, to offerings from SAP, SAS, and others -- that have flooded the market over the last 30 months. That's the point: vendors are chasing Tableau as well as self-service pioneer Qlik Inc. (which recently introduced a new HTML5-based discovery app, "Qlik Sense," to complement its mature QlikView product) and one-time data viz best-of-breed TIBCO Spotfire.

Is anybody catching up? Howson isn't sure. "In the last few years," she wrote in her TDWI recap of Tableau's recent user and analyst conference, "BI platform vendors have been adding visual data discovery capabilities to their portfolios, but it is hard to hit all three high notes equally well, and Tableau has had an eight-year head start."

Watson Analytics is IBM's biggest (or most explicit) business intelligence (BI) cloud play, too. Big Blue staked out a credible leadership position with respect to cloud in toto. In 2008, IBM invested almost $400 million in cloud computing infrastructure development. Big Blue also led aggressively with cloud hosting (what would today be called Platform-as-a-Service, or PaaS) services for its mainframe and Unix platforms.

The company, like other BI players, had been slow to introduce BI-in-the-cloud services. There are many good reasons for this, owing chiefly to the performance characteristics of analytic workloads, which are write-intensive, as distinct to read-intensive. (Reading from multiple, virtualized, often physically separate resources simply doesn't entail the consistency and reliability problems -- particularly for transactional integrity -- that writing to such resources does. The shift to in-memory computing, combined with the cost-economics and significantly increased density of physical memory modules, helps to change this.)

Tableau and other vendors have distinct leads in this regard, too. One such vendor is MicroStrategy, which -- notwithstanding the pioneering example of software-as-a-service BI pioneer LucidEra -- was arguably the first BI vendor to effectively leverage the strengths of the cloud paradigm, as distinct from simply translating an on-premises experience into the cloud.

On the whole, BI-in-the-cloud is comparatively less mature than is (for example) customer relationship management (CRM) in the cloud. Watson Analytics is a cloud-only play. IBM isn't offering it an an on-premises version. The practical effect, argues Kopp-Hensley, is that IBM is focused on what she calls "consumability" -- making it so people in a range of contexts can consume information from Watson Analytics and/or its constitutive services.

Part and parcel with this is the capacity to embed cloud services in operational applications -- be they in the cloud or on-premises, Hensley argues. "When we announced Watson Analytics, we gave a preview to say, 'We're trying to take all of these services and make them much more consumable, in the cloud.' Right now, you can buy all of these things a la carte, but as you move to the cloud, you'll want to take these services to be much more "composable" -- you'll want to be able to mash them up with other [cloud and/or on-premises] applications [and] with line-of-business [applications]," she says.

"All of these services are going to be built together with the goal of changing the experience of analytics in the cloud," Kopp-Hensley concludes.

TDWI Membership

Get immediate access to training discounts, video library, research, and more.

Find the right level of Membership for you.