In-Depth
SAP Takes the Predictive Analytic Plunge
With SAP Predictive Analysis, the company says it wants to bring predictive analytic capabilities to as many users as possible.
- By Stephen Swoyer
- 05/08/2012
SAP AG's pitch with its new SAP Predictive Analysis launch may sound like that of upstart rival Predixion Software. Like Predixion, SAP talks about putting predictive analytics into the hands of more users.
At one point, Jason Kuo, SAP's predictive analytic pointman, even trots out an "m" word: "many." SAP, Kuo says, wants to "bring predictive analysis capabilities to as many people as possible." Unlike Predixion, Kuo and SAP stop short of using that other "m" word -- "masses."
Kuo also adds a crucial caveat: "This tool will never be used by a general user."
In other words, he continues, don't even think about trying to get Grandma to interact (self-servingly) with predictive analytics. It's a nod to Predixion CEO Simon Arkell, who invoked his (proverbial) grandmother to describe Predixion's approach to pervasive analytics.
Kuo doesn't explicitly say Arkell's or Predixion's strategies aren't valid. He simply says that Predixion's way isn't SAP's way. At the same time, he argues, predictive analysis is the kind of problem that -- even in the age of big data, big insights, and (if futurist Ray Kurzweil is to be believed) nascent artificial intelligence -- can't be trivialized.
"Sure, grandma might be able to create the predictive model. That's super cool. But is it modeling the right thing? Is it getting data that's even relevant? Does it even have enough data? Is there enough variability? Is she implementing the algorithms correctly?" he wonders. "We're not naïve that it's as simple as just throwing a wizard on top of the product and giving it to users. It's going to take other innovative approaches."
A Not-so-Secret Weapon
SAP has been promoting in-memory analytics since before in-memory analytics was considered sexy. Back, that is, when in-memory analytics was merely a topic of interest.
So give it credit: in HANA, its in-memory analytic appliance, SAP isn't late to market. It has only been shipping for about 18 months, and -- while SAP likes to position HANA as just the thing to supercharge NetWeaver BW -- adoption of HANA is the subject of a great deal of speculation, as we'll discuss shortly.
Not surprisingly, SAP plans to leverage HANA to help promote SAP Predictive Analysis, too.
"The bigger story is the synergy with HANA. It increases the value [of SAP Predictive Analysis] by like an order of magnitude. Now you've got the big data screaming through the system, all in real-time, and processed in a way that ... has never been able to be done before [with predictive analysis]," he comments, noting that SAP added native support for data mining in its most recent HANA refresh. He compares this arrangement to that of analytics powerhouse SAS Institute Inc., which touts its ability to run natively in the context of Teradata, Netezza, and other engines.
There's an added wrinkle, insists Kuo, who contrasts SAP's position with that of rival IBM Corp., which acquired former predictive analytic best-of-breed SPSS Inc.
Thanks to its installed base, Kuo claims, SAP already "owns" a chunk of the operational data that's destined to serve as grist for the predictive analytic mill. (IBM, for the record, can likewise claim to have a line -- if not quite an inside track -- on this same data, by virtue of the extensive reach of its middleware and platform assets.) What, he asks, could be more natural than for users of R/3 and other SAP offerings to tap SAP Predictive Analysis to power their analytic practices?
"If you look at who owns the business data, it's SAP. That's where we feel the greatest opportunity for predictive analysis lies," Kuo argues. "The game has always been about how do you build a better statisticians' mousetrap. We feel we're in an amazing position to change that."
With the first release of SAP Predictive Analysis, the scope of SAP's better mousetrap is comparatively limited. SAP's ultimate goal with Predictive Analysis is to target business analysts, MBAs, and other power users. Its aim is to empower this user constituency to begin interacting with predictive analytics and to generate analytic insights that can drive effective decision-making. One practical effect of this scheme, claims Kuo, is that predictive analytic functionality will be pushed out beyond both of these core constituencies -- i.e., professional data analysts/statisticians and power users/business analysts -- to mainstream users, chiefly in the form of canned algorithms or analyses that can be embedded inside existing applications.
That's SAP's aim and hope. In the first rev of Predictive Analysis, Kuo concedes, it isn't quite there, which is to say that it isn't yet delivering an experience that takes many predictive analytics heavy-lifting (i.e., the creation and refinement of a predictive analytics model) out of the hands of professional data modelers. "In this release, we believe that the modern user interface and the quality of the visualizations, as well as the product's ability to shield the end user from the complexity of the algorithms ... we believe it's definitely more productive for the statistician, even than tools like SPSS," Kuo says.
"We believe it begins to allow the business analyst to use more predictive tools. Begins. We have internally many business analysts who are now using this tool," he continues. "In the future ... our plans are to have for example wizards and other types of options that make it even easier for the business analysts. This tool will never be used by a general user."
HANA, Predictive Analysis, and SAP's Long View
There's some question about SAP's tie-in strategy with HANA, too. Kuo, for example, describes the combination of HANA and SAP Predictive Analysis as a potent one-two punch, but he wasn't able to disclose just how widely or pervasively HANA is used inside of SAP environments.
SAP officials likewise declined to share information about HANA's usage.
Realistically, HANA uptake is probably still fairly limited. Although SAP has been talking about in-memory analytics for almost half a decade, SAP didn't start aggressively pitching HANA to its customers until last year. As a result, HANA deployments are still ramping up.
In a February posting on his DBMS2 blog, analytic database-watcher Curt Monash, a principal with Monash Research, wrote that "it's hard to sort out where SAP HANA has actually been used in production, and where people are just building something on HANA that they hope will work well." Monash concluded, however, that "the analytic case for SAP HANA seems decently substantiated -- there are years of experience with the technology and its antecedents, and column stores (including in-memory) are well-established for analytics."
In a posting responding to Oracle Corp. CEO Larry Ellison's downplaying of SAP's HANA strategy in Oracle's Q1 earnings conference call, Monash suggested that "SAP is still cherry-picking hard-core, committed SAP customers for its early HANA adoption."
SAP Predictive Analysis isn't a HANA-only proposition, Kuo stresses: SAP has another hot-shot analytic engine up its sleeve in Sybase IQ. Predictive Analysis will also run against relational repositories from IBM, Oracle, and Microsoft Corp., along with SAP's own BusinessObjects Universes and conventional flat files.