By using website you agree to our use of cookies as described in our cookie policy. Learn More


Has Analytics become a Meaningless Term?

Some say the industry has dumbed down what's meant by analytics, treating the term as little more than a marketing tool.

Analytics isn't what it used to be. As some business intelligence (BI) and decision-support veterans see it, the term itself has lost some of its meaning. Just how much is a subject of some debate.

In business law, it's called trademark dilution: i.e., what happens when a competitor uses a trademark to lessen its uniqueness, or to break its association with a specific product or use-case. Poet A.R. Ammons called it something else. "A word too much repeated," Ammons wrote, "falls out of being."

This, critics allege, is what's happening with analytics.

"The term analytics has become so overloaded in this industry that it's ridiculous," Dr. Robert Brammer, Ph.D., told BI This Week at this year's Strata + Hadoop World conference.

Brammer was speaking on behalf of object database company Versant Inc., for which he serves as an "independent director." Before joining Versant, however, Brammer served as CTO of Northrop Grumman Information Systems. He's also a principal with Brammer Technology, a Boston-based information management consultancy. In his capacity as CTO for a major defense contractor, Brammer says, he was asked to assess a good many putative "analytic" technologies.

"A lot of these companies talk about analytics in what I would regard as a ridiculously crude way," he points out. According to Brammer, the idea of a general-purpose analytic platform is fatuous. Any viable (or "non-crude") analytic platform has to be industry-specific. He cites industry- and domain-specific analytic packages from SAS Institute Inc. -- such as SAS' Fraud Framework for Insurance, or Integrated Merchandise Planning for Retail -- as representative cases in point. These packages bundle use-case-specific and industry-specific predictive models; industry-specific data quality routines; connectors to data sources that are (or once were) commonly used in certain verticals; and a passel of other optimized amenities.

"[T]he fact of the matter is that analytics is very dependent on which vertical markets you're trying to serve and what you're trying to do with them. What might be a very good analytical capability in one industry might be totally irrelevant in another," Brammer explains.

Marilyn Matz, CEO of streaming analytic database specialist Paradigm4, echoes Brammer's pessimism. In the database arena, the glut of self-styled "analytic databases" has created a backdrop of noise. As a result, Matz maintains, buyers are hard pressed to screen for viable signals.

"One of the problems we're having in terms of how we position [Paradigm4] is, how do we describe what we do? The word 'analytics' has been co-opted [to such an extent that] it's unhelpful. Everyone claims to be 'analytic.' We need something [i.e., a term or description] that explains why we're different," she says.

Unlike a traditional RDBMS, Paradigm4 -- which is based on industry luminary Michael Stonebraker's SciDB platform -- doesn't use or require tables; it's able to efficiently express spatial and time-series operators because (at the database level) it represents information as arrays and vectors.

"When you think about the kinds of operations that people want to do, there's the SQL analytics, but what people want to do these days is very large-scale principal component analysis ... and multivariate statistics. That's the basic math that underlies recommendation engines, clustering, regression, and risk-modeling," Matz continues. "This [math] also provides the underpinnings for a lot of the things people want to do with all of this [machine-generated and streaming data] they're collecting from all of these different sources."

Tim Boyd-Wilson, a data warehousing professional who first came to data management by way of statistics, says that he, too, is frustrated by what he believes to be the cavalier (or too-glib) use of the term "analytics."

"I'm an ex-statistician [and] researcher, who got into data warehousing the old-fashioned way. A long time ago ... as analysts [and/or] researchers, we needed data; the [IT] department couldn't supply it, and so we built our own warehouse. SAS was chosen as the software solution because of its advanced analytics," says Boyd-Wilson, a manager with a New Zealand government ministry.

Notwithstanding all the hype, not much has changed, Boyd-Wilson contends. "From my ... point of view, SAS is till the major supplier of industrial-strength analytics. I don't regard Cognos, Business Objects, or [Microsoft's] Office [i.e., Excel] as 'analytics.' The former two are good reporting tools -- significantly more elegant than SAS -- while the latter is fine for ad hoc exploration. However, those aren't analytics," he argues, conceding that IBM Corp. (which markets Cognos, but which also acquired the former SPSS Inc.), Oracle Corp., and the open source R statistical and programming language, in some sense "do compete with SAS."

Analytics Everywhere

Glen Rabie, CEO of Yellowfin, a business intelligence (BI) multi-national based in Australia, has a different take on the issue. Unlike Paradigm4, Yellowfin doesn't purport to be an analytics player. More to the point, Rabie insists, part of Yellowfin's pitch is that it doesn't do analytics; it focuses on "mass-production" reporting, exposing -- if anything -- "lite" analytic capabilities.

Rabie concedes that critics of "analytic dilution" have a legitimate basis for their criticism, but argues that traditional BI vendors -- including not just IBM and Oracle, but also companies such as MicroStrategy Inc., QlikTech Inc., SAP AG, and Tableau Inc. -- have built sophisticated (and undeniably creditable) analytic capabilities into their tools.

"The term [analytics] itself has grown to encompass a lot of the things that were traditionally associated with BI," says Rabie. On this basis, he adds, "a lot of what we do [in Yellowfin] could be considered 'analytics,' but we don't promote it that way."

Rabie doesn't necessarily see this as a cynical or disingenuous move on the part of vendors, either. "Really, what's driving this is [that vendors are] responding to customers, to what customers say they want," he notes. "If you do a search on Google trends for 'analytics' versus 'business intelligence,' you'll find that analytics is growing relative to BI, which is dropping off, so in a sense, vendors are being forced to use that terminology to stay relevant."

The Category Problem

One problem with the idea of analytic dilution is that it frames analytics as a narrow and unchanging category. At the very least, this is problematic.

Consider the sport of football: is it a narrow and unchanging category?

In spite of the fact that it gets played at many different levels -- from "flag" to "touch" to Pop Warner; from middle school to high school to college; from semi-pro to professional -- football's legitimacy, even in the case of the most hopeless of Pop Warner teams, never gets called into question. It does get qualified, however: as adjectives, we understand the descriptors "touch," "Pop Warner," "NCAA," and "NFL" to name four very different kinds of football.

In other words, when we talk about "football," we mean a category with a wide range of skill and sophistication. We could say the same thing about analytics. This much is true, after all: from the "lite" analytics of the (increasingly ubiquitous) analytic dashboard to the predictive analytics of a vertical-specific fraud detection tool, we're talking about a single overarching category that spans a wide range of skill and sophistication.

On the other hand, analytic dilution involves a self-serving aspect -- a kind of opportunistic disingenuousness -- that isn't addressed by analogy to football. If a Pop Warner player compares himself to -- say -- Robert Griffin III of the Washington Redskins, you can bet that few folks, other than (perhaps) his parents, are going to take him seriously.

What about when a vendor uses analytics -- or attempts to cultivate the patina of analytic legitimacy -- to boost its product marketing? That's a more complicated problem.

A better comparison is the category "audiophile." Two decades ago, few folks could have said just what an "audiophile" was; as a category, it originally described an enthusiast of boutique audio equipment: i.e., someone who prized audio reproduction and who (often) spent lavishly on designer audio gear. Over the last half-decade, the term "audiophile" has become much less niche-y: nowadays, it encompasses both the high-school kid rocking a pair of "Beats" earbuds and the traveling executive sporting a pair of noise-cancelling designer headphones. Some of them -- yes, even the high-school headbanger -- might even self-identify as "audiophiles."

From the perspective of an audiophile purist, this seems like heresy; from the perspective of vendors and consumers alike, it's hardly controversial.

That's the point, says industry veteran Mark Madsen, a principal with consultancy Third Nature Inc. "The IT market has always had a problem with terminology because the vendors want to differentiate themselves from competitors. Any chance to sound different while doing the same thing they've always done is grabbed ahold of. The buyer is typically not knowledgable about the depth of the domain, and [so] takes claims at face value," he points out.

"Add to [this] … the desire by management to add some new capability and the claims of vendors that they now do this [new capability] … and you have a recipe for [the] rapid dissolution of any meaningful terminology."

TDWI Membership

Get immediate access to training discounts, video library, BI Teams, Skills, Budget Report, and more

Individual, Student, and Team memberships available.