Analytic Insight: It's All About People and Processes

If you don't understand your business processes and don't have talented staff, even best-in-class analytics can't save you.

It's inevitable: in the midst of economic uncertainty, enterprises hunker down.

They curtail -- or even slash -- operating budgets, constrain or eliminate discretionary spending; cut spending on R&D, services, and support; postpone cancel pending projects; and scale back staffing levels (and the free coffee and bagels in the break room).

In short, they do anything and everything they can to cut costs.

This doesn't mean that they'll cease spending entirely, of course. In the current climate, many business intelligence (BI) vendors predict, enterprises will almost certainly increase their spending on BI and data warehousing (DW) technologies (see

With the economy looking increasingly troublesome and some firms hard-pressed to endure a deep and lasting recession, companies will focus less on BI and DW in general and on big-bang-for-the-buck analsytic technologies.

It's a tendentious proposition. Although few dispute the ROI potential of analytic technology -- particularly that of high-value solutions such as predictive analytics -- few believe that such technologies, in and of themselves, can deliver instant positive returns to prospective users. SPSS, for its part, contents that predictive analytics software and solutions can deliver positive returns quickly, usually within the first 6 to 12 months.

The building blocks of (and biggest impediments to) analytic success, proponents say, are the same people and process issues that frustrate all would-be transformative efforts. You must understand your business processes. You must have done the infrastructure heavy lifting to optimize those processes (by delivering timely connectivity to operational systems or by ensuring the quality of the data that you're feeding into your analytic pipeline). You must have creative, talented folks on staff. Without these elements, even best-in-class analytics (with the most elegant data models or smartest algorithms) won't help you.

"Our point of view is grounded in convergence -- the convergence of analytics, business processes, and the IT architecture. No longer can we approach our customers by only focusing on analytics; we have to broaden our view a bit and look at the bigger picture -- [such as] what is the existing business process that we're looking to optimize, or how can we integrate predictive analytics into the IT infrastructure that supports the business process," says Kris Hackney, vice president of enterprise solutions with statistical analysis powerhouse SPSS Inc.

"For our customers, that means SPSS is making investments in smart people, people with enterprise architecture experience, [and] people who understand Web infrastructure."

It isn't surprising that the players that tend to be most enthusiastic about high-value analytics are also the players with the most credible analytic portfolios -- namely, vendors such as SAS Institute Inc. and SPSS Inc. Even so, officials caution against a certain kind of irrational exuberance when it comes to the efficacy of analytic investments.

"It starts with the data itself. You're always measuring things. Stepping back, you need to look at what you're measuring. Are you measuring what matters most? What else could you measure? Where do you perhaps have a blind spot?" says Ann Milley, senior director of analytics strategy with SAS.

It's tempting, Milley says, to sift through everything. There's a finite amount of what she calls "analytic bandwidth," however. "There's a lot of data out there, so … what it comes down to is a case of analytic bandwidth. There's a shortage of analytic talent that can really help you tap into the value of your data. If you're measuring everything -- if you're not trying to understand what you're measuring and prioritize it accordingly -- you're going to saturate that bandwidth."

Hackney, too, picks up on the issue of analytic bandwidth. It isn't that organizations don't have analytic know-how in-house, she stresses. Instead, it's an issue of making the most of that know-how -- i.e., avoiding oversaturation.

"One of the things we see … is that there isn't a plethora of analytic expertise, particularly around data mining and statistics, sitting around in companies," she indicates. "There may be people who can write a neat select statement, but that doesn't mean they understand predictive analytics and data mining."

Analytic talent is a rare and pricey commodity -- one that tends to be difficult to attract (and considerably more difficult to retain) during times of economic tumult. The paradox of economic hardship, as Gartner Inc. and other industry watchers like to point out, is that companies that have the means will in many cases pay more to acquire and retain top IT talent.

In cases where there's too little analytic bandwidth -- or a scarcity of data mining or predictive analytic expertise -- Milley's suggestion is to outsource. (SAS itself maintains a not-inconsiderable analytic services practice.)

"Sometimes it's a good thing to outsource. Bring in a fresh perspective and a new set of eyes. Most of our Global 2000 [customers] typically have some analytic expertise, but you've still got departments where analytics isn't as widely adopted," she says. "Most companies know that they've got to do … something about the problem that they want to solve. If it's just a matter of analytic bandwidth, it might be time to say, 'Let's go outside of our organization and tap into someone else.' It might be temporary, but if we get good results, it might be project after project."

There's much to be said for a fresh perspective, she argues. "We get cases where [customers have] brilliant model builders but someone else will say, 'You know what? This is the way we've been doing it for a long time. We just want you to come in and take a look at our model and our approach and tell us if there's anything we can do to make it better.'"

There's a further wrinkle here. According to SPSS' Hackney, one goal of any analytic investment should be to make it pervasive: to expose it to as many consumers as possible and to present it in such a way that it can meaningfully influence (and in many cases drive) decision-making. In this regard, then, analytic technology must above all be usable, either on a standalone basis (via a spreadsheet or analytic workbench, for example), or embedded in third-party software.

If there's one knock against SAS and SPSS it's usability. Competitors -- few (if any) having achieved analytic parity with the pair -- like to gleefully pigeonhole both companies as purveyors of tools for pointy-headed statisticians. Getting information out of SAS or SPSS is a black art, competitors argue.

Both companies vigorously dispute such claims. There might once have been some merit to such marketing, SPSS' Hackney concedes, but her company has taken aggressive steps to ratchet up the usability of both its SPSS suite (now in revision 17) and its Clementine predictive analytic workbench (see

She points out that SPSS has worked hard to hone its Web services, SOA, and interoperability chops. The upshot, Hackney claims, is that developers can both invoke analytic functionality from existing applications and -- in a growing number of cases (e.g., call center applications) -- tap predictive features to guide a user (via dialog sequences) and control the application itself. The upshot, she says, is that many people who consume analytic insights aren't actually interacting with a traditional analytic toolset. In the future, she suggests, this scenario will become even more common.

"One area that we've paid a lot of attention to is how we deploy analytics or guidance around a decision, or how we integrate [analytics] around a business process. That means embedding analytics as part of the process" she asserts.

"If I am a call center agent for an insurance company and I have to walk a customer around a series of questions to help them determine how to handle their claim, I'm not focused on the predictive process that's going on in the background [that] drives those questions and determines where to route that data," Hackney continues. "A real focus for us has been how do we prepare an environment and build an environment where predictive analytics are being deployed seamlessly into where people already work?"