In-Depth

Both Sides Now to Business Intelligence

• By Stephen Swoyer
• September 1, 2005

Talking Points
ROI IS INTANGIBLE AND HARD TO GRASP

• BI describes nothing more than straightforward operational reporting running some kind of client application such as Microsoft's Excel spreadsheet against a relational database.
• BI today is a more expensive proposition than it was just a few years ago, when organizations frequently built their own reporting and analysis solutions.
• The biggest problems with shrink-wrapped BI are that it's expensive, and few vendors can claim to address all the pieces of the BI puzzle.

Profiting from business intelligence sounds like a no-brainer, doesn’t it? Profit, after all, is BI’s raison d’être. From the very beginning, BI has been pitched as a savvy, but evolutionary, way for organizations to squeeze more out of existing technology investments, revitalize mature profit centers and obtain actionable insights that enable the emergence of new ones. All the while paying for itself. That’s been BI’s charter for the better part of a decade now.

To a very real degree, most companies do profit from their BI investments. Few can cite ROI studies, and fewer still are willing or able to quantify ROI using the empirical metrics (such as revenue growth or income expansion) most palatable to CFOs. Many, however, can point to the revivification of established profit centers, or to cases in which BI investments have played an instrumental role in the identification of new profit-making activities.

Nevertheless, organizations tend overwhelmingly to find BI ROI in intangibles such as improved customer service, empowered knowledge workers and line-of-business customers, the delivery of competitively differentiated customer-facing products or services and the rapid identification of actionable business insights. (See related story, “In ROI we trust.”)

Technology innovation is a restless thing, however. With few exceptions, yesterday’s cutting-edge technology investment will over time become tomorrow’s de rigueur technology requirement. Just as one no longer—or rarely, in any case—speaks of profiting from ERP, fewer organizations are embracing BI because of its perceived ROI-generating abilities.

Instead, companies are doing BI because they can’t afford not to if they want to keep up with the Joneses, Wangs and Patels of the global economy, that is. “People talk a lot about ROI,” says Steve Snodgrass, CFO and interim CIO with Graniterock, a provider of stone and quarried materials. “I’m actually the company CFO, but I also have a background in IT, and I think it’s just hard to find an ROI on a lot of IT projects. You have to have them, really, if you’re going to remain competitive. ERP is one of those [ITprojects], and I think business intelligence is another.”

There’s another wrinkle, too. BI today is a more expensive proposition than it was just a few years ago, when organizations frequently built or cobbled together their own reporting and analysis solutions. For a number of reasons, companies are doing far less building and much more buying today. The consequences of this transition have also helped to muddy a once pellucid BI ROI picture.

Moving goalposts
How far have the goalposts moved in the BI ROI game? Graniterock’s Snodgrass cites his company’s own recent BI initiative—an RFID-powered information aggregation application based on Business Objects’ Crystal Enterprise reporting tool—as a textbook case of changing expectations.

Crystal Enterprise is a substantially more expensive product than the Crystal Reports software Graniterock used for half a decade. At the same time, Snodgrass says, ROI expectations, at least in terms of generally accepted accounting metrics, didn’t directly factor into the company’s project planning.

In fact, Graniterock didn’t expect to enhance its existing decision-making process—although the collection and analysis of RFID information has helped to do just that—nor did the company assume that its RFID jaunt would help identify new profit centers, although in time it will almost certainly do so.

Instead, Snodgrass says, Graniterock embraced RFID and centralized BI as a means to further differentiate itself from cut-throat competitors.

“Most construction companies compete on low price and poor service, but our paradigm is really high-quality materials and high-quality customer service,” he says. So we monitor on-time delivery statistics, we do statistical process control of quarried rock, because it’s sized, and it needs to be consistent in the sizing,” he explains. “This lets us provide a unique service that customers value because it gives them all sorts of information about the status of their order.”

The only question, he says, was whether Graniterock would build its own next-gen BI solution or tap a shrinkwrapped BI tool. Given the complexity of the proposed GraniteExpress 2 application, which assumed massive information collection activity and Webbased reporting capabilities, it was a no-brainer: Shrink-wrapped BI was the way to go. It’s a conclusion that a lot of former BI DIY-ers have also reached.

The case for shrink-wrapped BI
Basically, business intelligence describes nothing more than straightforward operational reporting—running some kind of client application such as Microsoft’s Excel spreadsheet, for example—against a relational database. There are other important pieces to the enterprise BI puzzle, such as extraction, transformation and loading, but in many cases, it involves making slight formatting changes to data when it’s exported from one source and imported into another. As a result, ETL can describe a process as simple as transferring data from one source to another via ODBC, perhaps with some script-driven transformations along the way.

The point is that commercial off-the-shelf software isn’t a prerequisite for doing, much less for profiting from, enterprise BI. In fact, given the upfront cost of COTS packages—ad hoc, or do-it-yourself—their licensing costs and the annual maintenance fees charged by most BI vendors, BI has traditionally had a lot to offer ROI-wise.

It’s little wonder, then, that organizations have overwhelmingly opted to build their own reporting and analysis applications instead of buying them.

Three years ago, TDWI put the build-vs.-buy split for analytic apps at about 2:1, but according to TDWI’s research, that ratio has narrowed over time. Why? Because even though there are still a number of very good reasons for organizations to embrace do-it-yourself BI, especially for applications such as straightforward operational reporting and spread-sheet-driven analytics, there are probably just as many reasons not to.

For starters, analysts say, homegrown BI invariably results in siloed BI. Consider the case of Excel, the most popular ad hoc query and analysis tool. Used by itself, Excel provides no central means of data management, data consistency, data collaboration or data security. In practice, non-homologous versions of the same spreadsheet tend to proliferate throughout an organization. There’s no single version of the truth.

This is what TDWI research director Wayne Eckerson famously called the “spreadmart” phenomenon. Purveyors of commercial reporting and analysis packages have tended to exploit spreadmart FUD in their marketing, but with the mainstreaming of SOA, there is a strong case to be made for centralized BI. And in this respect, commercial BI ISVs do have something to crow about: After all, almost all top-tier BI vendors have incorporated service-oriented underpinnings into their BI suites, and most of the ERP systems BI tools commonly report against have also been service enabled.

There’s another reason COTS packages make sense: proven scalability. This was the case with Graniterock, which found that the scope of its RFID initiative far exceeded the capabilities of its homegrown Crystal Reports-based reporting solution, which has well known scalability and licensing limitations with respect to concurrent user support. As a result, Snodgrass says, his company, which amassed a repository of more than 1,200 Crystal reports, ponied up the cash for Business Objects’ high-end Crystal Enterprise offering. Since doing so, Snodgrass confirms, Graniterock hasn’t looked back.

“Once we upgraded to [Crystal Enterprise], we realized that we could e-mail reports using Crystal automatically, so we actually have set it up so that our more sophisticated customers on a daily basis get a report e-mailed to them on the status of their shipments,” he says. “From a customer service standpoint, more and more we’re realizing that we have all of this data that we can get in our customer’s hands, and that Crystal [Enterprise] gives us so many ways to do it. The best part about it is that once we set up an automated scheduled report, there’s not a lot of overhead required to maintain it.”

Graniterock is by no means alone: Many avowed BI DIY-ers have warmed to the idea of buying some or all of their BI software from commercial best-of breed vendors. Take RufalloCODY, a provider of software and services for educational institutions and other nonprofit organizations, which has carved out a market-leading position on the strength of its analytic chops.

“The big issue we have is obviously the analytics of our business is really what drives profitability for our non-profit clients, so the more they can understand their data on a daily basis, the better they can use it for the purpose of segmentation [or] calling strategy, [and] the better off they are at raising money for their institutions,” says, RuffaloCODY CIO Bruce Lehrman.

Except that RuffaloCODYwas experiencing tremendous growth—on the order of 25 percent annually. As a result, Lehrman says, his company was quickly smashing up against the limits of its homegrown Crystal Reports and Excelbased BI infrastructure. And with annual growth projected to remain at or outstrip 25 percent over the long haul, Lehrman and RuffaloCODY determined that COTS was the way to go.

After an RFP process, the company opted for a BI software suite from relational OLAP (ROLAP) specialist Micro-Strategy. “The biggest issue for us was the strength they had in the online [analysis], which allows our employees all over the country, all over the world, to go in and dig into the information they want to see, as opposed to the [traditional] process of requesting reports and generating analytics on an ad hoc basis,” he says.

The case against shrink-wrapped BI
The biggest problem with shrinkwrapped BI, as almost any CIO will tell you, is that it’s expensive. Moreover, comparatively few vendors can claim to address all the pieces of the BI puzzle, such as operational reporting, ad hoc query and analysis, ETL, data cleansing and data profiling, which means customers must frequently purchase point solutions from different best-of-breed vendors. That gets expensive, too. On top of this, users say, commercial BI packages keep getting more expensive every year.

This last trend, in particular, is one that few customers think can, or will, be reversed, in part because BI requirements will only become more sophisticated. As a result, more and more organizations will replace functional DIY or ad hoc BI solutions with commercial or best-of-breed tools. This will almost certainly translate into even higher BI licensing and maintenance costs.

“[BI] is a pretty mature area, so there aren’t a number of big companies that are just starting down a business intelligence path, so the amount of green field business is diminishing,” says Matthew Datillo, CIO of PerkinElmer, a health sciences research, development and manufacturing company. “As a result, there’s much less pressure for [BI vendors] to accommodate [customers] on pricing, and there’s going to be even less [pressure] in the future.”

Datillo stresses that he’s very happy with his BI vendor, Business Objects. In fact, he explains, PerkinElmer was able to centralize and consolidate its highly heterogeneous ad hoc and commercial BI infrastructure—the culmination of departmental silo-ing and intensive merger-and-acquisition activity—on top of Business Objects’ Enterprise BI suite and Crystal Enterprise reporting toolset.

In spite of its next-gen Business Objects underpinnings, PerkinElmer still uses an ad hoc ETL tool—Oracle PL SQL-driven scripts. It’s something of an inelegant solution, Datillo concedes, but it’s a lot cheaper than the alternative—purchasing an enterprise ETL tool from Oracle (Warehouse Builder), Business Objects (Data Integrator) or another COTS BI vendor.

PerkinElmer isn’t alone: Many BI suite adopters bring some of their homegrown or ad hoc software forward with them. Still others remain on less-powerful BI point products they’ve already paid for. Given the high cost of shrinkwrapped BI software, some say they don’t have any other choice.

Take Jody Porrazzo, director of econometric risk strategy with APEX, an insurance services consultancy. Porrazzo describes herself as an extremely happy user of SAS Institute’s BI software. Her company has been running the SAS BI suite since 2003, when it implemented SAS8 in place of an homegrown product. Recently, APEX transitioned to 64-bit SAS9, deploying that suite on top of a new 64-bit Itanium2 server from Hewlett-Packard APEX’ SAS9 implementation was easily a six-figure effort, including hardware. But the company still uses a homegrown ETL solution, even though SAS markets one of the top three ETL tools available.

“We do not have an ETL tool. We’re still using a homegrown tool, although one of the things we are evaluating is the SAS [Enterprise] ETLtool,” she says. “But ETL from anyone is expensive, and we still haven’t made a decision on that.”

Sidebar: In ROI we trust

ILLUSTRATION BY PETE MCARTHUR