For more than a decade, the dominant image of electronic business analysis has been the spreadsheet: the two-dimensional grid of values formed by the intersection of columns and rows. Now, purveyors of a new generation of business intelligence software want to supplant the grid with the cube (or hypercube, as some call it). Imagine, they say, analyzing product sales in three dimensions: by time period, channel, and customer. Or four: add geography. Or five, six dimensions, with margin and scenario. Then imagine rotating the whole like some impossible Rubik's cube, constantly reorienting and parsing the data--all in the blink of an eye.

If a six-dimensional cube can't be visualized, such multidimensional rotation is nevertheless a daily reality for thousands of corporate managers and analysts. They are devotees of OLAP: online analytical processing. A technology defined by database pioneer E.F. Codd and colleagues in a 1993 white paper for Arbor Software Corp., OLAP is being aggressively marketed as the most advanced means of performing any number of critical business functions: sales forecasting, profitability and market-share analysis, budgeting, modeling, allocations, even activity-based costing.

Given the potential applications, there's no limit to potential customers. No wonder that dozens of vendors, from the obscure to the famous, now offer a multitude of sophisticated OLAP tools, from relatively inexpensive desktop clients to powerful servers with extensive multidimensional capabilities. Arbor, for one, currently counts more than 840 (mostly large) corporate customers for Essbase, its $37,500 multidimensional data-base. Another prominent vendor, Cognos Inc., has sold more than 200,000 copies of its $695 OLAP front end, PowerPlay. And software giant Oracle Corp. doubled the sales of its OLAP server technology just one year after purchasing it from Waltham, Massachusetts-based IRI Corp. Market watcher International Data Corp. (IDC), in Framingham, Massachusetts, predicts the multidimensional server market will more than double in size between 1995 and 2000, from $215 million to $476 million--a 17 percent compound annual growth rate.

But the most significant entry into the market was announced last October. In a deal that made few headlines, Microsoft Corp. acquired OLAP technology from Panorama Software Systems, an Israeli software developer. At a stroke, Microsoft put its imprimatur on the technology--which it plans to integrate into BackOffice, its server suite that includes Microsoft SQL Server database--and put the rest of the market on notice. Bill Gates and crew "won't have a real viable product for the next two years," predicts Howard Dresner, research director at Gartner Group, an information technology advisory firm in Stamford, Connecticut. But when they do, he adds, "it will suck the margins out of the business"--and spread the gospel of OLAP throughout Corporate America.

Thinking multidimensionally

Meanwhile, accounting software vendors are incorporating multidimensional features into their applications. Oracle, of course, can plug all of its own OLAP products into Oracle Financials. Hyperion Software offers Hyperion OLAP (which incorporates Applix/TM1's multidimensional server, TM1) and is bundling PowerPlay into its financial applications. Likewise, Lawson Software has integrated Essbase into its enterprise applications software. SAP, PeopleSoft, Coda, and Systems Union are a few more of the numerous other vendors offering integrated OLAP capability or supporting third-party tools.

There are compelling reasons for jumping on the bandwagon. Basically, OLAP software provides deeper, more powerful insight into what makes a business tick. It eliminates information bottlenecks, freeing end-users from dependence on IT staff to extract data from transactional and other systems. Connected to a data warehouse, data mart, or multidimensional database--all designed for multiuser decision support, with data modeled by subject--OLAP enables PC users to explore huge volumes of data at various levels of aggregation, and even, in some cases, to reach beyond to transactional data.

In short, OLAP software gives truth to two clichs: it turns data into information, and it empowers users.

True, a managed query software tool (like Seagate Software's Crystal Reports Professional or Cognos's Impromptu) can extract a great deal of what users need to know from relational databases. But for managers wanting a top-down view--to identify trends, model complex scenarios, or do free-form analysis--only OLAP is up to the job. "Query tools work well with small sets of data," explains Neal Hill, vice president of marketing at Cognos. "But let's say you have two product lines--each has 12 to 15 products in it, four geographies, 12 sales regions, five channels. If you want to explore this over eight quarters, you can't swim through line after line of data with a reporting tool."

"Businesspeople think of their data multidimensionally," comments Richard Creeth, president of Creeth, Richman & Associates Inc., a Norwalk, Connecticut, OLAP consulting firm. "A number belongs to a minimum of four dimensions: account, cost center, time period, scenario--even customer and product, making six." Report writers simply can't present this information as two-dimensional tables, he says. "The world is way past the era of static reporting--you want to be able to dynamically explore data," says Creeth. OLAP, he maintains, is the best way to do precisely that.

Creeth would get no argument from Jason Denman, manager of manufacturing and technology budgets and general accounting in Whirlpool Corp.'s North American Appliance Group, based in St. Joseph, Michigan (1995 sales: $5.1 billion). On January 31, the group's new OLAP application, Pilot Software Inc.'s Decision Support Suite, will go into production after 16 months of planning and installation. "We'll be able to quickly--important word--look at key drivers that we couldn't before," says Denman; in particular, volume, product mix, and product features. "These three drivers were very difficult to analyze," he explains. "Say you've got several hundred different SKUs at your plant: if you try to analyze volume across them, with different feature sets and so on, the complexity of the task grows quite rapidly."

The OLAP client/server application will use manufacturing data--manufacturing, material, labor, and freight costs--from a mainframe data warehouse, which is fed by various production and MRP systems. Some product feature data, which hadn't been tracked before, is being added manually. "Since we were creating this data [for the project], our data warehouse people said you had to prove that it belonged in the warehouse," says Denman. "In the long term, we'll automate it."

Power to the people

Empowerment, not complexity, was on the mind of Jay Forbes when his company, Halifax-based Nova Scotia Power Inc., rolled out OLAP software 18 months ago. "The number one reason [for buying the software] was to encourage people responsible for making decisions to understand the repercussions of their decisions," says Forbes, corporate controller of the $735 million (Canadian) utility. Between 75 and 80 managers and financial personnel use Cognos's PowerPlay to analyze budget variances, operating and maintenance costs, marketing activities, and customer service.

;"Take right-of-way clearing," he says, referring to the removal of trees and other barriers to distribution and transmission lines. Because customer service managers habitually have trouble keeping track of their budgets, says Forbes, "we'd find ourselves running surpluses most of the year, and then suddenly costs were flowing in toward November," resulting in unexpected deficits. The PowerPlay software will remedy the situation, although this particular instance may sound like a case of too much tool for the job. Why use OLAP instead of dedicated budgeting software? "Regardless of what the user wants to do, Cognos can accommodate it," responds Forbes, whether simple budgeting for customer service managers or complex trend and scenario analysis for Nova Scotia Power's financial analysts.

The OLAP client analyzes general-ledger and other financial data extracted from the company's accounting software. "We have people working throughout the province, and a sophisticated telecom infrastructure to support them," says Forbes. "The data extraction is pumped through the network after the monthly close. We give each of the users the information they need to analyze." The application administrator spends at most one day a month maintaining the system, he notes.

Data sharing and coordination on such a scale simply wasn't feasible with spreadsheets. "The beauty is that we can all communicate together, using the same source of information," says Forbes. (A data warehouse is on the drawing board for 1997.) It's all part of a larger design, to decentralize the finance group and create a flatter organization. "We're an engineering-oriented company," Forbes explains. "A lot of functionality has devolved to the field offices. We're formalizing that with the creation of strategic business units." Two divisional controllerships were created in 1996 as part of that effort, and the OLAP system will keep everyone on the same page.

"My philosophy is, create a tool that will allow you to implement change," says Forbes. The cost of Nova Scotia's tool? To date, under $100,000.

Across the universe

The ability to distribute OLAP capability throughout an organization, already abetted by LANs and WANs, is getting a giant boost from corporate intranets. Practically every OLAP vendor is incorporating Internet and World Wide Web technologies into its software, enabling users to gain access to remote servers and databases via the Internet with a Web browser. Customers will enjoy the economies of easier implementation (no need to install client software) and centralized administration and maintenance.

One company reaping the benefits of OLAP over the Internet is, not surprisingly, a networking company. Bay Networks Inc., a $2.1 billion vendor of networking hardware and software (routers, switches, hubs) based in Santa Clara, California, started out with Arbor Software's Essbase in 1995 to create a framework for management reporting.

"We use the following information: channel, inventory, point-of-sale information, bookings, and shipments," says Vito Palermo, vice president and corporate controller. The Essbase data is siphoned daily from Bay Networks's data warehouse, which is populated with source data from an SAP R/3 enterprise client/server system. "We cross-cut that by customer, sales geography, territory, and product," says Palermo, using Microsoft Excel as the front end. "How much business did we book yesterday? In Europe? What products? What margins?"

The multidimensional database allows users, numbering about 50 this past November, to make "very rapid and timely decisions," says Palermo. "Imagine having on your desktop a report: one sheet of paper. There's a set of buttons on the side: you can point and click, converting 2-D into 3-D views. I never see the [raw] data."

Soon, many more employees at Bay Networks will be pointing and clicking through multiple dimensions of data. In December, the company began implementing Arbor's Web Gateway for Essbase, a $10,000 Web server application. Using Netscape Navigator browsers (standard issue at Bay Networks), analysts will be able to access the Essbase database via intranet or Internet. Initially, a sales analysis application will be rolled out to between 150 and 200 users, and an executive information system application will go to about 50 managers. Eventually, hundreds more users will be able to tap into the system, according to Linda Bronson, senior IS manager of integrated data management.

"We have a lot of Web servers to go through to access either Essbase or our information warehouse," says Bronson. Queries through the intranet will be automatically steered to the optimal source. Those involving customer names, for example, will be directed to the warehouse. "You can't support 30,000 customer names in a multidimensional database" with a daily update of data, observes Bronson. The end-user won't be able to translate data into graphics on the browser, but the result can be cut and pasted onto Excel.

"People love the Web," remarks Bronson. While she expects the growth in Bay Networks's OLAP users to be "nonlinear," she doesn't think concurrent use will cause any performance problems. "We have international users--they come in waves," she explains. And although the number of users will multiply, a common, centralized security structure will enable database administrators to maintain strict control over the type and level of information that users have access to.

Beyond sex appeal

As OLAP becomes widespread, if not commoditized, vendors will emphasize less the sexiness of the new technology (which attracts IT people) and more its practicality (which attracts finance executives). "We're seeing increasingly that OLAP technology is being sold under cover of a business solution," says Henry Morris, research director for applications and information access at IDC. It's not the particular flavor of OLAP that's being promoted, explains Morris, but rather the value of the particular information provided, such as marketing intelligence or manufacturing process analysis.

But no matter how shrink-wrapped the product or persuasive the pitch, no OLAP application should be considered without corporate IT's full input. "OLAP is clearly a technology," says Richard Pope, vice president at Symmetry Corp., a San Rafael, California-based OLAP consulting firm. A company buying a multidimensional database management system will need to contend with a number of difficult issues, from data sourcing and modeling to systems integration.

How to choose the right OLAP tool for the job? First, says Pope, determine which type of product will best fit into the company's technical architecture--its client/server platforms, its networking systems, and so on.

Second, consider integration issues. If the source of data is the general ledger, does the OLAP vendor build links to the particular GL application? What kind of tools will be needed to integrate the product with a data warehouse? How do you convert the source files into the desired format? "Some products assume that you can get the data already prepped," observes Pope. "Other tools provide a programming language to model the data."

Third, which product best performs the desired application? "People do bake-offs," says Pope. "Ask the vendor, 'How will you build this app?' Get proof that it will work. Some products demo and prototype better than others; that's where references start coming into play. Get five references," advises Pope, "then put them aside and ask for five more."

Neal Ship, an OLAP product manager at Oracle, says financial users should consider the mathematical functions the OLAP product can perform, such as net present value, internal rate of return, and depreciation calculations. The ability to do simultaneous calculations is important, Ship adds; for example, when doing a cash flow projection, the software should be able to calculate at once how much money must be borrowed and how much interest will have to be paid on the borrowed funds.

As with data warehousing projects, sourcing and modeling data are perhaps the most critical, and certainly the most time-consuming, steps in an OLAP implementation (see "Tapping Your Hidden Assets," CFO, May 1996). Dimensions of data (such as product, customer, and channel), business rules (for calculating measures, such as net sales), and levels of aggregation must be defined. Source systems of data must be agreed on, and the data itself will have to be cleansed and transformed. Regular data loads to refresh the OLAP database will need to be scheduled--daily, weekly, and monthly.

"You've got to do a lot of up-front planning, both with businesspeople and IS people," says Whirlpool's Denman. "You've got to model the data correctly, because you're going to be slicing and dicing it in all different directions." The modeling can't be too thorough; if users want to add overlooked dimensions to an existing multidimensional database, the database must be physically rebuilt, top to bottom--no small task.

Properly configured, however, OLAP software gives free rein to the analyst's imagination. It's a technology whose time has come. Ask Jason Denman. "Initially, we planned for around 30 users [for Whirlpool's OLAP application], but we're already up to 45," he says. "And a lot of other people are saying, 'Gee. Neat system.' We'll have to go back to the steering committee and ask for more seats." *

OLAP tools that rely on stored, summarized data (sometimes called MOLAP, or multidimensional OLAP) are particularly useful for applications such as budgeting, financial modeling, profitability analysis, and forecasting. Data in a multidimensional database or cube is rolled up just above the individual transaction layer to the desired dimensions: time period, sales channel, customer, country, and so on. Because the data is structured the way users want to look at it, MOLAP tools can be extremely fast and agile. Multidimensional databases require specialized skills to administer, however, and are generally limited in size to about 20 gigabytes.

Since ROLAP tools don't store data but draw it from relational databases in response to specific queries, they are extremely flexible. And since relational databases can store much more than multidimensional databases, ROLAP tools can analyze a correspondingly larger amount of data. ROLAP is particularly well suited to retailers that want to sift through extremely fine grained data--daily sales, for instance--and to organizations that want to explore vast customer databases, such as insurers and financial-services firms. Some ROLAP products work best with "denormalized" databases designed for data warehousing, from software vendors like Red Brick. -- E.T.

Edward Teach is a senior editor at CFO.