But it sounds so impossibly boring," I protested to Ari. "Data warehousing has all the glitz of a zircon and I'm looking for a diamond."

Ari Fishkind, my IBM media relations guru, assured me with fitting fervor that it was not boring at all. "Data warehousing," he said, "gives you the power of the fabled Rumpelstiltskin?it allows you to weave the straw of daily business data into gold!"

"That's not a bad line," I laughed. "You should use it in one of your press releases." He had tried, apparently, but somebody nixed it.

"And if you're looking for a diamond," he continued, "do I have a deal for you. Cartier," he said, "is using a leading-edge, AS/400-based data warehousing application. I can put you in touch with their MIS manager in London."

"Good chap," I retorted, warming to the subject.

Frankly, I have a more modest view of data warehousing. I think of it as the answer to the age-old dilemma of what to do with all those archive tapes in the tape library. Row upon row of ancient tapes that hang from metal branches like inedible fruit, or decay in opaque plastic coffins?essentially worthless data, captured and stored at great expense. Turns out there's gold in that mountain of data, but until recently there was no cost-effective way to mine it.

With the availability of faster processors and larger storage capacities, the data can now reside online, where it becomes accessible and, therefore, useful. And, with a powerful new set of midrange tools from IBM, customers can perform sophisticated analyses that were previously either too time-consuming or simply impossible to generate.

Some of the nation's leading companies, including MCI, Citibank, and American Airlines, are beginning to see the value of learning from their own history. According to Business Week, Burlington has built its business around a 1.5 trillion-byte data warehouse that allows its managers and salespeople "to identify top-selling styles and brands, balance regional inventories, [or] measure the performance of one store manager against others." Interest in data warehousing is estimated by Price Waterhouse to top $20 billion by decade's end.

To warehouse such towering amounts of data, the host system should ideally be scalable and capable of parallel processing. That is, it must be capable of supporting incremental processors and DASD or, optionally, it should be able to run in tandem with other like processors that can share partitioned queries. Finally, it must have a powerful relational inquiry system that operates in a parallel/scalable environment. The industry moniker for such software is OLAP, which stands for On-Line Analytical Processing.

The AS/400 now provides all three capabilities. With the announcement of DB2 Symmetric Multiprocessing (available this month for users running V3R1), data mining aficionados can dredge with software that spreads database queries over multiple processors on a single scalable AS/400. With the power of additional processors, answers are returned up to four times faster.

Another new product, DB2 Multisystem, will allow customers with Opticonnect to exploit the combined power and storage of up to 32 AS/400s in parallel. It spreads queries over multiple systems, allowing them to solve individual parts of a problem concurrently. Expected availability is in the first half of 1996.

Stripped of the acronyms, data warehousing simply applies the flexibility of relational database management to massive amounts of data stored on single or multiple systems. Its advertised purpose, among others, is to unearth previously undetected trends and relationships.

The skeptics among you may ask, "So, what good is that?" Well, for one thing, there is the curious relationship between diapers and beer. A major retailer, sorting through the chaff of sales data, discovered that beer and diaper sales unaccountably increased on Thursdays. Further investigation revealed that young fathers typically shopped on Thursday evenings and apparently stocked up on the weekend's essentials. The store stopped short of erecting a Bud Light and Pampers display, but it did price beer, diapers, and related items accordingly and was thus able to increase sales.

The value of data warehousing in this case was in making connections that are not ordinarily made.

I wanted to find out how a data warehousing implementation actually worked. So, on Ari's advice, I got up early the next morning to call England.

"Where are you going?" asked my wife.

"I'm going to call Cartier in London." She brightened visibly. Even in her drowsy state, she had made some primal connection between Cartier and Christmas.

Michael Ellard was just leaving work when I rang. I could tell he was excited to take my call, but the British are stoic, so that was probably why he wasn't showing it. "Data warehousing," he acknowledged, "has been enormously helpful to us. Previously, people would key financial and sales data into their personal computers and run ad hoc reports. Now, the information is quickly available, cross-indexed any way they want to see it."

And, since he probably hadn't heard the Rumpelstiltskin analogy, he offered his own simile. "For us, data warehousing works like a three-dimensional cube," Ellard cordially explained. "One side allows us to view a company or distribution channel. The second side represents an item, and the third side allows us to specify a period of time." The information retrieved from Cartier's AS/400 Advanced Systems 310 allows managers to predict trends and make sounder marketing decisions.

Cartier, Ellard revealed, sells a variety of merchandise under its name. Beyond a sparkling $3.8 billion in jewelry sales, Cartier also markets watches, pens, tableware, leather goods, lighters, fragrances, and other accessories. Sales outlets range from airport boutiques to exclusive lottery-winner-destination jewelry stores. "We track sales and determine where the best fits are," said Ellard. Buying patterns also help determine the location and inventory mix of new outlets.

I asked about the software Ellard has installed. "We use Hoskyns financial systems, which are linked to our stock, order entry, and distribution systems," he said. "On top of that, we've installed AMIS/400 [Advanced Management Information Systems], also from Hoskyns. It gives us tremendous flexibility."

Hoskyns is but one of a growing number of providers of OLAP client/server tools. Other products include Brio Query from Brio Technology; SalesTracker and DataTracker from Silvon; and MIT/400 from SAMAC.

Cartier's was a good news/bad news implementation. "We really only had to make one change," Ellard confided. Unfortunately, it was to the general ledger. The G/L and chart of accounts numbers were modified to allow for more inquiry flexibility, which unavoidably touched numerous other applications. The entire implementation took 18 months and required the support of five employees and two consultants. Users, however, needed only one 2-3 hour training session before becoming proficient enough to use the tools at their desktops.

Ari, it turns out, was right: there is gold in that mountain of data. The magic comes in its application and that is only limited by imagination.

I thanked Ellard for his time and was just hanging up the phone when my wife came into the kitchen. "Tough luck, honey," I said. "They're out of diamonds."

"That's a shame," she countered, "but what a coincidence. The 49ers are out of season tickets, too."

Victor Rozek has 17 years of experience in the data processing industry, including seven years with IBM in Operations Management and Systems Engineering.

Also Read

IBM i Shops Pivot to Battle COVID-19

Energy, Vibe, and Atmosphere

Murphy’s Security Law