GOLDEN MEANS: IN THE BEGINNING
by Inderpal Bhandari, executive editor at large

In mid-1997, I founded Virtual Gold, Inc. to create innovative data mining technology and bring it to market. I assembled a crack team of software developers. The ideas flew fast and the code even faster. We were on our way. I could agree with Ambrose Bierce that the future is indeed a period in which our affairs prosper, our friends are true and our happiness is assured. Until I talked with Dollar, Mark, and Otto, senior executives who were representative of our potential customer.

Dollar was the chief technology officer of a major U.S. bank. I apprised him of our plans. "Data mining, oh yes, we have got it covered, been doing it for the last year or so." Upon further conversation, it became clear that actually they had been building a data warehouse. I pointed out that was not data mining. His response: The key is getting the data to one place. Once that is done, querying, analyzing, mining, call it what you will, is a simple matter.

Mark was Dollar's counterpart in a German financial institution. He too had "done data mining, for several years!" Upon further probing, it became clear that his company's data was in a central warehouse. Their marketing staff could then easily create customer lists for specific products. They brainstormed about the attributes of the customers likely to buy the new product and then queried the warehouse to build the lists. I pointed out that sounded more like database marketing to me. His response: Querying, database marketing, mining, call it what you will, the key was to convert the data to useful information, such as the customer lists being created by them.

Otto was the CIO of a large car maker. He understood data mining from a technological perspective. He was going to use it to analyze sales data augmented with demographic information. He hoped to learn enough to improve the sales of certain cars in certain regions. More on Otto later, but do note that while his analysis can certainly improve the sales figures for the year, it cannot improve the quality of cars being produced.

Dollar, Mark and Otto are by no means exceptions to the rule. Senior executives are often confused about the technology that underlies business intelligence. In his teachings, Gautama Buddha, the founder of Buddhism, proclaims that the first step on the path to Nirvana, i.e., escape from the wheel of reincarnation of life and death, is "to have the right idea about things." Well, my friends, with regard to data mining, if the current state of confusion prevails, we will definitely be spinning our wheels for some time to come. There are many folks with titles that have C's and O's who do not understand the difference between data warehousing, database query management systems, database marketing, on-line analytical processing, and data mining. Simply put, they do not have the right idea about these things.

Well, then, who does? For one, Tom Sterner, an assistant coach in the National Basketball Association (NBA) with the Orlando Magic. Sterner has been using the IBM Advanced Scout data mining program since 1994. He recently mentioned a data mining experience that showed he had come full circle.

Last year, Orlando was playing the Miami Heat in the playoffs. They went into the best-of-five series badly depleted with injuries among their starting unit. They were soundly beaten in the first two games at Miami, outscored by a total of 56 points. They returned to Orlando for Game 3 with their game plan for the series in shambles. Their backs were to the wall. They knew not what to do.

Until Advanced Scout revealed a startling pattern. When their players on the floor included Darryl Armstrong and Danny Schayes, the Magic had outscored the Heat by 23 points! Sterner was stunned. Both Armstrong and Schayes were bench players, with minor roles in the original game plan. Sterner checked and double checked. The figures were correct. The coaching staff then proceeded to pick apart the videotape of the game when those two players were on the floor. The result was a new game plan that centered around Armstrong, with a supporting role for Schayes.

Orlando won Games 3 and 4. They lost Game 5 in Miami but very narrowly. As Sterner puts it, by extending the series to five games, their data mining efforts had brought in millions of dollars in gate money and in arena sales, salvaged the honor of the team before the home crowd, and created an extremely exciting series for the television audience.

Data mining can reveal such unexpected, counter-intuitive patterns that otherwise one would never have thought to look for even if the data were all in one central warehouse and one spent all the time in the world brainstorming what to look for. Mostly, brainstorming is done in the box. Data mining, in contrast, leads to out-of-the-box thinking.

And that, in turn, suggests that data mining can be used to bring about fundamental changes to your core business. Unfortunately, that is not the use it is being put to. Today, data mining, even where appropriately applied, is largely restricted to improving the marketing and sales effort of a business and usually does not extend to improving its core products. For example, Otto was interested in figuring out how he could sell more cars but not how he could fundamentally improve the cars being produced. Otto's vision is inherently limited in the return of investment it can provide. Once the sales or marketing process has been optimized, significant as that may be, there can be no further impact to the business.

This is akin to applying data mining to the data of the ticket sales of the NBA as opposed to the game data analyzed by Advanced Scout. Sure, that will improve the sales of tickets. But it will not improve the fundamental product of the NBA, the game itself. Contrast this with Advanced Scout, where the target of improvement is the game of basketball itself. As is clear from the Orlando experience, the impact is more fundamental, reaching every aspect of the business of the NBA. If the game itself is made more competitive, it will rise in popularity, increasing ticket sales, broadcast revenues, merchandise sales, etc.

The difficulty, of course, lies in the fact that in a large business, there are many areas that can be targeted for improvement. For example, in the NBA one can mine data of ticket sales, or of broadcast relationships, or of merchandise sales, etc. The involvement of executives at the highest levels is required to target strategically significant areas that can produce fundamental impact.

The conclusion that I have come to after talking with several C*O's such as Dollar, Mark and Otto, is that cannot happen unless we make a concerted effort to educate senior executives about the technologies of business intelligence. Consequently, Virtual Gold now has a consulting and education arm in addition to its technology arm. If you would like to learn more about how data mining and the other technologies of business intelligence can fundamentally change your business, contact us ( http://www.virtualgold.com ).