The story is familiar: Company frets over the ability to serve customers and know what they need before they need it. This nightmare causes sweat-soaked nights and is driven by competitive paranoia and technological impotence.
In this case, the subject of the story is no ordinary company: Merrill Lynch & Co., the Wall Street juggernaut whose millions of individual customers and $1.3 trillion in assets under management makes the challenge of managing customer information all the more critical.
Though viewed by bank and broker competitors as the model for running a modern financial services company, executives inside the Merrill Lynch concluded two years ago that the firm s Achilles heel was its ability or inability, as the case may be to use millions of pieces of data to tap opportunities in its fabled U.S. private client group.
Until recently, the company s data resided on 25 different computer systems throughout the company in a DB2 environment which demanded time- intensive programming talent to begin to crack. Because Merrill Lynch managed marketing information on a mainframe computer which was also used to process millions of daily transactions, even basic queries which should take minutes could take days. The data was not well integrated, says Joe Hollander, director of business information management for Merrill s private client group, which includes individual accounts and the small business unit. Clearly, the system wasn t developed with information viewed as a strategic asset.
Always on guard
That has changed. Since last November, the company has been on-line with a new system called Midas, an enterprisewide data warehouse which centralizes customer data and enables individual business units to quickly extract data from their desktops. We re constantly critical of our own process. We try not to be complacent, Hollander says. You don t know if your next major competitor is Citibank or Microsoft. Information providers are playing in (our) space now.
Merrill Lynch s move to create Midas, which stands for management information decision analysis support, provides critical insight into the issues confronting financial services companies of all types and sizes battling to keep customers and develop richer relationships. Hollander s experience illustrates the need to educate end-users of the data; stresses the importance and economics of building scalable technology; punctuates the reality that even major leaps in technology are only first steps; and, perhaps most importantly, shows the need to understand how to get a return on investment from the millions of dollars which will be spent.
The business of managing customer data has become a rapidly growing industry in its own right. By one estimate, companies across multiple industries will spend billions this year to build more effective data warehouses or more specialized data marts. That figure is expected to rise dramatically in the next several years. Within the financial services industry, exact figures are hard to pin down, but most concede the search for profits begins with a company s own customer information.
For Merrill Lynch, the process of building a better data warehouse started in 1996. After Hollander s team assessed what needed to be done, it settled on a plan to create Midas to make access to data more democratic. The Merrill team settled on an IBM SP parallel-processing computer for the data warehouse, with an Informix database. The parallel processing system is powered by 14 engines and is easily scalable as the data warehouse grows. The data is centralized, cleansed for accuracy and then put into a centralized file for each customer.
a benchmark roi
While Hollander declined to discuss the specific cost of the system, he elaborated on the economics of such a project. As Hollander studied the scope of the project, he also focused on convincing senior management that the multi- million-dollar cost was an investment not an expense. Rather than focus on traditional cost saves, he focused on the upside that readily available information would generate. The estimates depended more on the opportunity cost of the use of the data warehouse, he says.
As a benchmark, Merrill Lynch set a target of 401 percent over three years. That standard was established by International Data Corp., which researched the financial impact of data warehouse projects. The study, which was underwritten by IBM and Informix and 18 other technology companies, focused on case studies at 62 organizations from Alaska to Finland to determine the payback for undertaking similar projects like the one Merrill Lynch planned.
ignore quick-fix hype
What the survey found is that, overall, the data warehouse implementations studied by IDC generated an average three-year return on investment of 401 percent. Over 90 percent of the organizations reported a three-year return on investment in excess of 40 percent, half reported returns of greater than 160 percent and one quarter showed returns greater than 600 percent. The average payback for the warehouse application was 2.3 years on costs which averaged $2.2 million.
Stephen Graham, the Toronto-based IDC analyst who wrote the study, says that companies have misdirected expectations and don t see the data warehouse for what it is: an opportunity for improved process and management. Most companies should ignore the hype regarding the quick-fix or single piece of information that can result in a single, one-time benefit, he says. A warehouse provides the underpinnings for an environment where management is better served to make decisions based upon data rather than intuition. One of the fundamental questions organizations need to ask themselves is what is the bottom-line benefit when information is provided that is more accurate, more timely and so on.
The financial benefits of a data warehouse are split, Graham says. We found that MIS savings and end-user productivity enhancements accounted for about half of all benefits, and process savings where the warehouse was cited as the sole source driving management decision-making accounted for fully 50 percent of the benefits, he says.
Ben Barnes, general manager of the IBM s Global Business Intelligence Solutions unit, says that while traditional cost savings are possible, the real motivation is to develop more profitable customers, to up-sell or cross- sell products and for retention. This is the red hot issue, he says.
Retention alone is a potent factor in justifying the cost of a data warehouse project, says Dave Watson, technical director at Informix, noting large-scale projects at First Union Corp., Fidelity Investments and Fleet Financial Corp. When they look at retention models, they can see value in keeping one or two percent of their customers, he says. Fleet sees a $50 million pre-tax gain from their project.
Merrill Lynch has laid the groundwork for similar results. Today, the company is registering more than 3,000 direct queries a month versus a history of a few select power users accessing data. Before Midas was available, I to compete for the time of programmers, says Mark Kolakowski, controller and business strategist of the Merrill s private advisory services unit which focuses on high net-worth individuals. The analysis I wanted would languish for weeks at a time because a programmer wasn t available. Midas has put information access on my desktop. Now I can do it on the fly, and I don t have to go through an intermediary.
Watson says that Merrill Lynch has major opportunities which can now be exploited. Merrill has relationships with many corporations, he says. They manage 401(k) (plans) for the employers, but those employees may not be an individual customer.
Hollander agrees that the new data warehouse makes it possible to capture external information to profile potential customers and to identify opportunities that were not apparent before. By example, Kolakowski says that a customer with $2,000 in an IRA may not seem like an opportunity, but that is overlaid with external information on corporate insiders who may have more investable assets or could need Merrill Lynch s advice on capital markets issues for their company.
Ultimately, Hollander expects more than 500 business analysts and other employees to use Midas regularly. The next phase is to start using dimensional tools OLAP applications to generate detailed reports, which help executives better manage their operations and to benchmark performance on a near real-time basis. Getting the users more knowledgeable about the data is much more of a challenge than getting the data together, he says. If we under did anything, it is educating them about what there is and how to use it. There are also projects underway to create data marts for more specialized needs. Informix s Watson says that plans are underway to develop a data warehouse for the company s mortgage lending unit. Adds Barnes, You never really quite finish a business intelligence system. You find new data to add. You find new users of the data. And you need new research tools.