A Business Approach to Technology: Shearson Lehman Commercial Paper Inc.
SIGNE A. VON VERDO (WEBER)
Shearson Lehman Commercial Paper Inc. (SLCPI) is a primary dealer in money markets and in mortgage-backed and government securities. SLCPI first opened its doors in 1963 as a wholly owned subsidiary of Lehman Brothers (subsequently acquired by Shearson/American Express). At that time, SLCPI's business focused on a limited number of products, all of which were easy to understand and easy to process. Starting around 1979, however, a number of innovative fixed-income securities were introduced into the marketplace, and business took off. There has been a tremendous growth of product types, product complexity, and transaction volumes since then. Today SLCPI is a low-margin, highly competitive business with a broad range of complex products. It is a profitable and powerful organization that accounts for a significant profit for its parent company. In 1986 SLCPI was using a set of manual processes and struggling to catch up with the business. It has since become an organization on top of the industry's tremendous growth. SLCPI has consciously cultivated a partnership between technology and business, resulting in redefined work flows and replacing manual processes with automated systems.
Looking at SLCPI rather simplistically, the organization has two basic functions, referred to as the front office and the back office. In fact, because this structure is found across the entire
industry, competitors such as Goldman Sachs or Salomon Brothers look fundamentally the same as SLCPI. The front office is made up of the sales force and traders, who deal directly with the client base—typically large institutions such as AT&T and IBM. The sales force maintains the ongoing client relationship, focusing on investment requirements and short-term financing needs as well as providing a key contact for error resolution. In SLCPI it also includes a number of "quants," or quantitative analysts, who perform analysis on trade data, pricing information, and economic data in an effort to support traders in evaluating market directions. Traders are authorized to execute the trades both for SLCPI's client base (per instructions of the sales force) and for the "house" account (SLCPI's money). Traders and the sales force are typically compensated on the amount of revenue they generate, and annual earnings of six figures are not uncommon. Back office personnel include clerical and professional staff who ensure that trades executed by the front office are processed correctly, accounted for accurately, and reported in a timely manner. It is important to note that many SLCPI products require settlement on the same day they are traded. This leads to a sense of considerable urgency in the back office to stay on top of its work load. Back office personnel are usually on straight salary and paid a bonus based on the overall performance of the organization.
THE ORGANIZATION AND THE TECHNOLOGY-1986 TO 1988
In 1986 the SLCPI environment was characterized by processing backlogs, high error rates, and overtime. SLCPI had 650 employees handling about 4,000 transactions a day in 28 major products. The daily trade total was $50 billion or roughly $12.5 million per trade. Because the dollar amount of any trade was so large, associated processing errors or backlogs became significant. As one SLCPI manager put it, "All the money made in the front office can just as easily be lost in the back office." Challenges due to high growth and increasing complexity of the business were compounded by predominately manual operations that had been put in place to handle the products of 20 years ago. It was a constant game of catch up, and senior management was becoming increasingly concerned over managing the inherent risk of SLCPI's high dollar trades. Management reporting was limited as was the time to assess where the problems were. Furthermore, support personnel were spread out across SLCPI and had limited opportunity to
share ideas. A few enterprising individuals had attempted to streamline their units where they could and were working with standalone personal computers. There were perhaps 10 or so of these units across various departments where spreadsheet applications were being used. Those individual efforts were uncoordinated, and ineffective. Apart from the few personal computers peppered throughout the organization, the bulk of data processing support was provided by a central management information systems (MIS) function. Requests for data processing support were made to the central MIS group, which also provided design and programming.
Today SLCPI has grown to 950 people and trades more than 40 major products plus an additional 100 derivative products. The business has grown more than 50 percent since 1986 and currently trades $85 billion on 5,500 transactions every day. Despite the high rate of growth, SLCPI has successfully eliminated the processing backlogs of 1986 and reduced error rates to a manageable level without increasing the number of back office personnel. (Sales and trader positions together with the addition of systems professionals accounted for the overall growth in personnel). Morale is high and overtime is down. Most of the manual back office processes have been automated across five major microcomputer-based networks linking more than 500 terminals and running 100 unique applications. Figures 1 and 2 show a work flow that was transformed from manual to automated processes. Systems support has been wrested from the central MIS unit and moved under the control of SLCPI management. SLCPI has further decentralized systems and programming support, pushing it into individual business functions in the organization.
SLCPI's transaction processing and back office operations (i.e., securities clearance and accounting, which take place once trades have been executed) were developed years ago for a different type and breadth of product line. Work flows were manually intensive and simply not able to handle the increasing product complexity and growing volume of trade activity. The clerks and line management who staffed these functions were also rooted in the past. The reactive and urgent nature of daily operations left little time or energy to step back and assess how the operations could be changed and improved. Not surprisingly, there was a limited awareness of systems or technology in general such that potential opportunities that could have helped were not explored.
Of all the problems facing SLCPI in 1986, management's inability to manage the risks posed by processing errors was the most critical. The size of SLCPI's individual trades meant significant revenue loss when errors or processing backlogs occurred. The interest lost on $12.5 million (the average size of a trade in 1986) came to thousands of dollars a day. When several trades were in error suspense or backlogged, the revenue loss multiplied.
Existing management reports did not provide sufficient timely information to determine the firm's position accurately. Consequently, SLCPI found it difficult to manage problems or even to know if there was a problem. Limited management reports became more of an issue with passage of the Securities Act of 1986, which required companies to demonstrate they were maintaining an acceptable ratio of liabilities to assets.
Ongoing requests for improved management reporting were primarily directed to Shearson's central data processing (DP) organization. Shearson, not unlike any number of other firms, was highly centralized in its data processing support and maintained a
large group of systems professionals who ultimately reported to the chief financial officer. Also not unlike other centralized DP shops, Shearson maintained a hefty backlog of programming requests. In addition, systems that were developed by DP and delivered to SLCPI were not finding acceptance with the people who requested them. Despite the increasing urgency of SLCPI's system needs and the number of resources the DP shop was willing to commit, system support was generally considered sluggish and not responsive enough by SLCPI management.
Thus SLCPI was faced with backlogs and processing errors, increasing regulations from the Securities and Exchange Commission, lost revenues, low morale, and a never-ending chase to complete the day's processing and book all trades before being ready to move on to the next day's business. All of these conditions existed in an increasingly competitive environment requiring streamlined and accurate processing and reporting.
Two key moves in 1986 began the transition that was to allow SLCPI to gain control over its operations. First, Steve Gott, a senior vice president from the First Boston Corporation, was recruited to clean up SLCPI's processing problems. Second, under Gott's direction SLCPI consolidated key processing areas—back office operations and accounting—into a new organization called Trading Services. The idea was to create an identity for previously far-flung operations and develop a synergy across the newly formed organization. Nearly 80 percent of Trading Services' 100 or so managers were asked to step aside, and a new team of handpicked recruits was brought in to replace them. The new organization then set about the task of defining the problems. Each of the Trading Services functions was scrutinized. Work flows were analyzed, skill levels assessed, and bottlenecks identified (see Figure 3). Associated reporting requirements were defined. All of the work was done in-house with the philosophy that line managers should be as close to their problems as possible.
During the analysis, Trading Services determined that effective information systems would be the key to achieving SLCPI goals. As a result, Trading Services proposed to pull MIS support entirely away from the central DP shop and put it under the direct control of SLCPI where it was hoped problems and solutions would find themselves more quickly. Vigorous discussions between SLCPI management and central MIS management ensued,
There was a general assessment of each person's abilities and potential, with the result that more than 100 people (in addition to the 80 departing managers) were replaced. The urgency to resolve quickly a number of critical problems pressed the need to bring in a higher level of technical and business professionals and replace those perceived as requiring time to come up to speed. A formal program was established for those asked to leave, and efforts were made to assist them in finding new positions either within Shearson or outside the firm. As new talent was brought on board, the emerging organization began to solidify. Teamwork was emphasized, and individuals were encouraged to develop a general understanding of SLCPI's business. Formal training was established to further this end, and each SLCPI employee was expected to sign up.
Top-quality systems professionals were recruited to staff the data processing support function that Trading Services had assumed. In an effort to ensure that these resources were close enough to the business problems that needed to be addressed, the organization was structured such that the systems professionals reported directly to the Trading Services department they supported. This highly decentralized structure provided an opportunity for business and systems people to work as closely together as possible. Adding a new twist, business and systems professionals (not including traders) shared equivalent job levels, base compensation, and bonuses determined by SLCPI's performance. Another innovation was to make the responsibility for developing successful systems an explicit role of the business manager rather than an implicit role of the systems professional.
The emphasis on a partnership between systems and business professionals worked both philosophically and from the standpoint of compensation to build a strong team environment. The further emphasis on SLCPI's performance not only helped in team building but also encouraged individuals to look around outside their immediate scope of responsibility. The result was an organization whose members had a good understanding of SLCPI's business in general and the freedom and encouragement to establish contacts as needed across the organization.
In addition to bringing in high-quality recruits, instituting team-building efforts, and winning the fight for local systems support, the head of Trading Services also introduced a new style of planning to the organization. It spelled out activities and expenditures for the coming year as well as the short-term system deliverables and paybacks expected in each quarter. Senior management was
presented with the Trading Services plan for 1987 and asked to review and approve the project list and associated expenses. Since the issues at hand were generally understood and payback periods were short, approval of the plan presented no particular obstacles. However, the capital acquisition process dictated by SLCPI's parent company raised problems by slowing acquisitions of computers and related equipment and software to about two months. In light of the aggressive nature of the Trading Services projects and associated commitment to short-term payback periods, an exception to Shearson's existing acquisition policy was necessary. This policy exception pushed the authority to make decisions for all technical expenditures within the approved budget down to each budget center manager and significantly accelerated the acquisition cycle for hardware, software, and approved personnel requisitions.
A new style of system development was also introduced and encouraged a certain degree of experimentation. Instead of the traditional phased approach to applications development often adopted by central MIS organizations, a more iterative, pilot approach was frequently employed to test a proposed system. Pilots typically incorporated roughly 80 percent of known requirements and anticipated changes from the start as the design evolved during use by the business community. As the pilot was moved into production, the iterative process for identifying changes and refinements continued. In fact, it was not unusual for the original requirements to undergo a total revamping as the business process was explored collectively by the design and user teams. Changes continued until the department determined that further modifications would not provide sufficient payback or that there was a more pressing priority. Because both business and technical resources were ''business smart'' and operated as a partnership, priorities for spending limited programming resources were rarely disputed.
From the start, five separate local area networks (LANs) were defined for each of the four primary product areas and the back office. The networks were designed to do the bulk of SLCPI's trade entry, back office processing, and accounting functions. Selected information locally stored on SLCPI's networks was in turn transmitted daily to the mainframe to update the firm's central books and records. Figure 4 shows a schematic of the SLCPI network structure. SLCPI believes that development components as well as development time are more manageable in a microcomputer environment than in a mainframe environment. Precisely
because SLCPI was so far behind in systems support, the firm was eager to select a solution that would allow the fastest development time possible for concurrent development efforts.
The technology SLCPI adopted was primarily microcomputer based. From the middle of 1986 to the end of 1988, more than 500 workstations were installed on five separate LANs, and more than 100 applications were developed. Specifically, SLCPI installed a 300-workstation Novell/Proteon LAN, an 85-workstation 3COM optical fiber LAN, and three Sun LANs supporting 80 workstations. The Novell LAN is installed in the back office and supports updates to SLCPI's income statement, balance sheet, and client accounts—all of which are maintained locally. The 3COM LAN is installed on the money market trading floor and is used by traders to enter trade information directly into an automated system. This system both captures data for transmittal to the Shearson mainframe (as well as subsequent updates to SLCPI's locally maintained accounting information) and updates a realtime securities inventory file accessible by each trader. The money market LAN also provides a number of analytical tools to assist traders in their assessment of market direction. Government securities, mortgage-backed securities, and the central funding desk are product areas supported by the three separate Sun LANs. The government securities system is similar to that used on the money markets trading floor. Trade information is captured and transmitted to the Shearson host, and the LAN provides a real-time inventory of government securities. The Sun LAN supporting mortgage-backed securities hosts an expert system that facilitates the calculation of complex mortgage-pool allocation algorithms. Finally, the central funding desk Sun system tracks the shortterm finance contracts created by this product area. Each LAN installation represents a separate set of departmental functions with a limited exchange of information across networks.
Although most of SLCPI's systems were developed on microcomputers, there was also a sizable amount of work with mainframe systems. Unlike microcomputer equipment, the mainframes were not owned and operated by SLCPI. The high-end hardware was resident at Shearson's central data center and maintained around-the-clock by MIS operations. Mainframe systems and programming efforts, however, were clearly under the direction of SLCPI Trading Services.
SLCPI established few technical standards during this time to avoid limiting its solution alternatives. SLCPI adopted a high-level computer architecture using UNIX as the standard operating
system, C as the standard programming language, and MIS-sanctioned telecommunication protocols for data going to and from Shearson's mainframe. Information going across SLCPI networks used the mainframe as an exchange hub.
In addition to the applied technology described above, SLCPI began work on a number of pilots that could reasonably be considered "cutting edge." Voice recognition enabled traders to enter real-time trades orally. Image processing was implemented for storage and retrieval of back office information, and an expert system was developed to assist in highly complex and time-critical mortgage allocation calculations.
CLIMATE FOR ACHIEVEMENT
Apart from specific management actions, such as realigning the organization and changing the approach to system development and planning, a number of environmental factors helped to set the stage for success. For a start, Steve Gott was a strong, aggressive leader. He was charged with turning around a highly visible set of problems that were clearly on the priority list of SLCPI's senior management. The SLCPI organization was profitable and powerful enough to ensure that funding was available for required capital and personnel expenditures. As a result, Trading Services was able to recruit systems people familiar with the microcomputer technology base best suited to solving SLCPI's processing problems quickly. They were also able to recruit line managers with demonstrated abilities and familiarity with pertinent data processing tools and techniques. The adage "you get what you pay for" is clearly supported by SLCPI's results since 1986.
SLCPI's processing problems of 1986 were fertile ground for immediate payback opportunities. The problems were real and the return on investment was quick and high. For example, lowering the transaction error rate through system validation and reporting techniques reduced the receivables balance and translated into significant savings. Likewise, eliminating a bottleneck of data entry by distributing the work from a central site to user areas reduced overtime demands. Project results could be defined in the short term, and there was a sense of achievement. Senior management could see results almost immediately. The poor perception of central MIS support was also firmly implanted and provided the impetus to bring systems and development in-house to SLCPI operations.
RESULTS AND FUTURE DIRECTIONS
SLCPI has successfully made the transition from a firm struggling to keep its doors open to one that is viable, profitable, and poised for future growth. The firm has supported a 15 percent annual rate of growth each year between 1986 and 1988 without adding additional support personnel, and it projects that it can handle at least three times the current transaction volume with minimal errors. In addition to accommodating industry growth, the organization has eliminated processing bottlenecks through the combination of changes in work flows and the simultaneous addition of automation. Structured, reliable information is now available to ensure monitoring of such things as backlogged transactions and to manage the firm's balance sheet. The product development cycle has also been significantly shortened, and new products are brought to market within one to two weeks of product definition. The formerly error-prone, tedious, and labor-intensive processing has been replaced by streamlined, redefined procedures and automated support tools. Overtime has been significantly reduced, transaction backlogs—once a primary concern to senior management—have been brought down to a manageable level, and aged receivables are no longer relevant.
SLCPI's systems installations have been both comprehensive and successful. The organization came from a point where it had limited systems support, no PC development, and no user-driven requirements. The organization shared a limited understanding of systems. Since then, more than 500 terminals across five separate local area networks have been installed and are in use. The major system development cycle from project inception to production (including both hardware and software) has typically been between 3 and 10 months.
SLCPI's general long-term direction is to move toward a completely electronic environment that provides high-quality, realtime information and is insensitive to daily fluctuations in volume. The microcomputer-based technology now installed provides a stable hardware platform on which to build the new applications and move them closer to the objective of total automation. SLCPI also has made considerable headway in exploring the potential of emerging technologies such as imaging, voice recognition, and artificial intelligence. It has approached the evaluation of these technologies in much the same way as when it began to use microcomputers. The firm is primarily concerned with business problems and how each proposed technology translates into a solution.