Three experts put this leading high-end system through its paces.

This issue's review of Summit Systems is a very different and bolder undertaking from previous test-drives. It is the first time Derivatives Strategy has evaluated a state of the art, Unix workstation­based, large-scale front, middle and back office derivatives trading system covering the full range of capital markets instruments. These include swaps, caps and floors, swaptions, bonds, bond options, deposits, FX, futures, FRAs, options and even such complex instruments as diff swaps, reverse floaters and exotic options.

Given the broad dimensions of the Summit system, we recognized in advance that there was no way we'd kick the tires of all the functionalities. Summit management was concerned that a mere magazine would not be able to grasp the range, power and flexibility of their system. At the same time we could not overly rely on company-fed information on system attri-butes. So we arranged for Summit to give a detailed presentation to, and field questions from, three hand-picked reviewers, who'd later get as much time on a Summit model installation as they could spare. The reviewers, we should note, are reporting on their first-time experiences only, not making ultimate judgments of system capability. A final caveat: our reviewers were exposed to Summit 2.4. This June the company plans to release version 2.5, which may or may not contain some features and enrichments that the reviewers felt were lacking.

Users Speak

by Maureen Nevin Duffy

In 1990 Jean de Fontenay, a former executive vice president of Quotient, a leading French financial software provider, founded Summit Systems in New York with several Quotient colleagues. Since then Summit Systems has been adopted by some 30 financial institutions in the U.S. and abroad. One of its major selling claims is that Summit is better than its rivals at making its system accessible to end-users and allowing them a strong hand in feature design and modification. Says de Fontenay, who speaks with a mild French accent, "We sell to the user, while some of our competitors sell to the programmer."

Summit is far from alone in the market terrain between off-the-shelf packages and full-blown customization of comprehensive front and middle office systems. Most rival package vendors in derivatives systems have come up with a tool kit or s vlibrary containing ready-made objects, interfaces or other programs that free clients to customize applications. Each has its distinctive approach: SunGard's Renaissance Software's adjustable modules, C*ATS's adaptable database structure and Infinity Financial Technology's class libraries.

Summit Systems' blend of standardization and customization is achieved via a client-server architecture with global capabilities, containing five communication layers (or sets of protocols) to allow clients' tailoring: the applications, API-extension, toolkits, data model and meta-data layers. APIs (application programming interfaces) allows users to extend customized functions, like proprietary pricing models. The toolkit is the means for adding object-based structures and functions. The data model allows for database extensions. And the meta-data layer permits all the adjustments in the structure and tables of the data model to be dynamically maintained in the central database. Thus the master system incorporates all the changes, even changes to screens, for instance, whether made by the client or by the vendor.

The systems capabilities span full analytics, including theoretical pricings, net present value, derivatives statistics, implied values and cash flows-all in a real-time position environment. It also performs flexible interest rate curve generation for all markets, plus sensitivity and hedge analysis at trade and portfolio levels, profit-and-loss reckoning, and automatic trades processing for rate resettings, interest payments and maturities.

Summit has shown a tendency to go in for deep developmental partnerships with clients. Two of its publicized alliances have been with Lehman Brothers and Toronto Dominion Bank (TDB). Lehman declined to answer our questions about Summit. TDB, where Summit worked on extensions of its product for credit risk management and RAROC (risk-adjusted return on capital) was more forthcoming. "We use Summit version 2.4 for our interest rate swap and option business," says Steve Tennyson, director of derivatives systems at TDB. "The system is being used to manage traders' positions, to track P&L and to monitor risk exposure. We are confident it will allow us to meet our future business requirements in the front, middle and back offices."

At ING Barings in London, Summit applications have expanded beyond traditional swaps and options to support a full range of financial derivatives, including the bank's entire LDC bond options platform. ING's risk management area has also embraced Summit as its core system. "We're now using it to support all of our asset swaps as well, which is where we do most of our credit trading," says Richard Prager, ING's director in charge of all debt derivatives worldwide.

Asserts Prager: "Summit allows us to be flexible. Whether we develop further applications within the system, or outside it and then have the [new application] talk to the system through an API, or whatever, [Summit] offers us the ultimate in flexibility. We also didn't want to get locked into any particular technical environment."

Nor does Prager believe that a vendor should be all things to the client. "That is not the best way to leverage the expertise of a vendor," he says. Consequently ING has built its own internal capabilities in the more mundane tasks, such as database management and report writing. For speed's sake ING will sometimes build a prototype outside the Summit system and then hand it over to Summit to be written into the Summit core. But ING also has the option to leave the changes outside the system, in which case it would change only the properties of the object.

For most core products ING deploys the same Summit system in the front and back offices. Summit is used as a subledger for feeding into its own general ledger. "It provides a ticketless environment and a reduction in manual intervention," says Prager.

Currently Summit is installed at some 70­100 sites. These include Crédit Lyonnais, Crédit Suisse Financial Products, Bank Indosuez, Hill Samuel, West Merchant Bank (the London subsidiary of Westdeutsche Landesbank) and Caisse des Dépôts et Consignations (CDC), to name but a few. That's proof indeed of the market acceptance of Summit's relational database and object-oriented design skills in the integration of front and middle office routines.

The Best of Both Worlds: Turnkey and Customization

A substantial part of my consulting business involves reviews of proprietary and third-party pricing tools and risk management systems. This job often involves evaluating the accuracy of those prices and Greeks that the tool, or system, produces for a given contract on the books-including accuracy and/or consistency of the contract, underlying risk factors, algorithms for estimating volatility and yield curves, pricing models, source code, input screens, available market data and, finally, the way traders actually use particular models. Such an evaluation usually requires running the software myself, and for a system as large and complex as Summit's this would take perhaps several weeks. It wasn't practical to spend that kind of time on this review. However, I was exposed to a very thorough presentation by Summit management, and subsequently had some all-too-brief personal access time on the system at the company's site.

The pricing and curve APIs

Although I wasn't able to use Summit's application programming interfaces (APIs) during my brief visit, conceptually they are among my favorite Summit features. There are three APIs in its toolkit: pricing API, curve API and data model API, which clients can use to customize Summit without Summit's help or even its knowledge. The pricing API facilitates creating new pricing models, as does the curve API for yield and volatility curves. If these work as advertised, then Summit offers the best of both worlds-a turnkey system plus tools that allow you to customize it.

Summit management says that APIs are available at each product and curve in the Summit system. At these points Summit delivers a prototype called "null API." The user must then (1) write C or C++ code that does the job he wants and conforms with the prototype, and (2) change the value of a flag variable to tell the system to use the proprietary API, rather than Summit's standard component. Conforming with the prototype consists largely of using the same inputs and outputs in precisely the same form and order. Null APIs are widely available for interest rate derivatives, less available for equity derivatives and, in release 2.4, not available for currency derivatives.

A user might employ these APIs extensively under the following conditions:

• To add new input screens, pricing models and risk reports without asking Summit to create them.
• To transfer new models off spreadsheets, which are notoriously easy to manipulate and difficult to control and audit.
• Where the user might feel that using Summit's standard models could put them at a competitive disadvantage, since rival traders might know too much about how the standard model performs under various market conditions.
• Where the customer has developed, or can acquire, models that Summit lacks, or thinks he has a better version of one of their offerings. I should add, however, that Summit boasts many models with excellent potential.
• Where the customer may want to know the nuts and bolts of a model, in order to be able to optimize code when he finds the opportunity for theoretical improvement.

Do these APIs in fact work as advertised? According to Summit, Commerzbank has used them to introduce some advanced models into their Summit installation, and Andrew Morton of Lehman Brothers has used APIs to bring his implementation of the Heath­Jarrow­Morton model into Summit.

Multiple models for any given deal

The input-output screen for any deal contains a button labeled "models." You can price the deal with any model whose name appears in black, but cannot use a model whose name appears in gray. You can switch from one pricing model to another in under ten seconds, with four clicks of the mouse. Of course, this assumes that you have done certain preliminary work, such as calibrating the HullWhite models.

The formula editor

A user who can decompose his customized derivative product into a sequence of cash flows may be able to use the formula editor. With it one can decompose a vanilla interest-rate swap into a fixed leg and a floating leg, and each leg into a sequence of payments on different dates. The formula editor allows the user to define a sequence of dates and a certain, or expected, cash flow for each date. He can define the cash flow in terms of the four standard arithmetic operations, max, min and many other functions that Summit has defined. Such functions include the CMS (constant maturity swap), quanto, digital and barrier option functions. The Summit system then simply present-values each expected cash flow and sums the present values to determine the price of the customized derivative product.

Summit executives say that they have implemented the formula editor fully for interest rate derivatives. It is present, but not ordinarily appropriate, for bond options. It is also present to a limited degree for equity derivatives. It is not available for currency derivatives, except in the sense that one can use it to introduce exchange rates into payoffs for interest-rate and equity derivatives.

A word of caution deserves emphasis: the pricing model of the formula editor replaces a risk factor with its forward value at each appearance, then present-values the forward value. Of course, Jensen's inequality tells us that this approach is not precisely correct. Thus Summit would produce a biased price for a non-linear payoff, such as one depending on LIBOR squared. Summit is well aware of this and says as much in the system's documentation. The company also says that their typical customers' quants recognize and understand the problem well.

A personal suggestion for a system improvement: that Summit at some future time consider permitting a random payoff in the context of their Hull­White trinomial model. This would appear to be a relatively elementary, mechanical exercise.

Other comments: caps and floors

I looked at how Summit's three available models priced a specific four-year cap struck at the "break-even" yield (the four-year, quarterly compounded par swap coupon). After calibrating the Hull­White models to Black's model for at-the-money caplets, the results were as I report in Table 1. I thought the prices were reasonably close (within six basis points of underlying notional) for three different models. Still, the differences point up how much of the profit-and-loss figures across Wall Street may be an illusion when the buyer's model overvalues the product while the seller's model undervalues it. Of course, the user would want to verify the prices for himself. The user should be aware that Summit reports the HullWhite analytic delta and vol are (1) not comparable to the Black­Scholes delta and vol because the two models differ and (2) are not useful for hedging. The Hull­White trinomial model is not integrated in the hedging algorithms. For the Hull­White trinomial model, price was a step function of the number of trinomial periods. Emannuel Fruchard, Summit's head of financial engineering, confirmed that this indicated that Summit's pricing algorithm uses a number of trinomial periods that depend on the specific deal, as well as on the number of periods the user specifies. Thus the user may input a larger number of trinomial periods, but Summit may not use it in calculating a new price.

Public and private data sets

I like the idea that a system administrator can limit write-access to "official" data to selected users. Meanwhile an individual user can define his own data set and use it for private purposes. Thus the controllers might have sole access to the curves they use for marking the books, while each trader can define his own curves for finding value in the marketplace. I wasn't able to see this feature in action during my demo, because Summit runs its demos in a controlled environment in order to avoid extraneous complications.

Conclusion

Summit's release 2.4 appears to be a turnkey front-end system ready to help fixed income derivatives desks. The back office system that will be incorporated into version 2.5 will greatly improve its value. Summit appears to have front office capability for some equity and currency derivatives. Clearly Summit has the incentive to complete that process. The future timing of coverage of commodity and mortgage-backed securities derivatives seems uncertain.

Summit's vision of the future impressed me greatly. In essence they plan to give their customers the best of both worlds: a turnkey system plus the tools that allow proprietary customization. If they continue to execute their plan well, many derivatives dealers may be able to use this as a substitute for their proprietary front and back office systems for derivatives based on interest rates and prices of bonds, commodities, currencies and equities. Coverage of currency derivatives is not as far along as fixed incomes, but I see a lot of potential.

William Margrabe Ph.D. started in derivatives two decades ago as Fischer Black's research and teaching assistant at the University of Chicago. Between 1987 and 1994, when he founded the William Margrabe Group, he developed models at various positions with Morgan Stanley, Salomon Brothers and Bankers Trust. He has taught at the Wharton School and NYU's Stern School and also lectured at the Federal Reserve Board.

Robust Functions Plus Powerful Customization

Summit is a comprehensive derivatives trading application with robust functionality for the front and middle office. The company's plans in back office, treasury and risk management should greatly improve Summit's functionality in future releases. This turnkey package addresses the most common user requirements and provides for flexible customization as well.

Front office

The system is strongest in the front office; it has a friendly graphical user interface (GUI) with the system. Screens are well-integrated and allow the user to view selected trades or books easily, to price new deals and to track market data and analyze hedging alternatives. However, user reactions to the GUI will very likely vary. Some may find that too many screens are required for some tasks.

Summit provides a choice of valuation models, including Black­Scholes variants and Hull­White models, in closed form and trinomial tree implementations that are suitable for pricing a wide range of derivative products. Sophisticated users with more exotic structures can apply the financial modeling API to easily implement proprietary models in real time without recompiling, as is necessary with some rival systems. Summit includes tools to calibrate model pricing and volatilities among different pricing models and between flat and forward volatilities. There is also an implied pricer that was quite helpful in the calibration of exotic structures with multiple legs. These tools and their flexibility have been battle-tested on major dealer trading floors and thus are among the better tools that are currently available in the package marketplace.

Flexible hedge analytics allow user definition of the universe of hedge products and selection of hedge population. The analytics scenarios include parallel shifts, perturbations and curve hedges based on a variety of shifts. However, version 2.4 does not include the flexible hedge optimization analytics that would be required to manage a large dealer option book efficiently. Analytics do include standard Treasury/ALM functionality for GAP, liquidity and cost-to-close analyses and customizable risk matrices. However, Treasury functionality for FX is currently limited relative to competitor systems.

Back office/accounting

Back office functionality is basic, consistent with Summit's evolution from a trading floor system. For example, confirmation support is not provided and settlement support does not include SWIFT payment/messaging. Moreover, the work flow processing currently lacks flexibility for a variety of processing environments. It requires a predetermined front-to-back work flow. Also, Summit does not provide back office functionality for securities, or for exchange-traded futures or options. However, the system does provide flexible, real-time accounting, using token-based architecture, which allows for a variety of G/L packages. Since Summit does not directly support a multicurrency trial balance (event postings are in a single currency), customization at the G/L interface may be required. Users, however, may define their own chart of accounts.

Architecture

Summit uses proven technology (Unix/C/C++/Motif/Sybase) to provide a turnkey application that is easily customized through the meta-data layer for database changes, the GUI builder and the financial modeling API. The technology is accommodating across a range of IT strategies and is notably strong where rapid implementation is critical.

Despite these merits, I do have some architecture concerns. The replication support that would normally be required for some institutions to run global trading operations could require substantial implementation efforts by the purchaser. Also, Summit uses essentially a two-tier client server implementation with a relatively "fat" client, thus requiring somewhat faster (and more expensive) client machines.

Looking ahead-upgrades

According to company sources, version 2.5 will offer significant back office functionality, fruits of a joint development with Credit Suisse Financial Products. Key features will include a strong settlements suite, with settlement advices and SWIFT payment/messaging; ISDA confirmation support; flexible work flow processing and more robust security controls. This functionality will move Summit from just covering the trading desk to providing more credible back office support similar to legacy swap dealer systems.

Architecturally, Summit has blueprints for significant enhancements. It is developing Oracle database support, which will provide its users with a broader choice among the top industry-accepted databases. To replace its two-tier architecture, Summit will introduce a three-tier client server implementation. This will provide a "thinner" client and thus improve performance. The use of faster multiprocessor application servers and compute clusters to handle computation-intensive chores, such as pricing, hedging and scenarios, will lighten the load on the client machine. An application server will run the core program, a database server will provide all deal and other data, while cheaper PCs will act as user interfaces (clients). This should reduce implementation and operating costs. Summit's planned Windows NT implementation also will not be just a simple port, but through object linking and embedding (OLE) will enable other applications to use Summit components as objects, which may facilitate powerful extensions of Summit into related business unit and enterprise applications.

Summary

Summit is a leading derivatives trading application which has earned a place on many potential customers' short lists thanks to strong functionality, especially for the front and middle office. Summit has positioned its product as an extendible turnkey application which provides significant functionality out of the box, rather than through the toolkit approach of its main competitor, Infinity Financial Technology. The advantages of the Summit approach include quicker implementation and lower IT support requirements. The primary disadvantage is that in some applications users will experience less flexibility due to greater reliance on the vendor, despite the extendibility features.

Edward T. Hida II serves some of Deloitte & Touche's largest derivatives and security dealer clients. He has substantial consulting experience in financial risk management, credit risk management, derivative financial instrument evaluation, information systems and operations, and internal control.

Smooth Hierarchy of Building Blocks

I received a presentation by Summit personnel in a half-day training session-hardly enough time to see all the system's facets completely, let alone offer material for a detailed review. Summit provides complete front office to back office functionality for derivatives trading, including trade capture, analytics, real-time price and rate feeds, sensitivity and hedge analysis, P&L, reporting and accounting.

Some aspects of the Summit functionality that I didn't get around to weighing were treasury products, credit reporting, general ledger accounting interface, interfaces to market data, curve hedging, hedge analysis (parallel shifts, perturbations), curve construction, middle office functionality, system configuration and equity derivatives.

Overview

Where Summit has set itself apart from its competitors is by their solutions-based approach, object-oriented design, API toolkits and intelligible extendibility. Summit's architecture resembles building blocks that are hierarchically structured and intuitive. This allows Summit clients to extend the system's functionality without having to recode when adapting to future versions of the software. Summit also provides a formula builder that allows a trader to build custom calculations for rates, spreads and prices. This solution approach differs from other vendors' toolkit approaches, which usually require extensive coding and take up more time and resources to match Summit's degree of extendibility. Summit's building block approach and API toolkit allows for maximum extensibility and for future upgrades to overlay with customers' extensions.

The rest of my comments will focus on two aspects of this very large and powerful system: its management reporting and value-at-risk components.

Value-at-risk

Over the past few years a number of high-profile companies have sustained major losses in the derivatives market. As a result value-at-risk (VAR) has become a popular standard by which to measure portfolio risks. VAR attempts to quantify risk in terms of the maximum potential loss over a given period of time within a specific confidence interval. In response to this need Summit has incorporated into their new suite of applications a number of VAR analytic programs and reports. Summit's VAR is based on JP Morgan's RiskMetrics, which is a correlation-based approach.

Summit version 2.5 goes beyond the core RiskMetrics parametric VAR calculations and extends their application to include convexity adjustment, plus historical simulation going back 500 days, and gamma and vega analytics for option-related products. Gamma and vega calculations are performed by delta-hedging the portfolios and then running scenarios against the portfolio, stressing volatilities and the underlying. The new analytics include:

• Basis VAR-C.P. versus LIBOR
• Interest Rate VAR
• Asset Trading VAR-Price-Based
• Interest-Rate Option VAR-Gamma and Vega

Summit's upcoming new VAR enhancement will be customizable, allowing the user to define the volatilities and correlations. An additional feature of Summit's VAR is the ability to backtest by tracking VAR predictions against actual results. Summit plans to continue enhancing its VAR analytics by evaluating portfolios with various scenarios. These models tend to be computationally intensive, and as a result Summit is looking into capturing the majority of risk with the Greek letters. For common groups of transactions, Summit will produce Greek letters and then perform the VAR calculations on them, reducing computation time.

Reporting

The Summit reporting suite comes with standard reports and the ability to filter records, by any number of criteria, when generating all of its standardized reports. For example, a trader could generate a daily trade journal by book, product or notional size. Summit's reporting suite provides reports for trade management, risk management, customer advices, P&L and credit analysis.

Summit could improve on its reporting suite by offering a GUI-based screen similar to those throughout the rest of the system, which allow the user to create ad hoc reports easily. At present when the user wants to build a new report, he or she must go into the report window and change the parameters through the options menu. This is somewhat cumbersome and clumsy.

In recognition of this problem version 2.5 will allow users to set and store reporting parameters through a GUI screen. Through Summit's API, clients can also create a wide array of customized analytic, operational control and P&L reports to meet their exact business requirements. This API supports several popular third-party-based report writers.

Paul A. Quirk Jr. began his derivatives career at Presidential Global Funding as manager of treasury operations, including interest rate swaps, options, and futures and FX options. Later, as a vice president of operations for the Bank of Boston, his responsibilities involved the complete automation of all spreadsheet and analytic tools.