Prospects for Operations Research in the E-Business Era

Copyright � 2001 INFORMS0092-2102/01/3102/0006/$05.001526–551X electronic ISSN

PROFESSIONAL—COMMENTS ONCOMPUTERS—E-BUSINESSCOMPUTERS—INTERNET

This paper was refereed.

INTERFACES 31: 2 March–April 2001 (pp. 6–36)

Prospects for Operations Research in theE-Business Era

Arthur M. [email protected]

Decisions, Operations, and TechnologyManagement

John E. Anderson Graduate School of ManagementUniversity of California, Los AngelesBox 951481Los Angeles, California 90095-1481

Ramayya [email protected]

Heinz School of Public Policy and ManagementCarnegie Mellon UniversityPittsburgh, Pennsylvania 15213

The digital economy is creating abundant opportunities foroperations research (OR) applications. Several factions of theprofession are beginning to respond aggressively, leading tonotable successes in such areas as financial services, electronicmarkets, network infrastructure, packaged OR-software tools,supply-chain management, and travel-related services. BecauseOR is well matched to the needs of the digital economy in cer-tain ways and because certain enabling conditions are comingto pass, prospects are good for OR to team with related ana-lytic technologies and join information technology as a vitalengine of further development for the digital economy. ORprofessionals should prepare for a future in which most busi-nesses will be e-businesses.

We may be living through the busi-ness equivalent of the Cambrian

explosion when, after 3.5 billion years ofsluggish evolution, a vast array of lifeforms suddenly appeared in only 10 mil-lion years. New business models andways of organizing and operating busi-nesses are appearing in comparably rapid

profusion, driven by stunning advances inthe information technologies.

As managers pursue organizationalchange on a massive scale, they have fo-cused to the point of preoccupation on theopportunities presented by modern infor-mation technology. The need for more de-cision technology has been growing all the

PROSPECTS FOR OPERATIONS RESEARCH

March–April 2001 7

while, but only recently have competitivepressures begun to build for moreapplications.

Fortunately, new data sources, opera-tional business software, intermediaries,and data-exchange standards are emerg-ing that are very conducive to decisiontechnology. Thus the digital economy isentering a phase in which decision tech-nology seems poised to rise and join infor-mation technology as a vital engine of fur-ther development.

What kinds of opportunities does thedawning e-business era offer OR, and howhas OR fared in taking them up? We haveinquired widely in academia and in indus-try over an extended period. We find that,after a slow start perhaps obscured some-what by secret projects (we encounteredmore than a dozen), OR practitioners arenow beginning to capture some of theirabundant opportunities.

Here are some examples, and we shallrecount many more in their proper places.—Hewlett-Packard Company uses realoption modeling for dynamic pricing andfor managing supply/demand risk as aparticipant in High-Tech Exchange, aprocurement-oriented electronic marketfor companies in computing and relatedindustries [Billington 2000].—MarketSwitch uses constrained optimi-zation to make ad- and promotion-placement decisions for Internet adver-tising networks and e-commerce sites.Its software takes into account the ad-promotion pool, associated advertisingand partner contracts, and other con-straints, working in real time as customersinteract with Web sites. (Underlinedphrases correspond to URLs, compiled at

the end.)—OptiBid, an integer-programming-based,multiattribute, Internet-enabled, combina-torial auction system, is used by dozens oflarge shippers to select transportation pro-viders in an online marketplace.—Strategic Data Corp. uses segmentationbased on hierarchical clustering, a rule-discovery data miner [Cooper andGiuffrida 2000], and a real-time learningengine to personalize Web-site contentfor its clients.—Trajecta, Inc. uses predictive and sto-chastic optimization models to help banksmanage their credit-card portfolios, in-cluding online applications [Smith 2000].—United Sugars Corporation uses a Web-based architecture to deploy its produc-tion, distribution, and inventory-capacityoptimization application [Cohen, Kelly,and Medaglia 2001].

Market forces will likely increase therate of progress, but we believe OR profes-sionals need to extend their technical rep-ertoire and change some of their ways toapproach their full potential in the newera. They need to—become better informed about applica-tion opportunities in the digital economy;—develop the skills called for by e-business projects, including those relatingto information and communication tech-nology, certain emerging OR topics, andthe design of e-business processes;—actively seek out such projects;—publicize their successes at least throughhouse organs, the trade press, andINFORMS.OR academics will find ample challengesin supporting the first two items with ap-plicable research and education.

GEOFFRION, KRISHNAN

INTERFACES 31:2 8

Figure 1: The e-business landscape has three major parts: the first centers on the needs of endconsumers, the second on the needs of businesses, and the third on the infrastructure needed tosupport the other two. The two-level classifications of the individual parts draw distinctionsneeded to understand this landscape. Numbers in italics are annual revenue estimates in bil-lions of dollars for the US except as noted. We synthesized the business-to-consumer (B2C) andbusiness-to-business (B2B) estimates from Kafka [2000], Putnam [1999], Sanders [2000], andWilliams [1999]. Russell and Keith [2000] provide the advertising estimate; Lawrence [2000] pro-vides the EDI estimate, and Barua and Whinston [2000] provide the network estimates, whichare worldwide Internet-related sales of US-based companies (these numbers are growing veryrapidly and are not directly comparable to the others since they pertain to four-to-five yearsearlier). The decision technology estimate is a worldwide lower bound inferred from Morris[2000]; associated services are worth a small additional integer multiple.

The first item is the main focus of thisarticle. Observations on the second alsooccur throughout, with more pointed com-mentary at the end.

In what follows, we sketch a view of thee-business landscape to provide some per-spective on a very large territory. Next weidentify where OR applications appearmost frequently within this landscape.Then we review applications and opportu-nities in four selected areas: informationgoods and services, supply-chain manage-ment, network infrastructure, and soft-ware tools for decision technology. Basedon this review, we present what we be-lieve are OR’s particular strengths in thecontext of the digital economy and selec-

tively illustrate these with discussions ofcustomer-relationship management andreal-time OR. We conclude by drawingsome actionable implications forpractitioners.The E-Business Landscape

Our view of the e-business landscape(Figure 1) serves as a framework for whatfollows and introduces essential concepts.We necessarily focus on online (mainlyTCP/IP-based) rather than offline activity,but these are converging so fast that it willsoon be regarded as archaic to speak ofone as though separate from the other.TCP/IP (Transmission Control Protocol/Internet Protocol) is the protocol suite onwhich the Internet is based.



The e-business landscape contains twomajor divisions, (1) consumer-oriented ac-tivity and (2) business-oriented activity,both supported by (3) the e-business in-frastructure. Consumer-oriented activ-ity comprises business-to-consumer,government-to-consumer, and consumer-to-consumer activity. Business-oriented ac-tivity comprises business-to-business,business-to-government, and government-to-business activity. We devote little at-tention to government activity, althoughUS government-to-consumer andgovernment-to-business activity amountsto some $450 billion annually in fees andfines (and trillions in taxes), and business-to-government sales of materials and ser-vices totals about $550 billion across alllevels of government [The Economist2000a]. Nor do we consider consumer-to-consumer activity, which thrives via suchconsumer auction sites as eBay, onlineclassified advertising, and portal sites thatfacilitate communication among individ-uals (for example, Web-based e-mail, in-stant messaging, and Web space for indi-viduals and communities). However,commercial consumer-to-consumer inter-mediaries and the fees and commissionsthey collect are part of business-to-consumer activity.

We subdivide consumer-oriented activ-ity according to revenue source: directlyfrom consumers through sales, accesscharges, subscription charges, commis-sions, and the like, or from nonconsumersin return for visibility to consumers (in-cluding sponsors and those who pay forreferrals). Non-revenue-producing onlineactivity can also be relevant to e-business(and to OR [Geoffrion 1998]). For example,

affinity-group Web sites (demographic,geographic, or topical) often contain seri-ous discussion of commercial productsand services and facilitate access to them.

We subdivide business-oriented activityinto that conducted (1) directly betweenthe essential parties, (2) with the help ofintermediaries (for example, electronicmarkets or exchanges), (3) via an Intranet,and (4) via EDI (electronic data inter-change). All but the last use Internet tech-nology exclusively. The first two use eitherthe public Internet or extranets, which aremembers-only networks run by individ-ual organizations, often implemented asvirtual private networks on the public In-ternet. Intranets use secure, intraorganiza-tional, Internet-technology-based networkstypically implemented as private networksor as virtual private networks.

Organizations use intranets for a varietyof internal, often operational, purposes—accessing customer records and internalreports, conducting administrative proce-dures, collaborating, distributing corporateinformation, providing personnel services,training, and automating work flows[King 2000]. They also use them to con-duct internal activity, mainly to transfergoods and services using internal prices(interplant transfer, chargeback for com-puter services, and so forth).

EDI predates Internet-based e-commerceby about two decades and still surpasses itin magnitude but is growing much moreslowly because of its more limited func-tionality, relative inflexibility, and highercosts. Legacy EDI activity will occur formany years, with some of it migrating toextranets, but most new EDI applicationsare implemented using Internet technol-

GEOFFRION, KRISHNAN

INTERFACES 31:2 10

ogy and XML (extensible markuplanguage).

We distinguish between physical goods,like cars and computers, and (digitizable)information goods, like magazines and

Electronic marketplaces coulddevelop into important ORapplication venues.

music. Similarly, there are physical ser-vices, like transportation, and (digital) in-formation services, like news feeds. Whencompletely digitized, information goodsand services incur negligible marginal pro-duction costs after the first unit, no mar-ginal transportation costs (except whencontent-delivery networks are used toserve content from caches), and no mar-ginal inventory costs (except for obsoles-cence costs in some cases). Moreover, theycan be reproduced and moved in unlim-ited quantity almost instantaneously.Physical goods and services do not enjoythese magical qualities. They require sup-ply chains whose annual costs approach$1 trillion in the US [Delaney and Wilson2000].

For physical business goods and ser-vices, we also distinguish between spotand prearranged transactions. The digitaleconomy is greatly increasing the fre-quency of such spot activity as buying andselling capacity and inventory. As Delaneyand Wilson [2000] argue, the stage is setfor an epic struggle between such marketsand traditional ways of doing businessbased on persistent relationships.

A huge e-business infrastructure sup-ports all of this activity. The foundation isthe network infrastructure, which includes

information appliances (networked PCs,personal digital assistants, cell phones,and other devices), servers, access technol-ogies such as cable modem and DSL (digi-tal subscriber line), backbone technologies,fiber optic cable, content delivery net-works, networking hardware and soft-ware, Internet-access and backbone serviceproviders, network storage providers, andsecure commerce infrastructure (for exam-ple, digital certificates). Components aresometimes packaged and offered as infra-structure targeted towards specific mar-kets. An important example is the ASP(application service provider) that oper-ates secure server farms in data centerswith highly available connections to ISP(Internet service provider) backbones.Such ASPs offer a platform for makingsoftware available as a service (Fig-ure 2).

Network applications build on networkinfrastructure to enable e-business. Theseinclude software for Web applications de-velopment, search, and transactions andsuch services as e-business consulting andonline training.

Decision technology adds value to net-work infrastructure and applications bymaking them smarter. We use decisiontechnology inclusively to refer to modelingand solution techniques in the tradition ofclassical operations research, statistics, andparts of mathematical economics, to sym-bolic and numeric reasoning techniquesfrom artificial intelligence, and to the im-plementation of these techniques in soft-ware. In our view, OR needs this breadthin the present era. Within decision tech-nology, we distinguish between applica-tions and the software tools that enable



Figure 2: An application service provider (ASP) offers access to application software runningon an application server via the Internet and a local client program (often a Web browser). Theapplication server has secure access to applications and databases, legacy or otherwise, throughapplication program interfaces (APIs). This is an increasingly popular way for software ven-dors to make their software available as a service.

them. For example, a Web-ready compo-nent library for optimization would be atool, whereas software for online invest-ment portfolio optimization that usesthis component library would be anapplication.

In our extensive search for applied workon the intersection of OR and e-business,we found the most activity in four areas ofthe e-business landscape (Figure 1):—within business-to-consumer andbusiness-to-business commerce in informa-tion goods and services, applications inonline financial services and travel-relatedservices;—within business-to-consumer andbusiness-to-business commerce in physicalgoods and services, applications insupply-chain management and electronic

markets;—within network infrastructure and appli-cations, applications in network designand quality-of-service improvement; and—within decision technology softwaretools, the packaged software component.

The remainder of the landscape mayhave received less attention to date fromOR practitioners, but do not discount itsfuture potential. For example, advertisingand customer relationship managementapplications are heating up rapidly. Thedecision technology applications categoryis a special case since every such applica-tion is associated also with some otherportion of the landscape.

We now review in some detail the fourareas of the e-business landscape, withparticular attention to opportunities.

GEOFFRION, KRISHNAN

INTERFACES 31:2 12

Information Goods and ServicesInformation goods include books, greeting

cards, music, software, video games, andvideotapes. Although these goods are tra-ditionally manufactured and delivered inphysical form, they can be digitized andso need not be materialized for customerswho have the means to receive them digi-tally—a solid state player to displaybooks, a screen to display an electronicgreeting card, and so forth. Over time,digitizable goods will increasingly be de-livered to customers in digitized form—almost completely bypassing, for down-loaded goods, physical manufacture andall that comes afterward [Johnson 2000].Supply chains for information goods inphysical form are therefore doomed to be-come less important.

The main categories of information ser-vices are entertainment, financial, news,telecommunications, and travel, but onlyto the extent that their delivery and busi-ness processes are digital. The ASP indus-try, expected to explode as software be-comes more of a service and less of aproduct [Apfel 2000], cuts across mostcategories. Since administrative, personal,and professional services usually requirepeople for performance, we categorizethem as physical services (an exception isthe Web-based SmartSettle negotiation-support system, which uses mixed integerprogramming at its core and does not re-quire a professional facilitator). In travel,we classify an airline’s direct-sales Website under physical services, since fulfill-ment requires flying airplanes, and agentsunder information services, since they dealonly in tokens for services that others willfulfill.

Based on detailed forecasts by ForresterResearch, we estimate that informationgoods and services, exclusive of EDI, willaccount for roughly one-third of the $187billion business-to-consumer activity pro-jected for 2004, a much smaller fraction ofthe $3 trillion business-to-business goodsprojected for 2004, and roughly half of the$220 billion business-to-business servicesprojected for 2003 (Figure 1).

One can distinguish two kinds of mana-gerial opportunities for OR in connectionwith information goods and services: tohelp design and operate e-business pro-cesses and to analyze business issues of aless recurrent nature.

The processes are of course very differ-ent from those for physical goods and ser-vices, although many analogies can bedrawn: information goods and serviceshave to be created, put in saleable condi-tion, made available for sale, and deliv-ered to customers. Content-delivery net-works and Web caching [Chan et al. 1999]are important parts of such digital supplychains.

Economists have had the most to sayabout business issues. A good example isthe book Information Rules, in whichShapiro and Varian [1998] cover pricing,product versioning and bundling,intellectual-property-rights management,lock-in, network externalities, and other is-sues for information goods and services.They argue that the digital economy is notas novel as often portrayed and that well-established economic ideas can be appliedto craft strategy and deal with common e-business issues. They include no explicitmathematics or models, but it is obviousthat models could be formulated using



standard ideas from economics, especiallyfrom microeconomics (see the simplemathematical models on the InformationRules Web site under “Teaching”).

This book and other economics-basedworks on the digital economy [Bakos 1998;Bossaerts, Fine, and Ledyard 2000; Smith,Bailey, and Brynjolfsson 2000; Vakrat andSeidmann 2000] contain an implied chal-lenge to OR professionals: to attemptmodeling applications that carry economicanalysis of e-business issues in the direc-tion of more detailed, quantitative studiesof particular problems faced by particularcompanies.

Moving from stylized economic modelsto real applications in the OR tradition

OR professionals need toextend their technicalrepertoire.

will require less reliance on convenient as-sumptions, more reliance on computa-tional rather than analytical solutions, andsometimes facility with such tools asMathematica that support symbolic aswell as numeric reasoning. Achieving sucha fusion of the two traditions is a majorchallenge facing OR, and success wouldalso strengthen the strategic capabilities ofOR practice, something not always seen asa hallmark of the field.

OR can also add value to informationgoods and services by direct incorpora-tion. For example, a book or technicalmanual downloaded to a portable displayunit containing a processor could includesome content computed on demand fromstored data and models. Tables andgraphs, for instance, need not be static but

could be computed as needed by a readerwho revises the default parameter settingsof an underlying model. In this way, ORcan become part of books, manuals, andbriefer works in any of its many applica-tion domains. (One enabling technologytoward this end is MathML, a markuplanguage for mathematical processing.)

Packaged software provides a more fa-miliar context for OR to add value. For ex-ample, incorporating OR into geographicinformation system (GIS) software canyield large practical benefits [Weigel andCao 1999]. Leading GIS software vendors,such as ESRI, now offer Internet map serv-ers and OR components that can serve asplatforms for OR applications. For in-stance, with real-time data about truck lo-cations and traffic, GIS and OR alreadyhelp retailers achieve greater efficiencies inproduct delivery [Brown 2000; Partykaand Hall 2000].

Spatial data modeling, which is greatlyfacilitated by GIS, will play an importantrole in mobile commerce (commerce en-abled by mobile information appliances,such as Internet-enabled cell phones andwireless personal digital assistants, whosenumbers are forecast to grow five-foldfrom 1999 to 2003 [Meeker and Mahaney2000]). The next generation of these appli-ances will be location-aware based on theGlobal Positioning System or ground-based networks or a combination of both.Given knowledge of the location of userscarrying these appliances, a range of infor-mation services that exploit the availabilityof spatial maps and OR algorithms arepossible. Two examples are real-time rout-ing to avoid congested roads and geo-graphically specific recommendation

GEOFFRION, KRISHNAN

INTERFACES 31:2 14

systems that consider interests and prefer-ences when suggesting restaurants orother stationary facilities.

“OR Inside” is still in the future formost information goods and services, butit is very much in the present for someInternet-based services.Financial services. OR methods have longplayed an important role in financial ap-plications [Jarrow, Maksimovic, andZiemba 1995; Luenberger 1998]. Examplesof practical successes include the analysisdone by the Management Science Groupat Merrill Lynch Private Client Group on astrategic service-pricing response to thediscount online broker challenge [Nigam,Labe, and Altschuler 2000]; products fromWeb-savvy, OR-based companies, such asAlgorithmics (enterprise-wide financial-risk-management software for market,credit, liquidity and operational risks) andBARRA, Inc. (portfolio analytics andenterprise-risk management); and high-impact work in asset and liability manage-ment at Towers Perrin-Tillinghast[Mulvey, Gould, and Morgan 2000] and inportfolio construction at GMO [Bertsimas,Darnell, and Soucy 1999].

Inevitably, OR-based financial productsand services are finding their way online.Online business financial services is thelargest of the business-services sectors,forecast to expand from $7.3 billion in1999 to $80 billion in 2003 [Putnam 1999].One such service is the Bond Connect on-line marketplace for fixed-income securi-ties, which uses integer programming tohelp match and price trades in a combina-torial auction setting. The novel design,based on work by Bossaerts, Fine, andLedyard [2000], promotes liquidity by per-

mitting contingent cross-market ordersthat reflect whole-portfolio considerations.Originally available only over a proprie-tary network, it is now available over theWeb. Another service is investment-portfolio risk management, in which re-cent advances based on scenario simula-tion enable ASPs (for example, CreditSuisse First Boston) to perform all theheavy model computations while theirinstitutional clients perform the light,portfolio-specific risk calculations withproprietary data [Dembo et al. 2000;

Sadly, many firms found noP in ERP.

Young 2000]. This partitioning of work en-ables one to take advantage of ASP func-tionality while still preserving client confi-dentiality. A similar service is available toindividual investors.

Among consumer services, automatedonline financial advice is forecast to ex-plode in penetration from an estimated 1.8million US households at the end of 2000to 20 million by 2005 [Punishill 2000].

Online advisory financial services in-creasingly embed optimization routines.An early example is Optimal RetirementPlanner from Sundown Software Systems,offered free through an ASP. It relies onlinear programming instead of on simpleformulas and rules as traditional retire-ment planning calculators do. Based oncurrent savings, planned contributions,and desired estate, it formulates a modeland computes a schedule of transfers anddisbursements from retirement accounts tomaximize the amount of annual after-taxmoney available for spending through the



term of retirement.A better-known ASP service for

optimization-based investment advice isFinancial Engines Investment Advisor,which advises on most tax-deferred retire-ment accounts. It offers basic features forfree and more advanced ones for a fee. Ituses quadratic programming for portfoliooptimization and Monte Carlo simulationto forecast probable investment outcomes.Several other firms, including BARRA,Inc., offer portfolio-optimization services,financial-risk management, and invest-ment analytics via the Web.

OR is also finding its way online inlending and in the analysis of electronicmarkets. ILOG, Inc. has developed a real-time constraint-programming application[Lustig and Puget 1999] to online lending,First Union Loan Arranger, for First UnionHome Equity Bank. It takes the customer’spreferences into account and to calculatethe best loan for the customer subject toconstraints imposed by the bank’s loan-approval policies.

Clemons and Weber [1997] applied in-formation economics and computer simu-lation to obtain strategically valuable in-sights into screen-based securities markets,and probably also into emerging electronicmarkets in which highly profitable ac-counts have traditionally subsidizedunprofitable ones (for example, somefinancial services, insurance, and tele-communications markets). The problem,they say (p. 1696), is that “. . . cross-subsidization attracts new entrants withtargeted marketing strategies, based uponopportunistic cream skimming. . . .” TheInternet is famous for facilitating this sortof market efficiency.

Other opportunities to embed OR surelywill follow from the success of online trad-ing and investing [Wall Street Journal 2000],from the trend toward online banking,and eventually from online insurance [TheEconomist 2000b].Travel-related services. Leisure travel(mainly airline tickets, lodging, and rentalcars) is forecast to be about 12 percent ofall business-to-consumer e-commerce in2004 [Williams 1999], a share considerablylarger even than the total of computerhardware and consumer electronics. Busi-ness travel is forecast to be about 17 per-cent of all online business services in 2003[Putnam 1999], which is about half ofbusiness financial services. Moreover,business travel is forecast to have by farthe largest penetration rate among themajor groups of business-to-business e-commerce services: about 32 percent of to-tal business travel (online or offline). Ourfocus here is on the information servicesrelated to travel, not the physical services.

The airline industry pioneered e-commerce about four decades prior to theadvent of the Web, deploying some of theworld’s first business-to-business systems.Thus, OR practitioners can look to the air-lines for experience, technology, and ap-plications that may be adaptable to othere-business domains. The discussion bySmith et al. [2001] of Sabre’s use of OR inthe airline industry is instructive even be-yond travel-related information services:(1) The Sabre computer reservation sys-tem was “the first real-time business appli-cation of computer technology.” This sys-tem and its competitors formed the basisfor today’s global distribution systems forreservations, which support multiple dis-

GEOFFRION, KRISHNAN

INTERFACES 31:2 16

tribution channels and vendors in manytravel-related industries. They are similarin function to, and predated by more than20 years, the vertical Web portals cur-rently the focus of frenzied activity inmany industries.(2) Sabre’s low-fare search engines havebeen used by travel agents since the mid-80s. The enormous user-base expansionthat followed public access to them via theWeb increased both the number and thekinds of database queries to be processed,forcing a substantial redevelopment effortto deal with the associated optimizationproblems. Sabre’s experience offers scal-ability lessons for other OR applicationsthat must migrate to the Web and possiblyalso for Web shopbots—programs that con-sumers use to search the Internet forgoods and services meeting low-price andother criteria.(3) The display-ordering problem is fun-damentally important to dynamic Web

E-prefixes will become passeas online and offlineeconomies fuse.

page design, especially for the results ofshopbots, recommendation systems, andsearch engines. The essence of this prob-lem is how to order the results to promotesome objective—such as to favor someproduct or vendor or to direct traffic tosome Web site—to the maximum extentpermissible under regulatory, ethical, andmarket constraints. This can be formulatedas an optimization problem if one canmodel how display order influencesviewer preferences. Sabre’s success withsuch applications should encourage others

to tackle similar opportunities in other e-business contexts.

(4) Yield management, originally ablockbuster airline application, is dif-fusing rapidly throughout the hospital-ity and transportation industries and be-yond (including business-to-business) ase-mail and the Web speed up communica-tion with customers. These technologiesallow companies to change prices—cus-tomized to individuals or groups—glob-ally and instantly as commitment dead-lines approach. They also enable rapidcustomer responses to price changes andsubsequent real-time or near-real-timenegotiations.

Online leaders are integrating yield- andrevenue-management systems with corebusiness processes in such areas as last-minute travel deals, natural gas trading,and ticket sales. For example, OptiYield-RT is a real-time yield-management sys-tem for truckload carriers based on loaddata from the carrier’s order-entry system,demand data from a cooperating Internet-based exchange, current truck locationsaccording to global-positioning tech-nology, and other data. Another yield-management product, NeoYield fromNeoModal.com, is designed for ocean car-riers to use on an ASP basis. It uses de-mand information from carrier forecasts,network-capacity data (vessel space, con-tainer availability, and terminal births),and carrier-cost data to decide what de-mand to satisfy at what price and howbest to reposition empty containers[Hingorani 2000].

In addition to the well-developed solu-tion techniques for perishable productsand services pioneered by such airline-



bred specialists as PROS Revenue Man-agement, Inc., Sabre Inc., and Talus Solu-tions (now part of Manugistics, Inc.), thereare techniques for other formulations ofthe dynamic-pricing problem [Andrews1999; Kephart, Hanson, and Greenwald2000]. Dynamic pricing will play a crucialrole in many online markets, including on-line auctions [Lucking-Riley 2000], and ORtechnology is very useful for designing as-sociated business processes. For example,Home Depot uses integer programming inan Internet-based combinatorial biddingapplication for contracting transportation-carrier capacity [Keskinocak and Tayur2001].Supply-Chain Management

Physical goods and services make upthe lion’s share of the e-commerce forecastfor 2004: roughly two thirds of the pro-jected $187 billion business-to-consumeractivity and a much larger fraction of theprojected $3 trillion business-to-businessactivity (Figure 1). If supply-chain costs(inventory, transportation, warehousing,and logistics administration) consumeanywhere near the same proportion of theonline share of the gross domestic productas they do of the offline share, about 10percent [Delaney and Wilson 2000], sup-ply chains will play a crucial role in thedigital economy.

OR-powered supply-chain managementwill be important in the online economy,as it has been in the offline economy, forstrategy, tactics, and operations. The digi-tal economy provides further reasons touse OR, including these:(1) Enterprise resource planning (ERP)and advanced planning and scheduling(APS) vendors are striving to become In-

ternet centric and to meet brisk demandfor better planning and scheduling capa-bilities within a context that is rapidly ex-panding to include e-business.(2) Business-to-business electronic marketsare growing rapidly, with spot marketsflourishing and dynamic pricing becomingpandemic.(3) The Internet is stimulating much moresupply-chain coordination and collabora-tion with suppliers and customers.(4) The Internet is stimulating expansionof supply chains to include product-designand customer-relationship management.(5) The trend toward mobile communica-tions and computing in the trucking in-dustry is accelerating Internet-baseddecision-support applications in trucking.(6) The Web greatly facilitates deployinghighly interactive applications.

These are some of the ways in which thedigital economy adds to the importance,opportunity, or effectiveness of OR-drivensupply-chain management. They warrantelaboration.(1) ERP systems have been implementedby roughly half of all large companies inthe US and Europe, are spreading rapidlyto midsize firms, and are incorporating e-business functionality. Nearly $10 billionis spent annually on ERP software licensesin the US, and still more on implementa-tion and associated services. Sadly, manyfirms found essentially no P in ERP. Thesesystems focused on transactions, with littleprovision for planning and scheduling.Worse, they often obliterated valuable ORapplications.

A brisk demand for supplemental APSsoftware developed. Optimization becamea panacea to such an extent that the word

GEOFFRION, KRISHNAN

INTERFACES 31:2 18

is now debased to mean merely “try toimprove.” Although everyone wants “op-timal” plans and schedules, most planningand scheduling software from APS andERP vendors produces solutions that areworkable but often far from optimal. Opti-mality or something close to it looks like along-term objective, and that is an endur-ing opportunity for OR. ERP vendors(most now with APS capability) have beenscrambling to incorporate better planningand scheduling into their products, as wellas Internet-enabled connectivity (for exam-ple, with EDI and XML-based document-exchange standards) and greater inter-operability with third-party software[Sodhi 2001; Sprott 2000].

An important legacy of the ERP move-ment is that it made available far morehigh-quality transactional data for model-ing purposes. OR practitioners can tap thisdata, but connecting model-based applica-tions to ERP systems requires using con-temporary, widely adopted application-program interfaces [Sprott 2000]. This canbe done. OptiFlow, for example, interfaceswith ERP, APS, and transportation-management systems; it optimizes routingand load consolidation for trucking com-panies each night, using the Web to obtaindata and to distribute solutions.(2) Electronic markets are forecast to ac-count for about half of business-to-business e-commerce trade in 2004 [Kafka2000]. Online markets (these exclude bilat-eral trade between pairs of trading part-ners) often take the form of auctions andexchanges hosted by buyers, sellers, or in-termediaries [Bakos 1998; Kaplan andSawhney 2000; Lucking-Riley 2000]. Theysupport prearranged and spot buying and

selling. In spot trading, there is a strongtrend toward transaction prices thatchange in response to the latest informa-tion about vendors (for example, capacity,demand, inventory, and scheduling) andabout buyers (for example, bidding behav-ior or “clickstream” history). (A clickstreamis a navigational path.) This creates oppor-tunities for analysis and for model-basedsystems to set prices, bids, and negotiatingpositions rationally. Keskinocak and Tayur[2001] explain some of these opportunitiesin a way that recognizes the roles of botheconomics and OR.

Economic theory is having a major prac-tical impact on online market design; forexamples, see Market Design Inc. (auctionmarket design) and Perfect.com (RFQ pro-cess automation for exchanges). Economicsbecomes stronger yet when more OR tech-nology is added, as in the followingexamples:—Beam, Segev, and Shanthikumar [1999]use Markov chains and optimization todesign multi-unit auctions taking into ac-count lot sizing and pricing.—Ishikida et al. [2000] use integer pro-gramming for deciding which offers to ac-cept in their design for a combined-valuecall market (a kind of combinatorial auc-tion) for trading air-emission permits. Thisdesign formed the basis for California’sAutomated Credit Exchange, which hasbeen operating successfully since 1995 asthe world’s first Internet-based business-to-business exchange. Subsequently, simi-lar technology has been used to buildother online, combined-value call markets(Net Exchange).—Vakrat, Pinker, and Seidmann [2000] usedynamic programming and Bayesian



learning to obtain methods for designingoptimal multiperiod, multiunit auctions.—Vakrat and Seidmann [2000] presentan empirical study of real business-to-consumer auction data, which they use tomotivate an economic analysis aimed atproviding a theoretical basis for their ob-servations and guidance for auctiondesign.

There are indications of an importanttrend among electronic market makers:that many will provide OR functionality toparticipating firms because this can addsubstantial value. Such functionality couldinclude optimal auction design and bid-ding services, individual and collaborativesupply-chain management, dynamic pric-ing services, and MCDM (multicriteriondecision-making) aids for procurement.Electronic marketplaces already offeringOR functionality include Digital Transpor-tation Marketplace from Logistics.com,FreightWise and eConnections (built withManugistics, Inc.’s ExchangeWORKS elec-tronic exchange platform), mysap.com,and marketplaces built using the i2TradeMatrix suite of components. Further-more, since some OR applications are bet-ter run by an ASP than by the companythey serve [Sodhi 2001], electronic market-places may pick up additional applicationsof this sort.

If OR functionality becomes wide-spread, electronic marketplaces could de-velop into one of the most important of allOR application venues. At a minimum, weexpect them to become an important newsource of market data and industry statis-tics with which knowledgeable participat-ing firms can build their own models.(3) Internet connectivity is causing

supply-chain management to expand inthe direction of suppliers and customers,thereby causing the physical scope ofsupply-chain models to expand [Sodhi2001]. For example, the Internet makesvendor-managed inventory more practical,thereby inviting extension of inventorymodels and software. For another exam-ple, the availability of data about capacity,inventory, and demand from trading part-ners’ ERP systems in business-to-businessmarketplaces sets the stage for integer pro-gramming to perform multiway matchingof supply and demand for manufacturingcapacity and inventory [Keskinocak andTayur 2001].(4) The Internet also facilitates expandingsupply-chain management in the directionof product design, sales, and customer-relationship management (CRM) [Sodhi2001]. Again supply-chain models must beexpanded, this time with respect to theirfunctional scope.

Keskinocak and Tayur [2001] and Sodhi[2001] elaborate further on these fourpoints and cite companies that are exploit-ing these OR opportunities.(5) Fuel is a major operating cost in thehalf-trillion-dollar trucking industry, mak-ing fuel-purchase optimization important.ProMiles Online was the first to offer thisservice online, with an ASP applicationthat makes the daily updating of fuelprices transparent to users. Another ASPservice, OptiStop, combines this applica-tion with optimal routing. It uses the Webto reach a carrier’s system, which in turnreaches the mobile satellite communicationterminal that many trucks carry; alterna-tively, owner-operators can use OptiStopdirectly via a Web-based interface.

GEOFFRION, KRISHNAN

INTERFACES 31:2 20

Such OR applications are likely to comeinto general use. Computers are alreadycommon in truck cabs (about 20 percent in2000), truck stops are increasingly com-puter friendly, and most trucks shouldsoon have mobile Internet connections andlocation-finding equipment. The latter en-ables new rendezvous tactics for swappingtrailers or parts of loads away from fixeddistribution facilities, which poses newmodeling challenges.(6) The Web makes it inexpensive for thefirst time to deliver highly interactive ORapplications at the point of need on aglobal scale. For example, it is now practi-cal to make MCDM methods availablewith the interactivity needed for such ap-plications as procurement. FrictionlessCommerce offers software that incorpo-rates such methods tailored to business-to-business purchasing for maintenance,repair, and operating supplies. Severalcompanies offer such methods forbusiness-to-consumer applications, includ-ing financial services. For another exam-ple, an LP-based, ASP-hosted contractnegotiation tool like NeoSelect fromNeoModal.com would be impractical todeploy other than online because of thenegotiations and many cost-service trade-off scenarios needing solution during atypical global ocean shipper’s contractingperiod, which may involve hundreds ofprospective trading partners [Hingorani2000].Network Infrastructure

The meteoric growth of the Internet iswell documented [Zakon 2000], compris-ing more than 93 million hosts in over 200countries as of July 2000. E-business appli-cations are both leveraging and further

driving the growth of this infrastructure.As these applications mature, people arepaying increasing attention to quality-of-service issues. According to a Forrester Re-search analyst, “Eight seconds isn’t fastenough any more when it comes to down-loading Web pages. . .expectations arearound four seconds, no matter what kindof connection you’re on” [Nelson 2000].Improving response time requires atten-tion to resource allocation and pricing,which OR technologies are well positionedto address.

From the consumer’s quality of serviceviewpoint, the network infrastructure fore-business has three major components:(1) Local consumer access to the Internet(the so-called last mile), which is migrat-ing from narrow-band, dial-up access tohigh-speed broadband access using eitherDSL or cable modem technology;(2) The Internet and networks that use In-ternet technology, especially intranets andextranets, which deliver the content from aserver to the user’s browser; and(3) The Web-server farms and transaction-processing systems that facilitate access tocontent.The delays users encounter may be causedby problems with any of thesecomponents.Local access networks. Even with thegrowing adoption of broadband technolo-gies, dial-up access to the Internet over thepublic-telephone network will remain thedominant mode until about 2005 [Meekerand Mahaney 2000]. Calls to ISPs differfrom voice calls in their statistical charac-teristics. Their greater length has resultedin an increase in blocked calls, both voiceand data. In redesigning their networks to



handle these fundamentally differenttypes of calls, local exchange carriers areusing queueing theory and new statisticalmodels of calling patterns that differ fromthe Poisson models traditionally used todesign voice networks [Willinger andPaxson 1998].

Researchers have also examined the de-sign and economics of local access to datanetworks, including broadband Internetaccess. Fryxell, Sirbu, and Wanichkorn[1999] discuss the economics of TCP/IP-based local-access networks, andKlincewicz [1998] reviews OR methods fordesigning tributary networks. The applica-tion of OR to these problems requires astrong background in communicationsnetworks and technology. Oppenheimer[1999] provides a good introduction tonetwork design, and Cahn and Chan[1998] cover OR approaches to wide-area-network design.Content delivery networks. Quality of ser-vice for content-delivery networks hasbeen studied from three perspectives.

The first is that of Web caching[Rodriguez, Biersack, and Ross 2000]. AWeb cache is a computer system in a net-work that monitors requests from clientsto servers and stores copies of requestedcontent. When it sees a request for a Webpage that has been stored, it serves its lo-cal copy. In this manner, caches reduce re-sponse time for accessing Web content.Serving from cache also conserves band-width that otherwise would be used to re-trieve content from originating servers, aplus because the cost of storage is lowerthan the cost of Internet bandwidth[Christy 2000]. Further, caching reducesthe workload of originating servers.

Content-delivery network companies,such as Akamai and Digital Island, owncaches, which are located at large ISP ac-cess centers. Content-delivery networksmust formulate caching policies. Thesepolicies are governed by contracts withcontent providers, who specify the num-ber or placement of their mirror sites, orby objectives based on response time orbandwidth usage [Chuang and Sirbu2000]. Typically, the policies seek to mini-mize response time and conserve Internetbandwidth. Given replicated content inthe caches, content-delivery networksneed to direct incoming requests for con-tent to the most appropriate caches. Theywant to do this distributed load balancingwell because they bill content providersfor bytes served. Akamai, in particular,makes extensive use of OR methods suchas min-cost multicommodity flow algo-rithms for this purpose (see Akamai pat-ent US06108703 for global system hosting).

The second perspective related to con-tent delivery is the use of pricing for allo-cating resources. Currently, Internet pric-ing is the same for delay-sensitive packets(such as telephony) and delay-insensitivepackets (such as e-mail), even though dif-ferent applications require different ser-vice qualities.

In general, analysts have focused oncongestion pricing versus pricing at mar-ginal cost of bandwidth. Congestion pric-ing reflects the delay the allocation ofbandwidth to one application causes toother applications. Some analysts use eco-nomic modeling [Gupta, Stahl, andWhinston 1999]. Others use statisticalmodeling to assess the demand for band-width and consumers’ willingness to pay

GEOFFRION, KRISHNAN

INTERFACES 31:2 22

for different levels of service [Varian2000]. This may prepare the way for ap-plying yield management to ISP services.Finally, Chuang and Sirbu [1998] discussthe pricing of multicast (many recipients)versus unicast Internet traffic. Pricing hasbecome even more important with the roll-out of Internet Protocol Version 6, whichpermits tagging packets with quality-of-service labels. While backbone ISPs do notcurrently reconcile accounts for packetsexchanged with other ISPs at network ac-cess points, pricing issues will likely be-come more important in the future.

The third perspective on content-delivery networks concerns their design,long a fertile application area for OR[Cahn and Chan 1998]. Jim Crowe, CEO ofLevel 3 Communications, which has builta coast-to-coast, fiber-based, Internet pro-tocol network, says: “Level 3, at its heart,is a technology-based company and,frankly, the central technology is opera-tions research and optimization, which Itend to think is going to revolutionize allof business over the next 20 years”[Horner 2000, p. 40]. Among other appli-cations, Level 3 uses OR to decide howmany empty fiber-optic conduits to put inthe ground to minimize expected long-runnetwork costs in light of cost and capabil-ity estimates for future generations of fi-ber. This is an important problem, sinceoptical technology is improving rapidlyand trenches are expensive.

The closely related topic of survivablenetwork design has become increasinglyimportant to ISPs. The aim is to maintainnetwork services and minimize congestionafter physical or software attacks destroylinks or switches [Medhi and Tipper 2000].

Optimizing quality of service on servers.The control and optimization of perfor-mance measures for high-volume Websites requires a fundamental understand-ing of user-request patterns and their im-pact. One successful methodological con-tribution toward this end used stochasticprocesses and statistics to characterize re-quest patterns for heavily accessed Webservers and tested the results on data fromthe 1998 Nagano Olympic Games server[Iyengar, Squillante, and Zhang 1999]. IBMnow includes these techniques in its lead-ing application-server product, Web-Sphere, to improve peak-period responsetimes.

To solve network-infrastructure prob-lems, OR practitioners need a good under-standing of network protocols (TCP/IP),particularly such application-layer proto-cols as HTTP, and such concepts as cach-ing. Good starting points are Kurose andRoss [2000] on TCP/IP and Zhang [2000]on the technology underlying packet-switching networks and Internet protocols.Software Tools for Decision Technology

We focus on packaged, rather thancustom-built, software tools for decisiontechnology that contribute toward improv-ing the infrastructure of e-business by fa-cilitating OR work in that context (Figure1, third division). Whether or not suchsoftware is sold over the Web is not ourconcern here (such sales would registerunder the second division of Figure 1), norare the very considerable employment op-portunities for OR expertise associatedwith its design and implementation.

Such software most enhances OR pro-fessionals’ productivity and the range ofproducts and services they can provide



when it has the convenience of Web-accessibility and when it can operate inconjunction with other software in anInternet environment [Bhargava andKrishnan 1998] and hence be incorporatedinto e-business software. This is perhapsespecially true for the “lone” practitioner.

Two types of software are particularlyuseful to OR practitioners: (1) free soft-ware suitable for internal purposes shortof incorporation into production-qualityapplication software and (2) commercialsoftware suitable for building Web-enabled, production-quality applicationsoftware or for doing OR analysis—espe-cially when it complies with e-businessapplication integration standards. Betweenthese two is open-source software, whichhas considerable potential. It is free, can beincluded in applications subject to licens-ing restrictions, and can be of very highquality. Yahoo! reportedly runs largely onsuch software. Open-source decision-technology software is just starting to ap-pear; for example, IBM’s Common Optimi-zation INterface.

Much useful software is now freelyavailable on the Web (sometimes with re-strictions against commercial use). For ex-ample, with just a Web browser and anInternet connection, anyone can under-take to—solve many varieties of optimizationmodels using optimization servers andother resources detailed by Fourer andGoux [2001];—do many kinds of statistical computa-tions using tools available through suchsources as StatLib, StatPages.net, StatPointInternet Statistical Computing Center, andSticiGui;

—produce US and multicountry macroeco-nometric forecasts using Fairmodel; and—use Web-enabled, education-grade soft-ware, such as WebGPSS (a Java-applet fordiscrete event simulation), and the Web-based software that is starting to accom-pany textbooks (for example, the Harvest-ing Customer Value site has sevenmarketing-oriented packages), and Java,server, and online offerings, such as thoselisted in Michael Trick’s softwarecompilation.Over time, application service providerswill be making more and more OR toolsavailable via the Web, usually for a fee.

For projects that require building Web-enabled, production-quality applicationsoftware or that could profit from makingmodels or results accessible via the Web orthat need industrial-strength tools, ORpractitioners have an expanding array ofoptions:—DRA Systems offers OR-Objects, a free-ware collection of 500 Java classes in-tended for OR-application development.—ILOG software components are avail-able for constraint programming, mathe-matical programming, business rules, andvisualization and can be embedded inclient-side and server-side Webapplications.—InfoHarvest, Inc. offers a Web hostingservice that puts models built with theirflagship MCDM product online for securedata gathering, preference surveying, andresults/analysis sharing.—InterNetivity Inc. offers the easily incor-porated analysis and reporting enginedatabeacon, which works in conjunctionwith HTML data tables to enable interac-tive graphing, drill-down, sorting, pivot-

GEOFFRION, KRISHNAN

INTERFACES 31:2 24

ing, and more.—John Galt Solutions, Inc. offers theForecastX Engine as an ActiveX control toadd forecasting and statistical analysis toWeb-enabled applications.—Lumina Decision Systems, Inc. offers theAnalytica Decision Engine (decision analy-sis, influence diagrams, Monte Carlo, sta-tistics) for incorporation into the back endof custom Web-based applications.—Meta Software Corporation offersMetaSoft Web for publishing WorkFlowModeler business process models on theWeb or an intranet.—SAS Institute offers Web-enabled prod-ucts for data analysis and mining, econo-metrics, forecasting, OR (decision analysis,discrete-event simulation, optimization,and project management), and statistics.SAS’s multitier application-server archi-tecture enables applications to be Web-enabled easily, including dynamicallygenerated HTML pages to communi-cate results [Cohen, Kelly, and Medaglia2001].—SIMUL8 Corporation offers a free playerfor viewing animated results produced byits simulation package and stored on aWeb server.—StatPoint, LLC offers StatBeans, statisti-cal components in Java for adding statisti-cal content to applications or HTMLdocuments.—ThreadTec, Inc. offers the Java-basedsimulation language Silk for developingWeb-accessible, discrete-event simulationmodels.—TreeAge Software, Inc. offers DATA In-teractive, an ActiveX component, for dis-tributing interactive models built withDATA 3.5 (decision trees, Markov analy-

sis, Monte Carlo) over an intranet or theInternet.—Vanguard Software Corporation offersDecisionPro 3.0, a Web-enabled and e-mail-enabled OR-modeling package sup-porting decision trees, forecasting,influence-diagram calculations, linear andnonlinear programming, and Monte Carlosimulation. It also offers DecisionScript,built around DecisionPro 3.0 using server-side JavaScript, for building analytic appli-cations with expert-system features.—WhiteLight Systems, Inc. offers theWhiteLight Analytic Application Serverand its component-oriented, drag-and-drop development environment for build-ing Web-based, spreadsheet-like analyticapplications.Other vendors choose to specialize inturnkey installations, such as ALT-C Sys-tems Inc. (which offers TimeTrends e-Forecasting). Still others function as part-ners, such as,—AlphaBlox Corporation Inc., which of-fers a Web-based platform and many com-ponents for the rapid assembly and cen-tralized administration of analyticapplications [Morris and Garone 1999];—ExpertCommerce, Inc., which offers theExpert Choice analytic hierarchy processimplemented as a Web-enabled business-to-business purchase-evaluation engine;and—Manugistics, Inc., which offers a suite ofoptimization engines targeted at supply-chain management and at pricing and rev-enue management, all of which can be em-bedded into its ExchangeWORKS platformfor building custom public or private elec-tronic trading environments.

Most analytical software companies will



likely have to Web-enable their products[Bhargava and Krishnan 1998] to remaincompetitive. The bar will have been raisedwithin a few years, and thereafter Web-enablement will be viewed as a competi-tive necessity rather than competitiveadvantage.

With the software categories just re-viewed changing so rapidly, OR profes-sionals will need to invest extra time stay-ing abreast of such software andcompatible implementation architectures.Why the Digital Economy Needs OR

Some of the kinds of value that OR canadd are particularly well matched to thedigital economy’s needs. Here are six no-table strengths of OR whose importancesprings from attributes of the digital econ-omy. They are not the ones most com-monly mentioned in the context of the off-line economy; OR is far more than merelya way to squeeze costs by another fivepercent.(1) OR excels at squeezing value out ofdata, which the digital economy is creat-ing in unprecedented abundance: auctionlogs, customer profiles, mobile scannerdata, point-of-sale data, process data gush-ing into data warehouses, transactionalERP data, Web logs, and more. By cou-pling statistics, data analysis, and datamining [Goebel and Gruenwald 1999; TwoCrows Corporation 1999] with decisiontechnology, organizations can turn thedata deluge into actionable knowledge.(2) OR can cope with complexity in de-sign, planning, and operations because itis inherently analytical. Analysis meansbreaking up a complex whole into its com-ponent parts; just what is needed to dealwith the increasing complexity brought

about by e-business.(3) OR can manage risk and deal with un-certainty using statistics, decision analysis,and probabilistic modeling. In an agewhen business and technological change isso rapid that the past is a poor guide tothe present, let alone the future, it is pru-dent to invoke model-based approaches todecision making that quantify uncertaintyand risk explicitly.(4) OR’s model-building approach is oneof the most reliable ways to achieve adeep understanding of business processesand issues. It can be argued that no phe-nomenon, even man-made ones like busi-ness processes, is truly understood until itcan be measured and modeled. Hence ORaccords well with the premium that man-agement increasingly places on intellectualcapital and methods for knowledgediscovery.(5) OR can perform virtual experiments(analyses, simulations) without riskingdamage to a company’s assets or financialperformance. With intuition-busting nov-elty the order of the day, businesses needto avoid failures whose consequences maybe exhibited and multiplied by the Web’sglobal reach.(6) OR can provide decision technologyfor operational software to handle entireclasses of decisions quickly (perhaps inreal time, as in many Web applications),repeatedly, and automatically. Even deci-sions that have long defied automation,such as approving home-equity loans, arenow made automatically using decisiontechnology. Information technology is sel-dom sufficient for automating business de-cisions, for rules of thumb that work welleven 99 percent of the time can generate

GEOFFRION, KRISHNAN

INTERFACES 31:2 26

huge economic penalties and dissatisfac-tion when used on a high-traffic Web site.The solution is to wed decision and infor-mation technology.

Other conspicuous strengths of OR,such as powerful ways to analyze deci-sions of great consequence, to achieve truecost minimization in many cases (not justcost reduction), and to make trade-offs,seem not to be especially advantaged bythe advent of the digital economy but re-main as useful as ever.

Every strength has already been illus-trated by cited applications (Table 1). Weelaborate further on the first and last.Exploiting the digital data bonanza.Customer-relationship management isone of the greatest beneficiaries of the IT-induced bonanza of data, and OR is takingadvantage of it. Software Magazine foundCRM to be the fastest-growing sector ofthe software industry during 1999, and In-ternational Data Corporation (IDC) fore-casts CRM to be the fastest-growing sectorof the packaged analytic-applications-software market through 2004, jumpingfrom worldwide revenues of $486 millionin 2000 to $2,374 million in 2004 [Morris2000]. The aim of CRM is to identify, ac-quire, serve, extract value from, and retainprofitable customers by interacting withthem effectively in an integrated wayacross the full range of customer-contactpoints for marketing, sales, and service—e-mail, in person, mail, phone, Web, andso on.

Doing this requires making sense ofmountains of pertinent data from opera-tional systems (order entry, finance, andshipping), customer-contact records (in-cluding call-center servers, reply cards,

and Web servers), and third-party sources(for example, credit histories and demo-graphics). This is best done through datamining, statistical studies, and models thatideally maximize lifetime customer value[Chatham 2000]. Some examples of OR insupport of CRM follow:(1) HNC Software Inc. uses neural-network and other predictive technologiesto recompute the predicted profitability ofindividual customers as transactions oc-cur, to better inform efforts to maximizecustomer lifetime value by balancingcredit risk, revenue potential, and cus-tomer attrition in financial services, tele-communications services, and otherapplications.(2) Market Switch Workstation is anenterprise-level decision-support applica-tion that uses constrained optimization forreal-time, multioffer optimization for in-bound marketing (call center, Web) andcustomer care. From a portfolio of options,the software determines what marketingpromotion, advertisement, or customer-service response will maximize the finan-cial return of that customer. It also doesmultioffer, enterprise-wide optimizationfor outbound marketing (direct mail, e-mail, telemarketing) to find the most prof-itable offers to release. This product is in-corporated into the products ofmarketing-automation-service providers,such as Prime@Vantage Optimizer andXchange Dialogue.(3) Trajecta, Inc.’s Optimization Simula-tion Environment uses neural networks,regression, and single-stage stochastic lin-ear programming to predictively modelkey customer behaviors and maximize(with robust optimization methods) the



These strengthsof OR. . .

. . .Help meet these challengesof the digital economy

Illustrative applications (discussedin the text)

Exploiting the databonanza

Digitally formatted data nowdominate nondigital data bymore than an order ofmagnitude and are increasingfar faster [Lyman and Varian2000].

Customer-relationship management(see the subsection “Exploiting thedigital data bonanza”); Electronicmarket services (DigitalTransportation Marketplace byLogistics.com, FreightWise byManugistics, Inc.); Web data mining[Cohen, Kelly, and Medaglia 2001];Web-site personalization (StrategicData Corp.).

Coping withcomplexity

The digital economy multipliesthe decision complexity of thetraditional economy with newtechnological options.

Combinatorial auction bid selection(Automated Credit Exchange [Ishikidaet al. 2000], Bond Connect [Bossaerts,Fine, and Ledyard 2000]); Content-delivery network operations (Akamaipatent US06108703); Logisticsoperations (OptiFlow).

Coping withuncertainty

Accelerating technology-drivenbusiness innovation and changegreatly amplify traditionalsources of uncertainty.

Scenario-based forecasting (FinancialEngines Investment Advisor); Mark-to-Future [Dembo et al. 2000]); Webserver quality of service [Iyengar,Squillante, and Zhang 1999]; Yieldmanagement [Smith et al. 2001].

Achievingunderstanding

The digital economy’s demandsfor action prompt decisionmaking without benefit of thecold light of reason.

Most applications based on validated,symbolic models in the OR tradition.

Experimentingwithout risk

Accelerating technology-drivenbusiness innovation and changeincrease exposure to risk.

Fiber network planning [Horner 2000];Investment portfolio analysis(Financial Engines InvestmentAdvisor); Most applications based onpredictive models in the OR tradition.

Automatingrecurrent decisions

Scalability and Web-userexpectations drive operationalautomation in e-business.

Home Depot combinatorial auctions[Keskinocak and Tayur 2001]; Onlinelending (First Union Loan Arranger);Real-time marketing optimization(MarketSwitch Workstation); Real-time yield management (OptiYield-RT).

Table 1: The digital economy needs these particular strengths of OR. Nearly every applicationmentioned in this article illustrates one or more of them.

GEOFFRION, KRISHNAN

INTERFACES 31:2 28

lifetime value of customers in specific pop-ulations, such as credit-card holders, sub-ject to constraints on such considerationsas near-term revenue and attrition. Busi-ness problems addressed include onlinecustomer acquisition, cross-selling, dy-namic repricing, and retention.

In addition, important CRM firms whichdo not yet apply OR, such as Siebel Sys-tems, have application-programming in-terfaces permitting the attachment of ana-lytic applications to their platforms.

A closely related application of wide in-terest is Web-site personalization, tailoringWeb-site content to individuals based onwhatever the firm may know or be ableto infer about them [Riecken 2000]. Someapproaches, especially the offlinesegmentation-oriented methods used todrive online rules-based personalization,are highly data-intensive and often in-volve applying data mining to clickstreamdata [Mulvenna, Anand, and Buchner2000]. (Mining clickstream data—recordsof individuals’ navigational paths on theWeb—is also popular for improving Web-site design and evaluating advertisingcampaigns [Cohen, Kelly, and Medaglia2001].) AI-oriented computer-science pro-fessionals, rather than OR professionals,have been the main architects of personali-zation and recommendation systems[Schafer, Konstan, and Riedl forthcoming].As with data mining, OR people can easilyunderstand the technologies involved andmay be able to add value by integratingthem into decision-support systems.Real-time OR. Decision technology em-bedded in software for e-business pro-cesses must often work in real time. Thisis true, for instance, for real-time CRM ap-

plications of the sort just discussed, sincethe very limited attention span of a livecustomer at a Web site or on a cell phonedemands immediate response [Nelson2000]. The same motive holds for someconsumer financial applications, for inter-active dynamic-pricing applications wherebidding, pricing, or purchase negotiationsoccur in one or two user sessions, and forrecommendation and preference-basedWeb-site personalization applications thattrack user activity and match it with otherusers’ activity for predictive purposes. Anexample is the custom recommendationsystem used by Amazon’s site. When aregistered consumer visits, the site offersproduct recommendations based on per-sonalization technology [Schafer, Konstan,and Riedl forthcoming] fed by multipledata sources such as the purchase patternsof “similar” consumers and explicit feed-back from the buyer.

A different motive underlies the needfor real-time operational decision supportin the transportation industry and in mo-bile commerce: equipment and people inmotion are changing state so fast that onlya speedy response can be useful.

A supplier’s online order-taking processillustrates a third motive for real-time per-formance. Customers usually ask when anorder can be delivered. The supplier basesthe available-to-promise answer on inven-tory, orders on hand, and possibly the cur-rent production schedule. If it cannot de-liver the order soon enough, it may avert alost sale by giving the capable-to-promiseanswer after rerunning the production-scheduling algorithm with the potentialorder included. For a large order, the sup-plier could extend the analysis to take ac-



count of spot-market or strategic-partneractions that might speed up delivery. Ide-ally, it should perform all calculations inreal time to close the deal before the cus-tomer has a chance to check with acompetitor.

In general, operational and even tacticalbatch-style OR applications will increas-ingly require conversion to real-timeoperation as business processes are auto-mated to meet e-business requirements.

Achieving real-time capability requiresmodels and solvers that reliably returngood solutions in compute windows thatmay be very much shorter than desired.Heuristics or approximation algorithmsare often necessary.

Some applications pose challenges be-yond short compute windows. One is thepossible need for stability in solutionspace across optimal (or near-optimal) so-lutions. For instance, it could be objection-able in a capable-to-promise applicationfor revised production schedules to differwildly for reasons that are essentially acci-dental to the reoptimization algorithm.There are several known approaches topromote solution-space stability. Anotherchallenge concerns how to think about re-petitive decision making with incompleteinformation. The most appropriate solverparadigm may invoke the notion of an on-line algorithm, more familiar in theoreticalcomputer science than in OR. Such an al-gorithm must execute each time new datacomes in, taking an action that might notbe optimal in light of subsequent data. Inthe popular competitive-analysis approach[Borodin and El-Yaniv 1998], an algo-rithm’s goodness is measured relative tothe best possible performance of an algo-

rithm that has complete knowledge of thefuture. There are other approaches, andopen research challenges in real-time ORabound.Implications for Practitioners

How can OR professionals best flourishin the new era? First, as always, by deeplyunderstanding the organization. Second,by playing to the six strengths listed previ-ously. Third, by improving their ability toexecute e-business projects:Stay current. Discovering opportunitiesfor OR applications and discussing themconvincingly with managers require abroad understanding of e-business pro-cesses and issues that is also deep in se-lected areas. This will not come easily.Stay current by reading thoughtful main-stream coverage of the digital economy,including The Economist, Harvard BusinessReview, and The Wall Street Journal Interac-tive Edition. Study reports from researchservices, such as Forrester Research, Inc.and Gartner Group. Follow suchtechnology-oriented magazines as Business2.0, Computer, The Industry Standard, andInformation Week. Periodically scan Yahoo!Inc.’s Tech Full Coverage, a good generalportal to other sources.Study the new decision technologies. Re-sponding effectively to new challengesand opportunities calls for an expandedtechnical repertoire. This means learningabout such topics as auction design, con-straint programming, data mining, dy-namic pricing and revenue management,personalization and recommendation tech-nologies, and real-time OR. Few of theseare covered in standard OR curricula, afact that warrants attention in academia.We have given sources for each of these.

GEOFFRION, KRISHNAN

INTERFACES 31:2 30

Monitor such OR sites as INFORMSOnline and e-Optimization.com for newdevelopments and professional-educationopportunities.Study economics. Brush up on economics,especially microeconomics, so that you canfollow pertinent work in this tradition andperhaps initiate the kind of hybrid studieswe called for earlier in which methodsfrom economics and OR are combined toproduce e-business analyses that go be-yond qualitative insights to solve realmodel instances. Unfortunately, micro-economics has not always been includedas one of the foundations of OR education.For a refresher, read Luenberger’s [1995]elegant Microeconomic Theory, whose view-point and development are closer to ORthan similar books written by economists.Study the new information technologies.Many OR practitioners program or workclosely with programmers. Since online e-business applications often require fairlynew computer languages and implementa-tion architectures, some investment in thisdirection is appropriate. Study Bhargavaand Krishnan’s [1998] paper to learn howto use Web technologies and distributedcomputing to deploy OR applications,study Fraternali’s [1999] paper and thetools he mentions to prepare for Web im-plementations that involve databases, readMaurer’s [2000] and Sprott’s [2000] articlesto find out why component-level program-ming is so important, and read Ousterhout[1998] to see why scripting languages areso important. To upgrade your program-ming skills, you could learn a scriptinglanguage like Per1 or JavaScript for gluingcomponents together. Those needing asystem-programming language should

consider Java, a very popular object-oriented language with excellent cross-platform mobility and good facilities forlinking to networks and packaging appli-cations as servers.Learn to work on Internet time. Internettime is not a fiction. Universities do notusually operate at this pace, nor do mostOR practitioners. Yet anyone who hopesto work in the e-business world needs tolearn to do professional projects faster.Preferably much faster. That means tailor-ing analyses to the time available and set-ting realistic targets for solution quality.Recent graduates will either make theirpeace with these heresies or soon enoughfind themselves relieved of their torments.Producing to Internet-driven schedulesalso means aggressively exploiting Inter-net communications options beyond e-mail, and for implementations, duly con-sidering using Internet standards tofacilitate integration, buying rather thanbuilding software or software components,using high-level and rapid developmenttools [Bhargava, Sridhar, and Herrick1999], and implementing solutions in ASPmode.

Following these suggestions should po-sition OR professionals for deeper involve-ment with e-business projects—perhapseven to strike out as Internet-based consul-tants or entrepreneurs. The head of onesmall (three-person) optimization-basedsoftware vendor offered these candidcomments:

“. . .When customers ask how many people wehave, I usually say ‘six’, quickly adding someof our agents. The company is totally virtual, inthat we all work from home. Communication isvia e-mail. No commuting, flexible work hours.We do need that, by the way, as the globality



of the Internet means that we have customersin all time zones. So, on Sunday afternoon forus, the Australians get back to work. . . Wehave customers in more than 20 countries. Oneof the key points of working via the Internet isthat you can look a lot bigger than you reallyare. . . sales over the Internet are totally not-regulated. That is, as long as our customersdownload things (no hard CDs, user guides,etc.), officially there are no import/export re-quirements, forms to fill out, duties to be paid,etc. So, this saves a lot of work.”

At the outset, we likened the advent ofe-business to the Cambrian explosionabout 540 million years ago, when but aninstant of evolutionary time produced allthe phyla from which today’s animals de-scend. The explosively evolving e-businessecosystem creates numerous opportunitiesfor many professions, most certainly in-cluding OR. But keep in mind that catas-trophes can also occur, such as theCretaceous-Tertiary boundary event 66million years ago when about half of thespecies then known disappeared. Somepundits predict this sort of future for thee-business ecosystem after the financialmarkets finish taking their toll, and net-work effects destroy all but a few domi-nant players in many e-business marketsor worsening security breaches destroypublic and managerial confidence. Sober-headed OR may be able to help companiesmanage such catastrophic risks.

Such specters may come to pass in somedegree, but the e-business genie can neverbe put back in the bottle. Surely the e-business mania will fade and e- prefixeswill become passe as the online and offlineeconomies fuse [Fry 2000]. Beyond that,we can be sure only that breathtakingchange will continue. For OR to thrive indifficult times as well as good, it mustadapt and evolve about as fast as the busi-

ness ecosystem is changing. The key tothat is a strong curiosity about e-businessdevelopments and a willingness to experi-ment with new hybrids of information anddecision technology.Online References

Underlined terms in the paper indicate on-line references.Akamai (www.akamai.com)Akamai patent US06108703 (www.delphion.

com/details?&pn�US06108703__)Algorithmics (www.algorithmics.com)ALT-C Systems Inc. (www.alt-c.com)application service provider (www.

aspindustry.org)BARRA, Inc. (www.barra.com)Business 2.0 (www.business2.com)Common Optimization INterface (oss.

software.ibm.com/developerworks/opensource/coin/)

Computer (computer.org/computer/)data mining (www.kdnuggets.com)digital certificates (www.rsasecurity.com/

rsalabs/faq/4-1-3.html)Digital Island (www.digisle.net)eBay (www.ebay.com)e-business application integration standards

(www.openapplications.org)e-Optimization.com (www.e-optimization.

com)ESRI (www.esri.com)ExpertCommerce, Inc. (www.

expertcommerce.com)Fairmodel (fairmodel.econ.yale.edu)Financial Engines Investment Advisor

(www.financialengines.com)Forrester Research, Inc. (www.forrester.com)Frictionless Commerce (www.frictionless.

com)Gartner Group (www.gartnerweb.com)Harvesting Customer Value (www.

valueharvest.com)HNC Software Inc. (www.hnc.com/pdf/

HNC.pdf)i2 TradeMatrix (www.tradematrix.com)The Industry Standard (www.thestandard.

com)ILOG software components (www.ilog.

com/industries/ebusiness/)InfoHarvest (www.infoharvest.com)

GEOFFRION, KRISHNAN

INTERFACES 31:2 32

Information Rules Web site (www.inforules.com/teaching.htm)

Information Week (www.informationweek.com)

INFORMS Online (www.informs.org)InterNetivity Inc. (www.internetivity.com)Java (developer.java.sun.com/developer/

onlineTraining/)JavaScript (webopedia.internet.com/

TERM/J/JavaScript.html)John Galt Solutions, Inc. (www.forecastx.

com)Lumina Decision Systems, Inc. (www.

lumina.com)Manugistics, Inc. (www.manugistics.com)Market Design Inc. (www.

market-design.com)MarketSwitch (www.marketswitch.com)MathML (www.w3.org/Math/)Meta Software Corporation (www.

metasoftware.com)Michael Trick’s software compilation

(mat.gsia.cmu.edu/Computer_Programs/)Net Exchange (www.nex.com)online lending (www.ilog.com/industries/

finance/onlinelend.cfm)open-source software (www.opensource.org)OptiBid (www.logistics.com/static/

static1-1.asp)OptiFlow (www.logistics.com/static/

static1-1.asp)Optimal Retirement Planner (www.i-orp.

com)OptiStop (www.logistics.com/static/

static2-1.asp)OptiYield-RT (www.logistics.com/static/

static2-1.asp)OR-Objects (www.opsresearch.com/

OR-Objects/index.html)Perfect.com (www.perfect.com)Perl (www.perl.com/pub)personalization (www.personalization.com)Prime@Vantage Optimizer (www.primeresponse.

com/products/optimizer.html)ProMiles Online (www.promilesonline.com)PROS Revenue Management, Inc. (www.

prosrm.com)Sabre Inc. (www.sabre.com)SAS Institute (www.sas.com)Siebel Systems (www.siebel.com)SIMUL8 Corporation (www.SIMUL8.com)SmartSettle (www.smartsettle.com)

StatBeans (www.sgcorp.com/statbeans.htm)StatLib (lib.stat.cmu.edu)StatPages.net (www.statpages.net)StatPoint Internet Statistical Computing Center

(www.sgcorp.com/on-line_computing.htm)SticiGui (www.stat.Berkeley.EDU/�stark/

SticiGui/index.htm)Strategic Data Corp. (www.

strategicdatacorp.com)ThreadTec, Inc. (www.threadtec.com)Trajecta, Inc. (www.trajecta.com)TreeAge Software, Inc. (www.treeage.com)Vanguard Software Corporation (www.

vanguardsw.com)vendor-managed inventory (www.cpfr.

org/Vendor-Managed.html)virtual private networks (www.vpnc.

org/white-papers.html)The Wall Street Journal Interactive Edition

(interactive.wsj.com/home.html)Web caching (www.caching.com)WebGPSS (webgpss.hk-r.se)WhiteLight Systems, Inc. (www.

whitelight.com)Xchange Dialogue (www.xchange.com/

solutions/dialogue/)XML (www.w3.org/XML/)Yahoo! Inc.’s Tech Full Coverage

(fullcoverage.yahoo.com/fc/Tech/)ReferencesAndrews, Whit 1999, “The new laws of dy-

namic pricing,” Internet World, December 15,pp. 26–34.

Apfel, Audrey 2000, “A new lease on soft-ware,” Executive Edge, Gartner Group, June–July. Retrieved December 9, 2000 from www.ee-online.com/jun2000/jun2000_feat_04.htm

Bakos, Yannis 1998, “The emerging role of elec-tronic marketplaces on the Internet,” Commu-nications of the ACM, Vol. 41, No. 8 (August),pp. 35–42.

Barua, Anitesh and Whinston, Andrew B. 2000,“Measuring the Internet economy,” CiscoSystems and University of Texas, June 6.Retrieved December 9, 2000 from www.internetindicators.com/june_full_report.PDF

Beam, Carrie; Segev, Arie; and Shanthikumar,George 1999, “Optimal design of Internet-based auctions,” Working paper 98-WP-1034,Haas School of Business, University of Cali-fornia, Berkeley, January.



Bertsimas, D.; Darnell, C.; and Soucy, R. 1999,“Portfolio construction through mixed-integer programming at Grantham, Mayo,Van Otterloo and Company,” Interfaces, Vol.29, No. 1 (January–February), pp. 49–66.

Bhargava, H. and Krishnan, R. 1998, “TheWorld Wide Web and its implications forOR/MS,” INFORMS Journal on Computing,Vol. 10, No. 4 (Fall), pp. 359–383.

Bhargava, H.; Sridhar, S.; and Herrick, C. 1999,“Beyond spreadsheets: Tools for building de-cision support systems,” Computer, Vol. 32,No. 3 (March), pp. 31–39.

Billington, Corey 2000, Personal communica-tion, Hewlett-Packard Company, Palo Alto,California, September 18.

Bossaerts, P.; Fine, L.; and Ledyard, J. 2000,“Inducing liquidity in thin financial marketsthrough combined-value trading mecha-nisms,” Working paper, California Instituteof Technology, March 27. Retrieved Decem-ber 9, 2000 from www.hss.caltech.edu/�jledyard/bfl_4�00.pdf

Borodin, Allan and El-Yaniv, Ran 1998, OnlineComputation and Competitive Analysis, Cam-bridge University Press, Cambridge, UnitedKingdom.

Brown, S. 2000, “How e-tailers deliver withinhours,” Fortune, Vol. 141, No. 11, May 29, pp.T210B–T210N.

Cahn, R. and Chan, R. 1998, Wide-Area NetworkDesign—Concepts and Tools for Optimization,Morgan Kaufmann, San Francisco, California.

Chan, Y. M.; Womer, J.; MacKie-Mason, J. K.;and Jamin, S. 1999, “One size doesn’t fit all:Improving network QoS through preference-driven Web caching,” Working paper,University of Michigan. Retrieved December9, 2000 from www-personal.umich.edu/�jmm/papers.html#osdfa

Chatham, R. 2000, “Managing customer profit-ability,” Report, Forrester Research, Inc.,Cambridge, Massachusetts, February.

Christy, P. 2000, “Content delivery architec-tures: Why doesn’t one size fit all?” Whitepaper, Internet Research Group, Los Altos,California. Retrieved December 9, 2000from www.irgintl.com/publications/whitepapers/iBeamCDDWP1-21.pdf

Chuang, John and Sirbu, Marvin 1998, “Pricingmulticast communications: A cost-based ap-proach,” Proceedings of the Internet Society

INET’98 Conference, Geneva, Switzerland,July 21–24. Retrieved December 9, 2000 fromwww.isoc.org/inet98/proceedings/6d/6d_2.htm

Chuang, John and Sirbu, Marvin 2000, “Distrib-uted network storage with quality-of-serviceguarantees,” Journal of Network and ComputerApplications, Vol. 23, No. 3, pp. 163–185. Re-trieved December 9, 2000 fromwww.isoc.org/inet99/proceedings/4q/4q_3.htm

Clemons, Eric K. and Weber, Bruce W. 1997,“Information technology and screen-basedsecurities trading: Pricing the stock and pric-ing the trade,” Management Science, Vol. 43,No. 12 (December), pp. 1693–1708.

Cohen, Marc-David; Kelly, Charles B.; andMedaglia, Andres L. 2001, “Decision supportwith Web-enabled software,” Interfaces, Vol.31, No. 2 (March–April), pp. 109–129.

Cooper, Lee and Giuffrida, Giovanni 2000,“Turning datamining into a management sci-ence tool: New algorithms and empirical re-sults,” Management Science, Vol. 46, No. 2(February), pp. 249–264.

Delaney, Robert and Wilson, Rosalyn 2000,“Logistics and the Internet: In the franticsearch for space, it is still about relation-ships,” Paper presented at the National PressClub, Washington, DC, June 5. Retrieved De-cember 9, 2000 from www.cassinfo.com/Press_Conference_2000.zip

Dembo, R.; Aziz, A.; Rosen, D.; and Zerbs, M.2000, “Mark-to-Future: A framework formeasuring risk and reward,” Report, Algo-rithmics Publications, Toronto, Canada.

The Economist 2000a, “Survey: Government andthe Internet,” special section, 34 pages, June24.

The Economist 2000b, “Survey: Online finance,”special section, 52 pages, May 20.

Fourer, Robert and Goux, Jean-Pierre 2001,“Optimization as an Internet resource,” Inter-faces, Vol. 31, No. 2 (March–April), pp. 130–150.

Fraternali, P. 1999, “Tools and approaches fordeveloping data-intensive Web applications,”Computing Surveys, Vol. 31, No. 3 (Septem-ber), pp. 227–263.

Fry, J. 2000, “What’s in store,” Wall Street Jour-nal, Industry and Economics, January 1, pp.R38–41.

GEOFFRION, KRISHNAN

INTERFACES 31:2 34

Fryxell, D.; Sirbu, M.; and Wanichkorn, K.1999, “An IP-based local access network:Economic and public policy analysis,” inCompetition, Regulation and Convergence: Cur-rent Trends in Telecommunications Policy Re-search, eds. S. E. Gillette and I. Vogelsang,Lawrence Erlbaum Associates, Inc., Mahwah,New Jersey, pp. 255–282.

Geoffrion, A. 1998, “A new horizon forOR/MS,” INFORMS Journal on Computing,Vol. 10, No. 4 (Fall), pp. 390–392.

Goebel, M. and Gruenwald, Le 1999, “A surveyof data mining and knowledge discoverysoftware tools,” SIGKDD Explorations, Vol. 1,No. 1 (June), pp. 20–33. Retrieved December9, 2000 from www.acm.org/sigs/sigkdd/explorations/issue1-1/survey.pdf

Gupta, A.; Stahl, D. O.; and Whinston, A. B.1999, “The economics of network manage-ment,” Communications of the ACM, Vol. 42,No. 9 (September), pp. 57–63.

Hingorani, N. 2000, Personal communication,NeoModal.com, Charlotte, North Carolina,October 19.

Horner, Peter 2000, “Something to Croweabout,” OR/MS Today, Vol. 27, No. 3 (June),pp. 38–44.

Ishikida, T.; Ledyard, J.; Olson, M.; and Porter,D. 2000, “Experimental testbedding of a pol-lution trading system: Southern California’sRECLAIM emissions market,” Social ScienceWorking Paper 1094, California Institute ofTechnology, March. Retrieved December 9,2000 from www.hss.caltech.edu/�jledyard/wp1094.pdf

Iyengar, Arun; Squillante, Mark; and Zhang, Li1999, “Analysis and characterization of large-scale Web server access patterns and perfor-mance,” World Wide Web, Vol. 2, No. 1–2, pp.85–100.

Jarrow, R. A.; Maksimovic, V.; and Ziemba,W. T., eds. 1995, Handbooks in Operations Re-search and Management Science, 9: Finance,Elsevier-Science, Amsterdam, TheNetherlands.

Johnson, C. 2000, “Digital downloads acceler-ate,” Report, Forrester Research, Inc., Cam-bridge, Massachusetts, January.

Kafka, S. J. 2000, “eMarketplaces boost B2Btrade,” Report, Forrester Research, Inc., Cam-bridge, Massachusetts, February.

Kaplan, S. and Sawhney, M. 2000, “E-Hubs:The new B2B marketplaces,” Harvard Busi-

ness Review, Vol. 78, No. 3 (May–June), pp.97–103.

Kephart, Jeffrey; Hanson, James; andGreenwald, Amy 2000, “Dynamic pricing bysoftware agents,” Computer Networks, Vol. 32,No. 6 (May 30), pp. 731–752.

Keskinocak, Pinar and Tayur, Sridhar 2001,“Quantitative analysis for Internet-enabledsupply chains,” Interfaces, Vol. 31, No. 2(March–April), pp. 70–89.

King, D. 2000, “Intranets: An Internet inside theorganization,” in Handbook of Electronic Com-merce, eds. M. Shaw, R. Blanning, T. Strader,and A. Whinston, Springer-Verlag, Berlin,Germany, pp. 521–541.

Klincewicz, J. 1998, “Hub location in back-bone/tributary network design: A review,”Location Science, Vol. 6, No. 1/4, pp. 307–335.

Kurose, James F. and Ross, Keith W. 2000, Com-puter Networking: A Top-Down Approach Fea-turing the Internet, Addison-Wesley Publish-ing Company, Reading, Massachusetts.

Lawrence, Stacy 2000, “Behind the numbers:The mystery of B-to-B forecasts revealed,”The Standard, February 21. Retrieved Decem-ber 9, 2000 from www.thestandard.com/research/metrics/display/0,2799,11300,00.html

Lucking-Riley, David 2000, “Auctions on theInternet: What’s being auctioned, and how?”Journal of Industrial Economics, Vol. 48, No. 3(September), pp. 227–252.

Luenberger, D. G. 1995, Microeconomic Theory,McGraw-Hill, New York.

Luenberger, D. G. 1998, Investment Science, Ox-ford University Press, New York.

Lustig, Irv and Puget, Jean-Francois 1999, “Pro-gram !� Program: Constraint programmingand its relationship to mathematical pro-gramming,” Technical paper, ILOG, Inc.Retrieved December 9, 2000 from www.ilog.com/products/optimization/tech/interfaces.pdf

Lyman, P. and Varian, H. R. 2000, “How muchinformation?” School of Information Man-agement and Systems, University of Califor-nia, Berkeley. Retrieved December 8, 2000from www.sims.berkeley.edu/how-much-info/

Maurer, Peter 2000, “Components: What if theygave a revolution and nobody came?” Com-puter, Vol. 33, No. 6 (June), pp. 28–34.

Medhi, D. and Tipper, D. 2000, “Multi-layered



network survivability—models, analysis, ar-chitecture, framework and implementation:An overview,”Proceedings of the DARPA Infor-mation Survivability Conference and Exposition(DISCEX’ 2000), Hilton Head Island, SouthCarolina, January 25–27, Volume 1, pp. 173–186.

Meeker, Mary and Mahaney, Mark 2000, “In-ternet: New media and eCommerce and PCsoftware,” Morgan Stanley Dean Witter,June. Retrieved December 9, 2000 fromwww.msdw.com/techresearch/globalprimer/Primer1.pdf

Morris, Henry 2000, “Analytic applicationsmarket forecast and analysis, 2000–2004,” Re-port #23498, International Data Corporation,Framingham, Massachusetts.

Morris, Henry and Garone, Steve 1999, “Com-ponent architecture for rapid delivery ofWeb-based analytic applications,” Report#W19016, International Data Corporation,Framingham, Massachusetts, May. RetrievedDecember 9, 2000 from www.alphablox.com/idcform.asp?source�www.alphablox.com (registration required).

Mulvenna, Maurice; Anand, Sarabjot; andBuchner, Alex 2000, “Introduction: Personali-zation on the Net using Web mining,” Com-munications of the ACM, Vol. 43, No. 8 (Au-gust), pp. 122–125.

Mulvey, J.; Gould, G.; and Morgan, C. 2000,“An asset and liability management systemfor Towers Perrin-Tillinghast,” Interfaces, Vol.30, No. 1 (January–February), pp. 96–114.

Nelson, Matthew 2000, “Fast is no longer fastenough,” Information Week Online, June 6. Re-trieved December 9, 2000 fromwww.informationweek.com/789/web.htm

Nigam, Raj; Labe, Russ; and Altschuler, Stuart2000, “Bullish on management science,”OR/MS Today, Vol. 27, No. 3 (June), pp. 48–51.

Oppenheimer, P. 1999, Top-Down Network De-sign, Cisco Press, San Jose, California.

Ousterhout, John 1998, “Scripting: Higher-levelprogramming for the 21st century,” Com-puter, Vol. 31, No. 3 (March), pp. 23–30.

Partyka, Janice G. and Hall, Randolph W. 2000,“On the road to service,” OR/MS Today, Vol.27, No. 4 (August), pp. 26–30.

Punishill, J. 2000, “Overhauling financial ad-vice,” Report, Forrester Research, Inc., Cam-bridge, Massachusetts, February.

Putnam, M. 1999, “Business services on thenet,” Report, Forrester Research, Inc., Cam-bridge, Massachusetts, January.

Riecken, Doug 2000, “Introduction: Personal-ized views of personalization,” Communica-tions of the ACM, Vol. 43, No. 8 (August), pp.26–28.

Rodriguez, P.; Biersack, E.; and Ross, K. 2000,“Automated delivery of Web documentsthrough a caching infrastructure,” Workingpaper, Institut EUROCOM, France. RetrievedDecember 9, 2000 from www.eurecom.fr/�ross/push.ps

Russell, Michael and Keith, Robert 2000, “Thee-Marketing Report,” Morgan Stanley DeanWitter, May 3. Retrieved December 9, 2000from www.msdw.com/techresearch/emarketing/emarketing.pdf

Sanders, M. 2000, “Global eCommerce ap-proaches hypergrowth,” Brief, Forrester Re-search, Inc., Cambridge, Massachusetts, April18.

Schafer, J. B.; Konstan, J.; and Riedl, J. forth-coming, “Electronic commerce recommenda-tion applications,” Journal of Data Mining andKnowledge Discovery. Retrieved December 9,2000 from www.cs.umn.edu/Research/GroupLens/ECRA.pdf

Shapiro, C. and Varian, H. 1998, InformationRules, Harvard Business School Press, Bos-ton, Massachusetts.

Smith, Barry C.; Gunther, Dirk P.; Rao, B.Venkateshwara; and Ratliff, Richard M. 2001,“E-commerce and operations research in air-line planning, marketing and distribution,”Interfaces, Vol. 31, No. 2 (March–April), pp.37–55.

Smith, Michael D.; Bailey, Joseph; andBrynjolfsson, Erik 2000, “Understanding digi-tal markets: Review and assessment,” in Un-derstanding the Digital Economy, eds. E.Brynjolfsson and B. Kahin, MIT Press, Cam-bridge, Massachusetts. Retrieved December9, 2000 from ecommerce.mit.edu/papers/ude/ude.pdf

Smith, Tim 2000, Personal communication,Trajecta.com, Austin, Texas, July 21.

Sodhi, ManMohan S. 2001, “Applications andopportunities for operations research inInternet-enabled supply chains and electronicmarketplaces,” Interfaces, Vol. 31, No. 2(March–April), pp. 56–69.

Sprott, David 2000, “Componentizing the enter-

GEOFFRION, KRISHNAN

INTERFACES 31:2 36

prise application packages,” Communicationsof the ACM, Vol. 43, No. 4 (April), pp. 63–69.

Two Crows Corporation 1999, Introduction toData Mining and Knowledge Discovery, thirdedition, booklet, Potomac, Maryland, Sep-tember. Retrieved December 9, 2000 fromwww.twocrows.com/booklet.htm

Vakrat, Y.; Pinker, E.; and Seidmann, A. 2000,“Multi-period optimal design of online auc-tions,” Working paper, William E. SimonGraduate School of Business Administration,University of Rochester, October 17.

Vakrat, Y. and Seidmann, A. 2000, “Implica-tions of the bidder’s arrival process on thedesign of online auctions,” Proceedings of theHawaii International Conference on System Sci-ences—33, IEEE Computer Society, January,pp. 1–10.

Varian, Hal 2000, “Estimating the demand forbandwidth,” Working paper, University ofCalifornia, Berkeley, August 29. RetrievedDecember 9, 2000 from www.sims.berkeley.edu/�hal/Papers/wtp/index.html

Wall Street Journal 2000, “Online investing,”Special report, June 12, pp. R1–22.

Weigel, D. and Cao, B. 1999, “Applying GISand OR techniques to solve Sears technician-dispatching and home-delivery problem,” In-terfaces, Vol. 29, No. 1 (January–February),pp. 112–130.

Williams, S. 1999, “Post-Web retail,” Report,Forrester Research, Inc., Cambridge, Massa-chusetts, September.

Willinger, W. and Paxson, V. 1998, “Wheremathematics meets the Internet,” Notices ofthe American Mathematical Society, Vol. 45, No.8 (August), pp. 961–970.

Young, Kung 2000, “The future is cubic,” TheBanker (Financial Times Supplement), June.Retrieved December 9, 2000 from www.algorithmics.com/press/releases/cubic.html

Zakon, Robert H. 2000, “Hobbes’ Internet time-line v5.2,” November. Retrieved December 9,2000 from info.isoc.org/guest/zakon/Internet/History/HIT.html

Zhang, Hui 2000, “Paper list for CS 15-844:Vivisecting the Internet,” Computer ScienceDepartment, Carnegie Mellon University. Re-trieved December 9, 2000 from www.cs.cmu.edu/�hzhang/15-844/papers.html

Prospects for Operations Research in the E-Business Era

Documents

Transcript of Prospects for Operations Research in the E-Business Era