If you're not careful, the dream of informationintegration can turn into a nightmare.

Putting the Enterprise intothe Enterprise System

by Thomas H. Davenport

ENTERPRISE SYSTEMS appear to be a dream come true. Thesecommercial software packages promise the seamless integrationof all the information flowing through a company - flnancial and

accounting information, human resource information, supply chaininformation, customer information. For managers who have struggled,at great expense and with great frustration, with incompatible infor-mation systems and inconsistent operating practices, the promise of anoff-the-shelf solution to the problem of business integration is enticing.

It comes as no surprise, then, that companies have been beating pathsto the doors of enterprise-system developers. The sales of the largest

ARTWORK BY CURTIS PARKER 121

ENTERPRISE SYSTEMS

vendor, Germany's SAP, have soared from less than$500 million in 1992 to approximately $3.3 billionin 1997, making it the fastest-growing softwarecompany in the world. SAP's competitors, includ-ing such companies as Baan, Oracle, and People-Soft, have also seen rapid growth in demand fortheir packages. It is estimated that businessesaround the world are now spending $10 billion peryear on enterprise systems-also commonly re-ferred to as enterprise resource planning, or ERP,systems-and that figure probably doubles whenyou add in associated consulting expenditures.While the rise of the Internet has received most ofthe media attention in recent years, the husinessworld's embrace of enterprise systems may in facthe the most important development in the corpo-rate use of information technology in the 1990s.

But are enterprise systems living up to compa-nies' expectations? The growing numher of horrorstories about failed or out-of-control projectsshould certainly give managers pause. FoxMeyerDrug argues that its system helped drive it intobankruptcy. Mobil Europe spent hundreds of mil-lions of dollars on its system only to abandon itwhen its merger partner objected. Dell Computerfound that its system would not fit its new, decen-tralized management model. Applied Materialsgave up on its system when it found itself over-whelmed by the organizational changes involved.Dow Chemical spent seven years and close to halfa billion dollars implementing a mainframe-hasedenterprise system; now it has decided to start overagain on a client-server version.

Some of the blame for such debacles lies with theenormous technical challenges of rolling out enter-prise systems - these systems are profoundly com-plex pieces of software, and installing them requireslarge investments of money, time, and expertise.But the technical challenges, however great, arenot the main reason enterprise systems fail. Thebiggest problems are business problems. Compa-nies fail to reconcile the technological imperativesof the enterprise system with the business needs ofthe enterprise itself.

An enterprise system, hy its very nature, imposesits own logic on a company's strategy, organization,and culture. (See the table "The Scope of an Enter-prise System.") It pushes a company toward full in-tegration even when a certain degree of business-

Thomas H. Davenport is a professor at the Boston Uni-versity School of Management in Boston, Massachusetts.His most recent book, Working Knowledge: How Organi-zations Manage What They Know, was published inby the Harvard Business School Press.

THE SCOPE OF ANENTERPRISE SYSTEM

An enterprise system enables a company to inte-grate the data used throughout its entire organiza-tion. This list shows some of the many functionssupported by SAP's R/3 package.

FinancialsAccounts receivable and payableAsset accountingCash management and forecastingCost-element and cost-center accountingExecutive information systemFinancial consolidationGeneral ledgerProduct-cost accountingProfitability analysisProfit-center accountingStandard and period-related costing

Human ResourcesHuman-resources time accountingPayrollPersonnel planningTravel expenses

Operations and LogisticsInventory managementMaterial requirements planningMaterials managementPlant maintenanceProduction planningProject managementPurchasingQuality managementRouting managementShippingVendor evaluation

Sales and MarketingOrder managementPricingSales managementSales planning

122 HARVARD BUSINESS REVIEW July-August 1998

ENTERPRISE SYSTEMS

unit segregation may be in its best interests. And itpushes a company toward generic processes evenwhen customized processes may be a source of com-petitive advantage. If a company rushes to installan enterprise system without first having a clearunderstanding of the business implications, thedream of integration can quickly turn into a night-mare. The logic of the system may conflict with thelogic of the husiness, and either the implementa-tion will fail, wasting vast sums of money and caus-ing a great deal of disruption, or the system willweaken important sources of competitive advan-tage, hobbling the company.

It is certainly true that enterprise systems can de-liver great rewards, but the risks they carry areequally great. When considering and implementingan enterprise system, managers need to be carefulthat their enthusiasm about the benefits does notblind them to the hazards.

The Allure of Enterprise SystemsIn order to understand the attraction of enterprisesystems, as well as their potential dangers, you firstneed to understand the problem they're designed tosolve: the fragmentation of information in largehusiness organizations. Every big company col-lects, generates, and stores vast quantities of data.In most companies, though, the data are not kept ina single repository. Rather, the information isspread across dozens or even hundreds of separatecomputer systems, each housed in an individualfunction, business unit, region, factory, or office.Each of these so-called legacy systems may provideinvaluable support for a particular business activ-ity. But in combination, they repre-sent one of the heaviest drags on busi-ness productivity and performancenow in existence.

Maintaining many different com-puter systems leads to enormouscosts-for storing and rationalizingredundant data, for rekeying and re-formatting data from one system foruse in another, for updating and de-bugging obsolete software code, forprogramming communication links between sys-tems to automate the transfer of data. But even moreimportant than the direct costs are the indirectones. If a company's sales and ordering systems can-not talk with its production-scheduling systems,then its manufacturing productivity and customerresponsiveness suffer. If its sales and marketing sys-tems are incompatible with its financial-reportingsystems, then management is left to make impor-

tant decisions by instinct rather than according toa detailed understanding of product and customerprofitahility. To put it hluntly: if a company's sys-tems are fragmented, its business is fragmented.

Enter the enterprise system. A good ES is a tech-nological tour de force. At its core is a single com-prehensive datahase. The datahase collects datafrom and feeds data into modular applications sup-porting virtually all of a company's business activi-ties-across functions, across business units, acrossthe world. (See the chart "Anatomy of an EnterpriseSystem.") When new information is entered in oneplace, related information is automatically updated.

Let's say, for example, that a Paris-based salesrepresentative for a U.S. computer manufacturerprepares a quote for a eustomer using an ES. Thesalesperson enters some basic information aboutthe customer's requirements into his laptop com-puter, and the ES automatically produces a formalcontract, in French, specifying the product's config-uration, price, and delivery date. When the cus-tomer accepts the quote, the sales rep hits a key; thesystem, after verifying the customer's credit limit,records the order. The system sehedules the ship-ment; identifies the best routing; and then, workingbackward from the delivery date, reserves the nec-essary materials from inventory; orders neededparts from suppliers; and schedules assembly in thecompany's factory in Taiwan.

The sales and production forecasts are immedi-ately updated, and a material-requirements-plan-ning list and hill of materials are created. The salesrep's payroll account is credited with the correctcommission, in French francs, and his travel accountis credited with the expense of the sales call. The

The growing numberof horror stories about failed

or out-of-control projects shouldcertainly give managers pause.

actual product cost and profitability are calculated,in U.S. dollars, and the divisional and corporate bal-ance sheets, the accounts-payable and accounts-receivable ledgers, the cost-center accounts, andthe corporate cash levels are all automatically up-dated. The system performs nearly every informa-tion transaction resulting from the sale.

An ES streamlines a company's data flows andprovides management with direct access to a wealth

HARVARD BUSINESS REVIEW |uly-AugU8t 1998 123

ENTERPRISE SYSTEMS

ANATOMY OF AN ENTERPRISE SYSTEM

At the heart of an enterprise system is a central database that draws datafrom and feeds data into a series of applications supporting diverse com-pany functions. Using a single database dramatically streamhnes the flowof information throughout a business.

CustomersSales force

and customerservice reps

Sales anddelivery

applications

Managers andstakeholders

Reportingapplications

V

VHuman

resourcemanagementapplications

Employees

Manufacturing^applications

Back-officeadministrators

and workersSuppliers

of real-time operating information. For many com-panies, these henefits have translated into dramaticgains in productivity and speed.

Autodesk, a leading maker of computer-aideddesign software, used to take an average of twoweeks to deliver an order to a customer. Now, hav-ing installed an ES, it ships 98% of its orders within24 hours. IBM's Storage Systems division reducedthe time required to reprice all of its products from5 days to 5 minutes, the time to ship a replacementpart from 22 days to 3 days, and the time to com-plete a credit check from 20 minutes to 3 seconds.Fujitsu Microelectronics reduced the cycle time for

filling orders from 18 days to a day and a half andcut the time required to close its financial booksfrom 8 days to 4 days.

When System and Strategy Clashclearly, enterprise systems offer the potential of bigbenefits. But the very quality of the systems thatmakes those benefits possible-their almost uni-versal applicability-also presents a danger. Whendeveloping information systems in the past, com-panies would first decide how they wanted to dohusiness and then choose a software package that

124 HARVARD BUSINESS REVIEW July-August 1998

ENTERPRISE SYSTEMS

would support their proprietary processes. Theyoften rewrote large portions of the software codeto ensure a tight fit. With enterprise systems, how-ever, the sequence is reversed. The husiness oftenmust be modified to fit the system.

An enterprise system is, after all, a generic solu-tion. Its design reflects a series of assumptionsabout the way companies operate in general. Ven-dors try to structure the systems to reflect bestpractices, but it is the vendor, not the customer,that is defining what "best" means. In many cases,the system will enable a company to operate moreefficiently than it did before. In some cases, though,the system's assumptions will run counter to acompany's best interests.

Some degree of ES customization is possible. Be-cause the systems are modular, for instance, com-panies can install only those modules that arc mostappropriate to their business. However, the system'scomplexity makes major modifications impracti-cable. (See the insert "Configuring an Enterprise

System.") As a result, most companies installingenterprise systems will need to adapt or even com-pletely rework their processes to fit the require-ments of the system. An executive of one companythat has adopted SAP's system sums it up by saying,"SAP isn't a software package; it's a way of doinghusiness." The question is. Is it the hest way of do-ing husiness? Do the system's technical imperativescoincide or conflict with the company's businessimperatives?

Imagine, for example, an industrial productsmanufacturer that has built its strategy around itsability to provide extraordinary customer service infilling orders for spare parts. Because it is able toconsistently deliver parts to customers 25% fasterthan its competitors-often by circumventing for-mal processes and systems-it has gained a largeand loyal clientele who are happy to pay a premiumprice for its products. If, after installing an ES, thecompany has to follow a more rational hut less flex-ible process for filling orders, its core source of

CONFIGURING AN ENTERPRISE SYSTEMConfiguring an enterprise system is largely a matterof making compromises, of balancing the way youwant to work witb tbe way the system lets you work.You begin by deciding which modules to install.Then, for each module, you adjust the system usingconfiguration tables to achieve the best possible fitwith your company's processes. Let's look more closelyat these two configuration mechanisms:• Modules. Most enterprise systems are modular, en-abling a company to implement the system for somefunctions but not for others. Some modules, such asthose for finance and accounting, are adopted hy al-most all companies that install an ES, whereas others,such as one for human resource management, areadopted hy only some companies. Sometimes a com-pany simply doesn't need a module. A service busi-ness, for example, is unlikely to require the modulefor manufacturing. In other cases, companies choosenot to implement a module hecause they already havea serviceahle system for that particular function orthey have a proprietary system that they believe pro-vides unique benefits. In general, the greater the num-ber of modules selected, the greater the integrationbenefits, but also the greater tbe costs, risks, andchanges involved.

• Configuration tables. A configuration table enablesa company to tailor a particular aspect of the systemto the way it chooses to do business. An organizationcan select, for example, what kind of inventoryaccounting-FIFO or LIFO-it will employ or whether

it wants to recognize product revenue hy geographicalunit, product line, or distribution channel. SAP's R/3,one of the more comprehensive and complex ES offer-ings, has more than 3,000 configuration tables. Goingthrough all of them can take a long time. Dell Com-puter, for example, spent more than a year on the task.

Although modules and configuration tahles let youcustomize the system to some degree, your optionswill be limited. If you have an idiosyncratic way of do-ing business, you will likely find that it is not sup-ported hy an ES. One company, for example, had longhad a practice of giving preferential treatment to itsmost important customers by occasionally shippingthem products that had already been allocated to otheraccounts. It found that its ES did not allow it the fiexi-hility required to expedite orders in this way. Anotbercompany had always kept track of revenues hy hothproduct and geography, but its ES would allow it totrack revenue in only one way.

What happens when the options allowed by the sys-tem just aren't good enough? A company has twochoices, neither of them ideal. It can actually rewritesome of the ES's code, or it can continue to use an ex-isting system and build interfaces between it and theES. Both of these routes add time and cost to tbe im-plementation effort. Moreover, they can dilute theES's integration benefits. The more customized anenterprise system heeomes, the less able it will be tocommunicate seamlessly with the systems of suppli-ers and customers.

HARVARD BUSINESS REVIEW |uly-August 1998 125

ENTERPRISE SYSTEMS

advantage may be at risk. The company may inte-grate its data and improve its proeesses only to loseits serviee edge and, in turn, its customers.

This danger becomes all the more pressing inlight of the increasing ubiquity of enterprise sys-tems. It is now common for a single ES package tobe used by virtually every company in an industry.For example, SAP's R/3 package is heing imple-mented by almost every eompany in the personal

A speedy implementation ofan enterprise system may be awise business move, but a rashimplementation is not.

computer, semiconductor, petroehemical, and to aslightly lesser degree, consumer goods industries.[R/3 is the client-server version of SAP's software;R/2 is the mainframe version.) Such convergencearound a single software paekage should raise asobering question in the minds of chief executives:How similar can our information flows and ourprocesses be to those of our eompetitors before webegin to undermine our own sources of differentia-tion in the market?

This question will be moot if a company's com-petitive advantage derives primarily from the dis-tinctiveness of its products. Apple Computer, forexample, has many problems, but the loss of eom-petitive differentiation because of its ES is not oneof them. With a strong brand and a unique operatingsystem, its computers still differ dramatically fromcompeting offerings. But Apple is an unusual case.Among most makers of personal computers, differ-entiation is based more on service and priee than onproduct. For those companies, there is a very realrisk that an enterprise system could dissolve tbeirsourees of advantage.

Compaq Computer is a good example of a com-pany that carefully thought through the strategicimplications of implementing an enterprise sys-tem. Like many personal-eomputer companies,Compaq had decided to shift from a build-to-stockto a build-to-order business model. Because the suc-cess of a build-to-order model hinges on the speedwith which information flows through a company,Compaq helieved tbat a fully integrated enterprisesystem was essential. At the same time, however,Compaq saw the danger in adopting processes in-distinguishable from those of its eompetitors.

It realized, in particular, that in a build-to-orderenvironment an important advantage would accrueto any company with superior capabilities for fore-casting demand and processing orders. Compaqtherefore decided to invest in writing its own pro-prietary applications to support its forecasting andorder-management processes. To ensure that thoseapplications would be compatible with its ES,Compaq wrote them in the computer language

used hy its ES vendor.

Compaq's course was not the obvi-ous one. It cost the company consider-ably more to develop the proprietaryapplication modules than it wouldhave to use the modules offered hy theES vendor. And using customized ap-plications meant forgoing some of theintegration benefits of a pure enter-prise system. But Compaq saw the de-cision as a strategic necessity: it was

the only way to protect a potentially critical sourceof advantage.

For companies that compete on cost rather thanon distinctive products or superior customer ser-vice, enterprise systems raise different strategic is-sues. Tbe huge investment required to implementan ES at large companies - typically ranging from$50 million to more than $500 million-need to beweighed carefully against the eventual savings thesystem will produce. In some cases, companiesmay find that by forgoing an ES they ean actuallygain a cost advantage over eompetitors that are em-bracing the systems. They may not have the mostelegant computer system or the most integrated in-formation flows and processes, but if eustomers areconcerned only with price, that may not matter.

Air Products and Chemicals, for example, sawthat many of its competitors were installing large,complex enterprise systems. After a thorough eval-uation, it decided not to follow their lead. Its man-agers reasoned that the cost of an ES might force thecompany to raise its prices, leading to lost sales insome of the commodity gas markets in which itcompetes. The company's existing systems, whilenot state-of-tbe-art, were adequate to meet itsneeds. And since the company had no plans to ex-change information electronically with competi-tors, it didn't worry about being the odd man out inits industry.

Of course, the long-term productivity and con-nectivity gains created by enterprise systems areoften so compelling that not adopting one is out ofthe question. In the petrochemicals industry, forexample, enterprise systems have improved theflow of information through the supply chain to

126 HARVARD BUSINESS REVIEW [uly-August 1998

ENTERPRISE SYSTEMS

such a degree that they have become a de facto oper-ating standard. Because participants in the industryroutinely share information electronically, itwould today be hard for a company to survive in thebusiness without an ES. Still, the cost of implemen-tation should be a primary concern. It will often bein a company's interest to go ahead and rework itsprocesses to fit the system requirements. The alter-native-customizing the system to fit the processesor writing proprietary application modules - willsimply be too expensive to justify. As the CEO ofone large chemical firm says, "Competitive advan-tage in this industry might just come from doingthe best and cheapest joh at implementing SAP."

The Impact on an OrganizationIn addition to having important strategie implica-tions, enterprise systems also have a direct, andoften paradoxical, impact on a company's organiza-tion and culture. On the one hand, by providinguniversal, real-time access to operating and finan-cial data, the systems allow companies to stream-line their management structures, creating flatter,more flexible, and more democratic organizations.On the other hand, they also involve tbe centraliza-tion of control over information and the standard-ization of processes, which are qualities more con-sistent with hierarchical, command-and-controlorganizations with uniform cultures. In fact, it canbe argued that the reason enterprise systems firstemerged in Europe is that European companiestend to have more rigid, centralized organizationalstructures than their U.S. counterparts.

Some executives, particularly those in fast-grow-ing high-tech companies, have used enterprise sys-tems to inject more discipline into their organiza-tions. They see the systems as a leverfor exerting more management con-trol and imposing more-uniform pro-cesses on freewheeling, highly entre- ,preneurial cultures. An executive at O W n lOiJlC O na semiconductor company, for exam- ^pie, says, "We plan to use SAP as abattering ram to make our eulture lessautonomous." The manager of the ESimplementation at a computer company expressesa similar thought: "We've had a renegade culture inthe past, but our new system's going to make every-body fall into line."

But some companies have the opposite goal.They want to use their enterprise systems to breakdown hierarchical structures, freeing their peopleto he more innovative and more flexible. TakeUnion Carbide. Like most companies implement-

ing enterprise systems. Union Carbide is standard-izing its basic business transactions. Unlike manyother companies, however, the leaders of its ESproject are already thinking in depth about howthe company will he managed differently when theproject is completed. They plan to give low-levelmanagers, workers, and even customers and suppli-ers much hroader access to operating information.Standardizing transactions will make Union Car-bide more efficient; sharing real-time informationwill make it more creative.

For a multinational corporation, enterprise sys-tems raise another important organizational ques-tion: How much uniformity should exist in the wayit does business in different regions or eountries?Some big companies have used their enterprise sys-tems to introduee more eonsistent operating prac-tices across their geographically dispersed units.Dow Chemical, for instance, became an early con-vert to enterprise systems beeause it saw them as away to cut costs by streamlining global financialand administrative processes. (A good idea in prin-ciple, although it became much more expensive toachieve than Dow had anticipated.) Some largemanufacturers have been even more ambitious,using the systems as the basis for introducing aglohal lean-production model. By imposing com-mon operating processes on all units, they are ableto achieve tight coordination throughout theirbusinesses. They can rapidly shift sourcing, manu-facturing, and distribution functions worldwide inresponse to changing patterns of supply and demand.This capability allows them to minimize excessmanufacturing capacity and reduce both compo-nent and finished-goods inventory.

Owens Corning, for example, adopted an ES to re-place 211 legacy systems. For the company to grow

An enterprise system imposes itscompany's strategy,

culture, and organization.

internationally, its chief executive, Glen Hiner, feltit was critical to coordinate order-management,financial-reporting, and supply chain processesacross the world. Having implemented the systemand established a new global-procurement organi-zation, the company is now able to enter into larger,more advantageous international contracts for sup-plies. Finished-goods inventory can be tracked daily,both in company warehouses and in the distribution

HARVARD BUSINESS REVIEW July-August 1998 127

ENTERPRISE SYSTEMS

channel, and spare-parts inventory has been reducedby 50%. The company expects to save $65 million hythe end of 1998 as a result of its adoption of theseglobally coordinated processes.

For most companies, however, differences in re-gional markets remain so profound that strictprocess uniformity would be counterproductive. Ifcompanies in such circumstances don't allow their

Enterprise systems can delivergreat rewards, but the risks theycarry are equally great.

regional units to tailor their operations to local cus-tomer requirements and regulatory strictures, theyrisk sacrificing key markets to more flexible com-petitors. To preserve local autonomy while main-taining a degree of corporate control-what mightbe called a federalist operating model - a very differ-ent approach to enterprise systems needs to hetaken. Rather than implement a single, glohal ES,these companies need to roll out different versionsof the same system in each regional unit, tailored tosupport loeal operating practices. This approachhas been taken by a number of large companies, in-cluding Hewlett-Packard, Monsanto, and Nestle.Tbey establish a core of common information-financial, say-that all units share, but they allowother information-on eustomers, say-to be col-lected, stored, and controlled locally. This methodof implementation trades off some of the purity andsimplicity of the enterprise system for greater mar-ket responsiveness.

The federalist model raises what is perhaps themost difficult challenge for a manager implement-ing an ES: determining what should be commonthroughout the organization and wbat sbould beallowed to vary. Corporate and business-unit man-agers will need to sit down together-well beforesystem implementation begins-to think througheach major type of information and each majorprocess in the company. Difficult questions needto he raised: How important is it for us to processorders in a consistent manner worldwide? Does theterm "customer" mean the same thing in everybusiness unit? Answering such questions is essen-tial to making an ES successful.

Different companies will, of course, reach verydifferent decisions about the right balance betweencommonality and variability. Consider tbe starklydifferent approaches taken by Monsanto and

Hewlett-Packard. Monsanto's managers knew thatdifferent operating requirements would precludethe complete standardization of data across itsagrochemical, biotechnology, and pharmaceuticalsbusinesses. Nevertheless, they placed a high prior-ity on achieving the greatest possible degree ofcommonality. After studying the data require-ments of each business unit, Monsanto's managers

were able to standardize fully 85% ofthe data used in the ES. The companywent from using 24 coding schemesfor suppliers to using just one, and itstandardized all data about materialsusing a new set of substance identifi-cation codes. While customer and fac-tory data have not been fully stan-dardized - differences among the units'

customers and manufacturing processes are toogreat to accommodate common data-Monsantohas achieved a remarkable degree of commonalityacross a diverse set of global businesses.

At Hewlett-Packard, a company with a strongtradition of business-unit autonomy, managementhas not pushed for commonality across the severallarge divisions that are implementing SAP's enter-prise system. Except for a small amount of commonfinancial data necessary to roll up results for corpo-rate reporting, HP's federalist approach gives all thepower to the "states" where ES decisions are con-cerned. This approach fits the HP culture well, butit's very expensive. Each divisional ES has had to beimplemented separately, with little sharing of re-sources. Managers estimate tbat well over a billiondollars will be spent across the corporation beforethe various projects are completed.

Doing It Right at Elf AtochemConsidering an ES's far-reaching strategic and orga-nizational implications, the worst thing a companycan do is to make decisions about a system based ontechnical criteria alone. In fact, having now studiedmore than 50 businesses with enterprise systems,I can say with some confidence that the companiesderiving the greatest benefits from their systemsare those that, from the start, viewed them primar-ily in strategic and organizational terms. Theystressed the enterprise, not the system.

Elf Atochem North America, a $2 billion regionalchemicals suhsidiary of the French company ElfAquitaine, is a good case in point. Following a se-ries of mergers in the early 1990s, Elf Atochemfound itself hampered by the fragmentation of criti-cal information systems among its 12 businessunits. Ordering systems were not integrated with

128 HARVARD BUSINESS REVIEW fuly-August 1998

ENTERPRISE SYSTEMS

production systems. Sales forecasts were not tied tobudgeting systems or to performance-measure-ment systems. Each unit was tracking and report-ing its financial data independently. As a result ofthe many incompatible systems, operating datawere not flowing smoothly through the organiza-tion, and top management was not getting the in-formation it needed to make sound and timelybusiness decisions.

Tbe company's exeeutives saw that an enterprisesystem would be the best way to integrate the dataflows, and they decided to go with SAP's R/3 sys-tem, wbich was rapidly hecoming the standard inthe industry. But they never labeled the ES projectas simply a technology initiative. Rather, theyviewed it as an opportunity to take a fresh look atthe company's strategy and organization.

Looking beyond the technology, the executivessaw that the real source of Elf Atochem's difficul-ties was not the fragmentation of its systems butthe fragmentation of its organization. Although the12 business units shared many of the same cus-tomers, each unit was managed autonomously.From the customer's perspective, the lack of conti-nuity among units made doing husiness with thecompany a trial. To place a singleorder, a customer would frequentlyhave to make many different phonecalls to many different units. And topay for the order, the customer wouldhave to process a series of invoices.

Inside the company, things wereequally confused. It took four days-and seven bandoffs between depart-ments-to process an order, even though only fourhours of actual work were involved. Because eachunit managed inventory and scheduled productionindependently, the company was unable to consoli-date inventory or coordinate manufacturing at thecorporate level. More than $6 million in inventorywas written off every year, and plants had to be shutdown frequently for unplanned production-linechanges. And because ordering and production sys-tems were not linked, sales representatives couldn'tpromise firm delivery dates, which translated intolost customers.

Management knew that in the petrochemicalsbusiness, where many products are commodities,the company that can offer the hest customer ser-vice often wins the order. So it structured the im-plementation of its ES in a way that would enableit to radically improve its service levels. Its goal wasto transform itself from an industry laggard into anindustry leader. Even though many competitorswere also adopting the R/3 package. Elf Atochem

knew that if it could achieve a tighter, smoother fitbetween its business processes and the system, itcould gain and maintain a service advantage.

The eompany decided to focus its efforts on fourkey processes: materials management, productionplanning, order management, and financial report-ing. These cross-unit processes were the ones mostdistorted by the fragmented organizational struc-ture. Moreover, they had the greatest impact on thecompany's ability to manage its customer relation-sbips in a way that would both enhance eustomersatisfaction and improve corporate profitahility.Each of the processes was redesigned to take full ad-vantage of the new system's capabilities, in particu-lar its ability to simplify the flow of information.Layers of information middlemen-once necessaryfor transferring information across incompatihleunit and corporate systems-were eliminated inorder to speed the flow of work and reduce the like-lihood of errors.

To maintain its focus on the customer, the com-pany chose to install only those R/3 modules re-quired to support the four targeted processes. It didnot, for example, install the modules for human re-source management or plant maintenance. Those

Those companies that stressedthe enterprise, not the system,gained the greatest benefits.

functions did not have a direct impact on cus-tomers, and the existing information systems thatsupported them were considered adequate.

Elf Atochem also made fundamental changes toits organizational structure. In the financial area,for example, all the company's accounts-receivableand credit departments were combined into a singlecorporate function. This change enahled the com-pany to consolidate all of a customer's orders into asingle account and issue a single invoice. It also al-lowed the company to monitor and manage overallcustomer profitability-something that had beenimpossible to do when orders were fragmentedacross units. In addition. Elf Atochem combined allof its units' customer-service departments into onedepartment, providing each customer with a singlepoint of contact for checking on orders and resolv-ing problems.

Perhaps most important, the system gave ElfAtochem the real-time information it needed toconnect sales and production planning-demand

HARVARD BUSINESS REVIEW July-August 1998 129

ENTERPRISE SYSTEMS

and supply-for the first time. As orders are enteredor changed, the system automatically updates fore-casts and factory schedules, which enables thecompany to quickly alter its production runs in re-sponse to customers needs. Only one other com-pany in the industry had this capability, whichmeant that Elf Atochem gained an important edgeover most competitors.

The company understood, however, that justhaving the data doesn't necessarily mean the data

Only a general manager isequipped to act as a mediatorbetween the imperatives of thetechnology and of the business.

will he used well. Computer systems alone don'tchange organizational behavior. It therefore estab-lished a new position-demand manager-to be thefocal point for the integrated sales and production-planning process. Drawing on the enterprise sys-tem, the demand manager creates the initial salesforecast, updates it with each new order, assessesplant capacity and account profitability, and devel-ops detailed production plans. The demand man-ager is ahle to schedule a customer's order-andpromise a delivery date-up to six weeks ahead ofproduction. Previously, production could he allo-cated to individual orders no more than a week inadvance. Now central to the company's operation,the role of demand manager could not even have ex-isted in the past because the information needed toperform it was scattered all over the company.

The way Elf Atochem is managing the imple-mentation effort also reflects the breadth of itsgoals. The project is being led by a 6o-person coreimplementation team, which reports to a memberof the company's executive committee. The teamincludes both business analysts and informationtechnologists, and is assisted by a set of so-calledsuper users, representing the business units andcorporate functions. These super users help ensurethat decisions about the system's configuration aremade with the broadest possible understanding ofthe business. They also play a crucial role in ex-plaining the new system to their respective depart-ments and training people in its use.

The team is installing the ES one business unit ata time, with each unit implementing the same sys-tem configuration and set of procedures for order

processing, supplier management, and financialreporting. The unit-by-unit process ensures thatthe effort is manageable, and it also helps the teamrefine the system and the processes as it proceeds.For example, the second unit to implement the sys-tem found that it didn't adequately support bulkshipments, which are the main way the unit gets itsproducts to customers. [The first unit uses packageshipping for all its orders.) The system was thenmodified to support bulk as well as package ship-

ping, and the new configuration be-came the new standard.

Using the large and broadly repre-sentative implementation team, to-gether with the unit-hy-unit rollout.Elf Atochem has been able to staff theeffort mainly with its own people. Ithas had to engage only nine outsideconsultants to assist in the project-far fewer than is usually the case. Thereliance on internal resources not

only reduces the cost of the implementation, it alsohelps ensure that Elf Atochem's employees willunderstand how the system works after the consul-tants leave.

Elf Atochem's ES is now more than 75%complete-9 of the 12 business units are up andrunning on the new system-and the rollout isahead of schedule and under hudget. Customer sat-isfaction levels have already increased, and thecompany is well on the way to its goal of confirm-ing 95% of all orders with one call, a dramatic im-provement over the previous average of five calls. Inaddition to the service enhancements, the companyis operating more efficiently. Inventory levels, re-ceivables, and labor and distribution expenditureshave all been cut, and tbe company expects the sys-tem will ultimately reduce annual operating costshy tens of millions of dollars.

The Role of ManagementEvery company that installs an ES struggles withits cost and complexity. But the companies thathave the biggest problems-the kind of problemsthat can lead to an outright disaster-are those thatinstall an ES without thinking through its full busi-ness implications.

Managers may well have good reasons to movefast. They may, for example, have struggled foryears with incompatible information systems andmay view an ES as a silver bullet. They may belooking for a quick fix to the Year 2000 problem(enterprise systems are not infeeted with the mucb-feared millennium bug). Or they may be trying to

130 HARVARD BUSINESS REVIEW July-August 1998

ENTERPRISE SYSTEMS

keep pace with a competitor that has already imple-mented an ES. The danger is that while an enter-prise system may help them meet their immediatechallenge, the very act of implementing it may cre-ate even larger problems. A speedy implementationof an ES may be a wise business move; a rash imple-mentation is not.

A numher of questions should be answered he-fore any decisions are made. How might an ESstrengthen our eompetitive advantages? Howmight it erode them? What will be the system'seffeet on our organization and culture? Do we needto extend the system across all our functions, orshould we implement only certain modules?Would it be better to roll the system out globally orto restrict it to certain regional units? Are thereother alternatives for information managementthat might actually suit us better than an ES?

The experience of Elf Atochem and other suc-cessful adopters of enterprise systems underscoresthe need for careful deliberation. It also highlightsthe importance of having top management directlyinvolved in planning and implementing an ES. Not

only is Elf Atochem's executive committee over-seeing its ES project, but its entire hoard reviewedand approved the plans. At Compaq, the decision togo with an ES was also made at the board level, andthe senior management team was involved withthe implementation every step of the way.

Many chief executives, however, continue toview the installation of an ES as primarily a techno-logical challenge. They push responsibility for itdown to their information technology depart-ments. Because of an ES's profound business impli-cations-and, in particular, the risk that the tech-nology itself might undermine a company'sstrategy-off-loading responsibility to technolo-gists is particularly dangerous. Only a general man-ager is equipped to act as the mediator between theimperatives of the technology and the imperativesof the husiness. If the development of an enterprisesystem is not carefully controlled by management,management may soon find itself under the controlof the system. ^

Reprint 98401 To order reprints, see the last page of this issue.

T

"It appears the Ethics Committee has gone over the edge.

HARVARD BUSINESS REVIEW luIy-August 1998 131