251
Implementation as mutual adaptation of te ology and organization Dorothy LEONARD-BARTON Hwvmd Graduate S&o1 of BuMess, Cmbidge,
MA 02123, LISA.
Final version received February 1988
1. lWrodu&on New production technologies are widely recognixed as competitive weapons [51,61]. However, their implementation, i.e, getting them up and p a, ; g b d&i oper&c~s, & a$ I~Q PC ~h~lbnux.3u~;?z.z-i-;i~ ing a managerial problem as their invention (see e.g. [47]). The initial implementation stage, that is, the period during which the technology is first removed from its laboratory setting and introduced into the user environment is especially critical. This paper places that process of initial implementation under the microscope in order to understand better its dynamics - the mutuai adaptation that occurs between technology and user environment as developers and users strive to wring productivity increases from the innovation to benefit the whole organization (see fig. 1). The conceptual framework describing the implementation process and presented in the following pages derives from the intellectual interplay between academic literature and empirical fieldwork. Thus it is “grounded theory” [25]. Many concepts from innovation research in fields as diverse as agriculture and engineering and in settings as varied as steel mills and local government offices proved relevant and were empirically confirmed. However, as elabnrpted below, the only academic perspective that captured fully the important interaction between tec*hno@cai and organizational change during the initial stages of implementation was a study of educational innovations - research setting far removed from the
Research Policy 17 (1988; X1-267 North-Holland
environment of interest here. Therefore, the contribution attempt herein is to extend as well as to integrate previous theory. The fieldwork was comprised of 12-in-depth case studies of new technology (10 of them develoned e&rely internally) introduced into the operations of large corporations within the past five years. Although most were retrospective (historical) studies, one was a longitudinal (three-year) clinical study and three spanned four to eight months of implementation activity. Thus the methodology employed here is the replicated case study approach explained by Yin [67! and compar4 by Campbell in his foreword to Yin’s book to repeated experiments - with the additional
TECHNICAL 0 DELIVERY SYSTEM a .DSPSnPMANPC PPltCPlA _I.. Y . . . . I.-n..Yb “ill, L-m,,l
Fig. 1. Mutual adaptation of technology and organimion.
0048-7333/88/$3.50 6 1988, Eisevier Science Publishers B.V. (North-Hoiland)
D. Leonard-Bartun / Impkmentation CISmutual adhptatkm
252
refinement that one of the “experiments” was studied in reaMme. ’ The objective of this article is to present some broad, generic concepts that I believe (based on literature review) describe most processes of initial technology implementation, wherever they occur, although the research described here was conducted only on production technologies and in very large companies. * The major thesis of the framework is that initial implementation of a new technology is an extension of the invention process. That is, instead of the predictable realization of a preprogrammed plan, implementation is a dynamic process of mutual adaptation between the technology and its environment. As Van de Ven notes 1641““Innovations not only adapt to existing organizational and industrial arrangements, but they also transform the structure and practice of these environments” (p. 591). This thesis has a number of implications, each of which is explored in a section of this paper.
1
..e._n.. w.-m-+-.qca4 note [631, balsruuy wlIJuubuGu are by no means casual or unstructured investigations. In the 12 cases used for illustrative purposes here, interviews ranged from only eig 1 people (in the instance of one very small project) to a much more representative 31 in another; in two cases, large-scale personal or telephone surveys were conducted (N = 145 and N = 268), at multiple company sites, and in the cases of MRPiI and the Purchasing information Systems, “sub-ca&’ of implementation situations at each of four different plants were nested within the two overall cases Ccxqmy archival data, observations derived from nttanAi*n rrVoot;nnr ~nri #.““““’ nddichd wwrw+c were nlq ~4. Y\.ULY1.* .U_U..puI U._ __---_ ---.- -_ Data have %zn reviewed and confirmed by interviewees in all of the cases. Portions of six of the cases, after extensive review by the companies involved, have been published for classroom use as teaching vehicles and are listed in the references. 2 The medium of a relatively short article imposes two limitations wher describing this d&ynamicprocess systematically: (a) Ds&:.:ET 2~. s~&~~pfified. Therefore, elements are .J*K:,_..A ..n cxtrcnic ends “X rrc P 0 u-uuu4 r-tn.m Crn%ll” gr UGllIl~ OJ (e.g., &‘ “*II-. %rge”) without any attempt to accommodate all the gradations in between and categories (e.g., “technical” or “ir,:,astructural”) are presented as if they were mutually exclusive, whereas in real life, they may overlap. (b) Within the general framework, there is ample latitude to formulate contingency theory - propositions about the specific conditions under which certain descriptors predominate or about the relationship of certain predominant conditions to the success or failure of the implementation (231. These refinements and extertiions of the framework are covered elsewhere, along with a more detailed examination of the cases [42].
As Yirl and caiipb&:
.
. .
.--e-e.*
baasa
First, t& adaptation process is necessary because a technology almost never fits p&e&y into the user environment. 3 Even though developers rtiuce ‘the uncertainty inherent in the innovation process by technical iterations and prototyping, %s soon as [the technologyJ gets into the hands of the.. . production department the complexity will increase again” [lo, p. 291. This “complexity,” it will be argued and illustrated below, takes the form of misalignments (poor fits) between the technology and: (a) technical requirements, (b) the system through which the technology is delivered to users, or (c) user organization performance criteria. These misalignments must be addressed if the implementation is to succeed. Second, these misalignments can he corrected by altering the technology or changing the environment - or both. These alterations (termed “cycles of adaptation”) vary in magnitude - both for the technology and the user environment and elicit different levels of effort and resources. B~CWXthis mutual adaptation process is interactive and dynamic, technology may determine structure or vice versa, depending upon when the relationship is observed (see 1241).In fact, it is assumed that an organization’s progress obtains from a beneficial disequilibrium between technology and structure, akin to the creative tension between invention and efficiency noted by many authors [1;27;36]. Thus, the approach taken here is akin to the systems theory concept of “equifinality” [18] in that success can be attained through different sets of “equally effective alternative designs” (p. 520). Third and finally, the term “adaptation” is intended to be neutral. Some authors have assume that adjustments in an innovation as it diffuses are
3 Berman [6] recognixes a spectrum
of implematation strategies, from heavily preprogrammed to adaptive. Combining l ha* rnnmnt with nnd Mnnsnn’s of -a.-. --“-er’ .-e-m Pd7’c - _--__-_-_____ _ Ccj~l ._ _ * exnlsnation _.“ _---_--.. “levels of originality” yields the insight that technologies at the “borrowing level” (i.e., those adopted as is from a source outside the organization) are more amenable to the pr* grammed approach than the kinds of technologies studied here. Becwase the technologies examined herein are mternally __. ravr . (“origination level”), they have not been previ+l@lnmad ously implemented at multiple sites and are therefore still relatively immature. Their interaction with the organization is not readily predictable (see [39] for a discussion of the resultant organizational prototyping); therefore, some adaptation is almost inevitable.
D. Luonard-Bartan / Implementation as mutual aabptatbn
undesirable. Zaltman et al. [68, p. 1661 for inctancP: -_* lint “forces altering the innovation” as one ‘W-W’, form of “resistance” during initial implementation. IIowever, as noted below, others see adaptation as a positive process of “re-invention.” I will argue that in fact, the mutual adaptation process can take either beneficial or detrimental forms as has been found in an educational setting 17) and that management of the implementation process makes L difference.
2. Relevant concepts from literature The literature on organizational innovation is vast and diverse. The studies focusing on invention [2;32;34;56;65], on the initial adoption decision [44;21] or technology transfer among organizations [13;57] are less relevant to understanding post adoption issues than is the research focused specifically on implementation 121,501.However, frequently the tec’hnoiogies studied in this iaikr research tradition are “abstract” innovations such as federal IS] or local government policies 1661 rather than the “concrete non-living” 1581 machines, tools and software targeted in this paper. The concept of adaptation central to the framework developed herein has been researched elsewhere more as a process of either technology mutation or organizational impacts than as an interaction between the two. Thus researchers approaching the topic from the perspective of innovation diffusers have noted the phenomenon of “re-invention,” i.e., alteration of the original innovation as users change it to suit their needs [52] or use it in ways unforeseen by developers [31]. User participation in aligning or altering the technology to fit their needs has been posited as an important influence on user satisfaction - if not necessarily on the quality of the final product [28]. . D ac.anwm sk?ymg tha r\ff /\f &la” k!mA IIunaY’VII “1 t?w .11” tef4-l CIIIL*.Gawdws nology from design engineers to manufacturing have noted the important of the receiver’s “imbedded technology capability,” i.e., technical skills and resources, presumably because the recipients of the technology need to continue shaping the innovation once it is in their hands [59]. Souder [62] suggests that a “Task-Dominated” model of the hand-off is most successful for difficult projects, in part because the technology “moves back
253
and forth among several parties to adjust, modify,
fit and bend it...” (,D-227). As these examples illustrate, most of these literatures view innovation essentially from the perspective of the developers, addressing the issue of how the technology should be altered to fit its ultimate environment. There is an equally rich literature that addresses a complementary issue: how the oetion is affected by, and altered after, the introduction of the new technology. Much of the debate in this tradition focuses on whether individual wo&fs are “de&illed” Ill] or skilled by the advent of a new technology 1691but a few researchers such as Damanpp~w and Evan [16] have conceived of technological impacts in terms of overall organizational response. Whereas Evan noted the general tendency for organizational adjustment to lag behind technological change, Ettlie [20] found that better performing organizations synchronize the adaptation of administrative policies with the introduction of the technology. Most of the innovation literature relevant to ;~~i~~~~A~A~~~
. __^__^^ 133uc3, ~~=~=f~i~, fade
~zi ~~~~i
what can be done to the technology to adjust it to its environment or what is done to the organization & the technology. Although the sociotechnical literature explicitly addressed the possibility of altering both technology and organization, it does so primarily at the level of task and individual job redesign l&j. -Therefore, this paper takes a deliberately cross-displinary stance in suggesting that initial implementation of technical innovations is best viewed as a process of mutual adaptation i.e., the re-invention of the technoiogy and the simuitaneous adaptation of the organization. Such mutual adaptation was found to be associated with SUCcessful implementation in a study of educational innovations among school districts [?I. The research reported here extends those observations to the case of relatively “concrete” technologies. It alcn Aenntto by VieWbIg UIVY -Vr-‘_- frnm _____-pr&QhJs e_t&lg that mutual adaptation process as occurring at multiple levels within the organization. Adaptation does not necessarily occur in equal proportions to both technology and organization in all cases, of course, but I will argue that the successful management of technology transfer from de= velopers to users requires that managers recognize and assume responsibility for both technical and organizational change. See table I for a summary
254
D. Leonard-Barton
/ Implementation as mutual adaptation
Table 1 Description of twelve cases used in derivation of framework A
Polymer Process: A chemical reaction producing a new polymer that is used to produce toner for copier machines. Developed in corporate research laboratories of “Alpha” corporation on the East Coast and transferred into manufacturing operations in the Midwest.
Highly successful. In full use in plant as scheduled.
B
XCON (expert CONfigurer): An expert system that checks the configurations of complex computer systems before they are ordered from manufacturing to make certain that all needed components are included in the design and that they are compatible, i.e., that the system can be manufactured. Originated at Carnegie Mellon Unive.sity, tranafcz~edinto A! system group withb manufactluring at Dig&I Equipment Corp. and then into manufacturing operations to be used by 14 to 17 technical editors. [Sl]
Highly successful. In use of all 16 technical editors, for 98% of orders.
C
Rolling Mills Diagnostic: A mathematical computer program designed to detect the source of irregularities oocurring in the rolling of ahuninum sheeting. Originated in “World Aluminum Corporation” Research Laboratories, transferred into rolling mills.
Highly successful. In regular use by plant engineers.
D
MRPII: a manufacturing resources planning software program that tracks data on the entire manufacturing process, from order purchasing and receipt of parts, to their incorporation into products and shipment to customers. Emphasis in study on purchasing Module, at three plants.
Highly suc4ze&ul in terms of use.
E
Purchase: an automated purchasing system comparable in diction to the purchasing module of the MRPII system. Tracks parts from orders through receipt and .lispersement to stock or production floor. Studied at three plants.
Highly successful in terms of use.
F
CcrmputerAided
[email protected]&ng: -2. deqn tool intended for use by engineers who are designing circuit packs. CAE generates a thematic representation (drawing) of the circuitry that then goes to drafting for fine-tuning and on to a manufacturing site to guide the production of prototype circuit boards. Designed in Corporate Information Systems group of “Baker” Corporation and transferred to dozens of organizations nationwide. [40]
Moderately high success. use in 70% of intended cases after 18 months in field. (Use strongly mandated.)
G
XSEL: An expert system designed for use by sales representatives in a computer company that sells complex customized systems. XSEL checks the sales reps’ configurations for accuracy and completeness before the order is submitted. Designed in AI systems group (located in manufacturing) and transferred to more than 2000 sales representatives nationwide. (391
Moderate success. Use by 50% intended users 18 months after fulIScale introduction to field. (Use not mandated, with scattered exceptions.)
Structured System Analysis: A structured methodology for constructing software, used by systems analysts in their first attempts to “scqe out” the business their system is to serve. Accompanied by computerized aids that help generate the diagrams and stand&-,-d notations used to communicate the analysis (thus akin to v ;?rdprocessing). Developed in Corporate Information Systems group of “Cairn” C.qoration and transferred &rot&hut the corporation worldwide to hundreds of design groups and thousands of system analysts. [37]
Moderate success. Adoption by an average of 60% intended users 36 months after fullscale introduction to field. (Use weakly mandated.) - .I rauure at first. Eventually implemen ted, after months of difficulty.
Copper Process: ‘The substitution of copper for silver in the production of hybrid circuit boards. Originated in northern research laboratories, transferred to southern manufacturing site:. Smartwave: An expert system (type of artificial intelligence that mimics human judgments and can represent heuristics) designed to diagnose defects occurring during the process of passing circuit boards over a molten wave of solder to connect lead wires on the backs of the boards. Developed in a manufacturing AI systems group, transferred to one manufacturing site. [63]
Implementation failure in original form Revived in different form after study ended.
D. Leonard-Barton / Implementation as mutuaI adkptation Table 1 (continued)
L
of animal bones and hides into gelatin. The original process, taking six weeks, involves decomposition in a lime pit. Solagcn takes 48 hours. Originated in Kodak Research Laboratories and implemented on a pilot plant b&s at Kodak Park in Rochester, New York. [17]
pilot planp. withdrawn from active
Jumping Ring Circulator: An electromechanical pump designed to stir recycled aluminum scrap into a molten metal bath. Designed in Corporate Research Laboratories of “World Ahuninum Corporation,” manufactured outside and implemented on an experimental basis in rolling mills. [38)
Total failure. Withdrawn from operations after months of expeGmentation_
SMSlS.
description of the cases used hereafter to illustrate the argument.
is,
3. Implementation misalignments
ment completely. For instance, it is not feasible to build a $2.5 millicn furnace in a laboratory to test a new metallurgical process. Some technologies are notoriously subject to “scabup” problems when they go from laboratory prototype to full system. For example, chemical processes can react unpredictably when brought from petri dish quantities up to full production volume and large integrated software programs may “crash” when all the modules of software are finally interconnected. Therefore technologies often fail during initial implementation 4 to meet their technical targets. i.e., their specifications. Project managers are left with the options of returning to the drawing boards and starting over, continuing to “tinker” with the technology, or rethinking the specifications, and hence redefining implementation success. The case of the Solagen process to produce gelatin for Eastman Kodak’s film products illustrates the dilemma (see [17]). The more than W&year-old process of gelatin making entails the slow (4 to IO-week decomposition of animal bones and hide in a lime pit until an expert determines that the glutinous mass (ossein) has the right consistency by squeezing a handful, and then
The implementation of new technologies seems inevitably to occasion temporary losses of productivity [27]-often more than anticipated. Just as invention is often triggered by the recognition of performauce gaps, 155, p. 36; so adaptation is precipitated by implementation misalignments mismatches between the technology and the organization recognized at the time of initial or trial use. In the 12 cases studied, these misalignments can be categorized as one of three types (although each type can be stimulated by more than one cause): (1) Technical: the technology with its original specifications or with the production process into which it is introduced; (2) Delivery System: the technology with user organization infrastructure (supporting hardware, software or educational programs); (3) Value: the technology with job performance criteria in the user organization. Each of these kinds of misalignment is discussed below and illustrated with reference to one or more of the 12 cases studied.
before their creators have solved all of the basic scientific problems.
ninrkincs ysAs”-.b
3.1.1. Misalignment of technology with it: original specifications It is unlikely that anyone would attempt to implement a technology unless some level of technical feasibility has been established in a laboratory setting. However, time pressures or miscalculations sometimes conspire to push technologies out of the nest before they can really fly [15], that
ia 1,
tr\ C”
nhca~~e ““UYI.”
h&-r AA”..
;t .L
rrodc ywru
~~r~~r UWUJ
frcpl
ES
phenomenon is not confined te process technol~a of course. The famous project SAPPHO, which examines cases
4 This
of s~ccessf’~l and unsuccessful innovation, reported modifications after commercial sales in 53 percent of the chemical products and 80 percent of the instruments [N]. MO~~XMX there can be severe problems in meeting externtiy set s@fications, e.g., those originating from government environmental regulations, but here we consider only those specifycations set by the corporation itself.
256 fingernail. This mass is Str&efj off from as
then boiled and the gelatin many as eight Successive “cooks.” The Output from each of the eight batches is stored separately and graded for quality and type of gelatin. III the first “cook,” the gelatin usually reaches a level of transparency highly desirable for use in making film - an average of 0.013 in “yellow density,” while successive “cooks” yield higher density that are more appropriate for less color-set&i+% applications. Solagen, a chemical pXXe= reducing the decomposition period to 48 hours, was developed in the Kodak corporate research laboratories and tested in a pilot plant. The new technology offered a number of potential benefits beyond the time savings: a more predictable ;Ind controllable process and output that was tiot only more susceptible to scientific contrrl bGt at least as good as current production output for use in the fihn and paper processes downstream. However, production on a large scale would require building a new plant at an estimated cost of somewhere between $17.5 M and $46 M. Moreover, although output from the Solagen process matched current levels of the lowest yellow density (i.e., O.O134XI4), b&e project’s technical targets specified a yet lower level: 0.011. Technicians working on the project insisted that that level, which had been reached on small laboratory batches, would be achieved in time at full scale, if the project continued. The misalignment between targeted and achieved specifications was a major issue. 3J.2. Misalignment of technology with praduction process stage of knowledge Jaikumar and Bohn’s [29] characterization of production processes accoroing to the “stages of knowledge” embodied is useful in understanding the mismatches that can occur between the operations and the technology (see table 2). The more unpredictable and ill-understood is a production nln a* 1c ;t as i hub b&L to an %?f h=m ctxb =v.-, Process, +ha tdkd through human observation and judgment. The more predictable, proceduralized and understood the Pi-S, he more “scientific” it is, and the more amenable to control by technologies with a specialized and hence limited, repertoire of responses. A pejgct match between technology and the production process is not desirable. Usually techn010gies are introduced into processes to increase the quality Of Output or increase efficiency,
Table 2 Stages 0: knowledge a Recognition of prototype (e.g., what is a good product). Recognition of attributes within prototypes; (i.e., ability to define some conditions under which process gives good output). Discrimination among attributes (those that are important; recognition of patterns. Experts may differ about relevance of patterns; apprenticeship is common). Measurement of attributes (some key attributes; measures may be qualitative and re!ativej.. Local control of attributes (repeatable performance; process designed by expert, but technicians can performj. Recognition and discrimination of contingencies; production process can be mechanized and monitored manually. Control of contingencies; process can be automated. Complete procedural knowledge and control of contingencies. (Process is completely understood.) a Jaikumar and Bohn [29]; slightiy adapted.
and this improvement depends upon introducing more systematic order into the process (moving it up to higher stages of knowledge). Therefore, the transfer of new technologies into production always implies some degree of beneficial misalign-em* ba+wraam 4cn\r “a ef he tprknnl/ruv and the lAlblAC IJbcwbkAA*ha uaw a1.p.u &SAY .was.u”r”pJ UII w-v operations. However, introducing a small island of scientific rigor into a craft-dominated production process can appear far from beneficial to advocates of the old technology. In fact, improvements wrung from the old process in response to the challenge to a new technology can result in what Foster [22] has termed the “sailing&ip phenomenon,” in reference to the initial superiority of improved clipper ships over the rival technology, steamboats, when the latter were first introduced. Old technologies subject to this phenomenon outperform the innovation at first, thus making further advances in the stages of knowledge seem unnecessary and undesirable. In the case of Solagen, the operators in the ---_-z-_ gela?in plant were critical of the very prwstz calibration required for the new chemical reaction process, compared with the old decomposition process. With Solagen, their production window for removing the ossein shrank from four or five days to a few hours, even minutes. “If you don’t hit that baby just right,” explained a foreman, “ you’re in Zrouble.” And another long-time expert explained that sticking a six-foot pole down into a pit to determine if a residue of undecomposed
D. Leonard-Barton / Implementation as mutual aabptation
bones remained at the bottom might seem primitive, but it was more rxrtain than relyiig ‘upon figures on a dial. The leap from art to science in this case seemed more dangerous than any potential benefits could justify. 3.2. Delivery system misalignments The technical characteristics with which an innovation was endowed by its inventors are augmented (either positiveiy or negatively) by the technical attributes of the system (hardware, software or educational programs) through which it is delivered to the users. For instance, the performance of a new pump for molten metal is affected by the size of the furnace it is put in; a computer program’s speed and potential for useroriginated change is influenced by the language in which it is written; the transferability of a new production process depends upon the skills of the individual teaching it. Although the resulting misalignment between technology and its delivery system seems quite obvious3 the ratifications often are not. For instance, users frequently find it difficult to know how many of the innovation characteristics they experience are inherent in the technology rather than in the delivery system. The general impression created by the equipment or training through which a technology is delivered can have a kind of “halo” effect on judgments about a new technology’s accuracy or reliability. As psychologists and market researchers have shown, a general, global evaluation of a person or product can alter one’s perceptions of individual specific charatiteristics [481. Therefore, a misalignment between the technol--~-*athrcxq$ which it is delivered ogy and the ~yDb~~aa to users is not the simple technical issue it masquerades as. Inadequacies for which the devel--.-W-m~GllGl RdaSfi”rbll., hn*rn ** rl;mra+ meA*~&L;l;+rr L,.t UUL upma amy llQVb IfiU UIIbkb I~JpVAIBAuAubJ that users may blame on the technology can lead users to reject, underuse or even sabotage the innovation. In the cases studied, an inadequate delivery system surfaced time and again as a seriously underestimated and hence undermanaged I&alignment. In the case of the new polymer composition (table l:A), the laboratory process involved a two-step chemical seaction requiring two
257
separate sets of reactor equipment. However, the ~1nuLIenginee_rs, vichku tc ZX&~ ear&d WWF._ nmt19v V’UWJ, ----Y attempted to redesign the process to fit their onereactor equipment. The first production runs were not successful. The composition solidified prematwely, bringing the production line to a halt as the equipment was choked by solids. Similarly, in several cases, inadequate computer hardware slowed software performances to a snail’s pace. XSEL (table 1 :Gj, is an “expert SELling a&stant” that aids Digital Equipment Corporation’s salespeopie in designing the customized and highly complex computer configurations needed by customers, before these orders are set to manufacturing, where they are checked by XCON (table 1:B). XSEL was rejected by many intended users because the response time was so slow. Users associated slowness with the software more than the hardware (computer and communications links) through which they received it; thus the misalignment between the technology as designed and the delivery system through which it reached the sales personnei was a critical deterrent to XSEL use (see [39]). In the case of the MRPII system studio; inadequate !~ tn a ~c)rn-__- h~~~w~i~ --__ plete plant shutdown at one site as the critical information linkages lagged too far behind the response deadlines to be useful and people returned to their manual systems in the belief that the software system logic was at fault (table 1 :D). 3.3. Value misalignments: Technology with performance criteria Technology disrupts and reforms the organi=tional fabric, often in fairly unpredictable and situation-specific ways [3,12]. That is, organizational structure interacts organically with technology . Therefore, managers cannot foresee all the specific changes that may need to occur in the .nE** *~.r&*~rn*~+ ;m *n Ub%,C C” nv=ln~+ Wrry‘“lruQ *em, 1l-w UJWfiw*1.‘1l “I&&L‘WIBC 11fi*r&W VIUWLknr** te&.nology. TWO generic dimensions of performance criteria predictably int;eract with te&iol- _ ogy to produce misalignments: (1 j the significance to job performance criteria of the activity or task that the technology impacts; and (2) the nature of that impact, i.e., whether the net impact is more negative or positive. Performance criteria at numerous levels in the
258
D. Leonard-Barton / Implementation as mutual adaptation
MANAGEMENT SIGNIFICANCE OF ACiiiiiTY TO JOB PERFORMANCE
NEGATIVE
POSITIVE
EXPECTED NET IMPACT OF TECHNOLOGY ON ACTIVITY Fii
2.
Potentialmisalignments betweentechnologyand user perFormawerewards.
user organization may interact with the tec~hnoiogy: -
individual operators; the operations management; the business unit management; corporate in;_
Lack of interest in new technology, which developers freqttently interpret as an emotional “Not Invented Here” response, may more accurately represent “an arguably logical Not Interested Here”, response [45, p. 1541, because the innovation’s potential or actual usefulness is not clear to any or all of these four user levels. These four levels have often been lumped together as if there were one common user’s perspective. In realitt, there are at least four differing perspectives and hence potentially four different sets of performance criteria. Moreover there are often discrepancies among the four reward systems that can substantially affect a technology transfer effort. 3.3.1. Significance Each set of users evaluates the technology first in terms of the significance the activities being altered have to job performance criteria. The tech-
A_?_ noiogy is ‘highly significant to performance criterrd if it impacts one or more core activities by which an operation’s or person’s success is judged, i.e., one or more “critical success factors” [54]. If the activities affected are peripheral, significance is low (see fig. 2). 3.3.2. Impact on activities The second dimension affecting a technology’s value is its expected net impact on the activity being altered, positive or negative. Impacts include effects on the profitability of the activity, but of course financial measures are not the only ones to be considered. Technologies can “cost” in terms of lost time, decreased status, or unpleasant routine, and they can “benefit” in terms of increased skills, better quality output, and so forth. Imnart chnnlA he ratin nf nncitive - tn r9.w. “*-Y-a-_ Y_ rnnderd ~___“__~_~_ R - _--_ __ r_‘---. _negative effects rather than an absolute, because all technologies have both positive and negative’ effects: Technology assessment on the two dimensions - significance and impact - provides a perceived benefit/cost ratio. That is, each evaluation of the expected net consequences of using a technology LL A -_rr*:--.X-c\ ( the expected 11~~hhpraSei 1s we&h&d bjr the sttibjective significance that the expected consequence l
D. Leonard-Barton / Implementation as mutual rldaptation
will have for the reward system under which the users operate. 5 The two dimensions of performance rewards, significance and impact, interact to produce four possible extreme conditions of misalignment, ranging from very minor to severe as illustrated clockwise in fig. 2:
I. II. III. IV.
High significance, Low significance, Low significance, pzgh sign&ance,
positive impact; positive impact; negative impact; negative iwpa&
’
Of course this description oversimplifies; many more subtle combinations of the two dimensions are possible. However, useful distinctions can be drawn among the quadrants. Quadrants I and iI, in which impacts are positive, both represent relatively __1, ~l?ror misalignments (assuming p2ifWt alignment is impossible), although the forms of adaptation used to address those misalignments will likely differ between the two quadrants because of the difference in significance. The technologies characterized by Quadrant III (low significance and negative impact) are unlikely to be tbra ;r too rrnim, transferred at all La#W+..O~ UNUUUW CIlW+Bck &LLuR au UIIIIII portant to warrant expending effort to overcome the misalignment. Technologies that are reasonably well aligned with the performance criteria applied at a business unit or corporate level can still be severely misaligned at the individual user’s level. This circumstance is particularly likely in the case of productivity tools introduced for the good of the overall production rate but that individuals have little incentive to use, or of technologies attached to production tasks that are much more significant to upper management than to individual operators. Several of the innovations studied had a beneficial effect on the performance criteria applied to upper managemen.t, but a deleterious impact on individuals’ jobs. For instance, in the case of a computei-aided engineering tool (CAE tabJLe1 :F), the circuit de-
The direction of performance is implicit. That is, it is assumed that if an activity is critical to job performance, then better performance in that activity is desired. Thus, if an operation is judged on the quality of output, producing high quality is the critical activity. It is also assumed that each. business, operation, and person is judged by performance on more than one activity.
259
signers were not interested in capturing their *“II”-UU”” on an e!ectronic system They wae CFh~mQ*;PC accustomed to handing over draft sketches to draftspeople for translation into precise schematics. Therefore the task was a relative nonsignificant one, and the technology impacted it negatively by requiring greater precision md attention to detail from the engineers [a]. In the SSA implementation study, the software enginee= believed that the task of scoping out the software program was a significant part of their job - but many of them felt using SSA impacted the task negatively, because it was such a “time sink” at the beginning of the development process [37]. On average, the 145 programmer_/ analy its interv.iewd about SSA believed fi&& stiperT-&G judged their performance chiefly by whether or not they were on time with their projects. Moreover, technically-oriented analysts believed that “system development methodologies inhibit my creativity” and tended to agree with the statement, “I do not need to use structured application development techniques.” Thus for individual users, the attributes of SSA conflicted with job performance criteria - both those they perceived management used to judge their work and those many of them used personally. This conflict may partially explain why, almost three years after its release to the field and despite corporate headquarters’ requirement to use it for all software development, SSA was still used by only 60 percent of the analysts for whose jobs the technology was highiy relevant.
4. Adaptive cycles - large to small Misalignments cycles 6 6 ~~~~
~=“_ L” in IICI~IV as a_ return to exactly the .Su-, Apf&pA same starting point, the term cycle is used in a metaphorical rather than a strictly orthodox sense in this paper. The Cycles described in this paper never return the technolo~ or the user environment to exactly the same situation from which the cycle of adaptation started. Time, external events, and
learning effects - to name but a few infIuences - have altered the original situation, i.e., the decision context, by the time the decision point is revisited. As Eg. 1 SU@$S& the adaptation process is more accurately lx%traYedas 100~~ h a sgird, but fie te&&ogy
proved to be cumbersome.
~!%xhted
vd.!i *kit ~GEFjd
D. Leonard-Barton / Implementation as mutualahptation
260
TECHNGLiiG’t ADAPTATION CYCLES:
ORGANIZATIONAL ADAPTATION CYCLES: Levels of Performance Criteria
Development of Technology
Overall System
Corporate
Idea Generation
Delivery Systems
\ LARGE CYCLE
z Problem Solving
\
Line of Business
\ Concept Definition
\
Business Function (e.g., Manufacturing)
’
Test of Feasibility Plant or Department
I
z Laboratory Prototype
f PiiOi PiOChibi2 Prototype
f+oduction Prototype
/
Task
SMALL CYCLE
Production-Ready Technology
‘Technology
Interaction ’
\ Status Quo
Fii 3. Large and small cycles of redefinition.
because the process is one of circling back to revisit a decision point - reopening issues of technical design that the developers assumed were resolved, redesigning delivery systems in the user environment or %nfreezing’ 1431 organizational routine to reexamine the goals implied by current performance criteria [35]. These adaptive cycles vary in magnitude, depending upon how fundamental is the change to be made. In the case of technology adaptation, a large cycle would mean that the developers return to the drawing boards, whereas a small cycle would entail a shift very low it: the “design hierarchy” [14], that is, a minor adaptation such as a new module of software code or a different nose cone piece on an electronic PumP We fig 3)Similarly, z&rge cycie in the delivery system would entail a much more sizeable outlay of capital for major equipment than would a small cycle. Finally, a small adaptive cycle in performance criteria might entail the redesign of a particular task or individual role whereas a large cycle implies a strategic shift for the plant, if not for the business unit or even for the corporation, since it means rethinking the critical success factors [54] on which the organizational unit is judged. Tech-
nologies can foster such organizational changes by opening avenues for competition not previous recognized or by forcing entirely new routines as part of a desired strategic shift. Such large-cycle adaptations are always very costly, of course, in human as well as monetary terms. Developers whose technical designs are rejected may leave the firm [53]; current accounting practices make it difficult to cost-justify new technologies in comparison to current depreciated plants [33] and large-scale shifts in performance criteria can deskill large numbers of workers [ll]. In the cases studied, there were relatively few examples of adaptive cycles falling at the extreme “iarge” end of the spectrum because these imply massive and, one would expect, relatively infrequent change. ‘Moreover, two of the 12 projects studied were cancelled as failures precisely because their continuance implied large-cycle adaptation that the organization rejected as uneconomical or infeasible. A brief look at one of these cancelled projects compared with two others that entailed large-cycle adaptation is instructive. In the previously described Solagen project, the new chemical reaction being substituted for part of the gelatin-making
D. Leonard-Barton / Implementation CLT mutual adaptation
261
process achieved only parity with the old system’s technical performance, and implementation would have necessitated a large cycle of ?.taptation in performance criteria (plant-level shift). Output from the current technology was not judged by its consistency, batch to batch, because some proportion of the successive “cooks” always produced ossein of adequate “yellow density.” Nor was process control valued for its own sake, because the needed type and quality level of gelatin would be mixed from the barrels of dried ossein repre-
during the abortive attempts, into the final, successful “expert system.” Thus, although the developers did go “back to the drawing bo&s” to select a new tedmobgy, the problem they sought to solve was much better understood and struttured because of the failed first attempt. Implementation of the mammoth Materials Resource Planning software package (MRPII) i& lustrates large-scale adaptation of performance criteria. For each of the plants into which it was introduced, the software required a major over-
senting
Lkn..i llQUl
ouigut
from
a i&&y
of
bg@s3
ail
“cooks.” Because building a new plant to support Solagen would have been very expensive, and current capacity was adequate, the Solagen project == ‘; .Tu1upycu. 1=----a 1LGz3b -I- rt,k_m:r,9*IPI Wdb &ii&i*-..A LWG__A__.3 y”Lu” l+*aUbC(lrl, IIVVV~WCII, some top managers questioned whether the decision had been made on “too narrow financial criteria,” because a key factor in competition with Japanese manufacturers appeared to be control over production processes - that is, the ability to progress through the “stages of knowledge.” In . . C~% ~~g~_~~~~~ ~~~~~~~~~~~~ ides sf Tut;
i;ei~~i_
mance criteria and infrastructure might have been salutary5 despite their cost. An example of a large cycle of technical adaptation is Digital Pquipment Corporation’s struggle with the problem of checking the accuracy of manufacturing orders for their complex computer systems. Looking ahead to a probable exponential growth in the configuration task’s complexity (because of an ever-increasing number of component options and therefore of possible combinations to create systems), the managers foresaw no way their technical editors could continue to check the order flow manually. Yet the first attempts to automate the configuration checking task through a traditiunal software program failed. Only when Digital cycled back to select a completely different technology - a form of artificial intelligence - was the solution adequate [41]. The resulting “expert crr=tPm” ..YY namwl .*uyII XPfiN am-__- leXnt=rt \“_~“” CONfieurt&), _ -- --G o>O.“... ~UQC Such large cycles of technical adaptation are beneficial if the abortive attempt yields learning that could not have been captured otherwise, that is, if the organization “fails forward.” This is the essence of “learning by doing” [19]. In the XCON project, the automatic checker of manufacturing order configurations could probably not have been constructed had not a kno*&&zeable engirreer ---- --- -o-spent much time embed&rig what he had learned
G A32
rewd
sysRms
-
&nm &au-
ni71nC p&us&
snancscrJps-yaws
Ofi _”
the most junior purchasing agent. This adaptation helped the plants move to high-performance “Class A” manufacturing status (as judged by mnt;nan 1A~iri*r~
DrnArPtifiql i Aij-ir.lc-iru
anri lnventtwv &e-d-_ a-_. -,iiiu;J
Cnntrnl L--_j_z_ i_
Society standards) by imposing discipline on all operations. Small-cycle adaptations, being much iess costly, are common to all technology implementation situations, although of course, some technologies and some organizations are much more adaptable lhnm r\thn+c Thmrdnr~ & cnme PPCPC tawhniral CULUA “&=“I0. 1 aa~X”l”.“, “Y-I -“I” -w-e_changes predominate; in others organizational changes do. In substituting copper for silver in the production of hybrid circuit boards, for instance, a number of technical changes were necessary once the process was transferred into the production facility (table I :I). The laboratory had used a copper alloy paste that had e; Lellent conductance properties, but poor adhesion, especially sunder full production conditions. Manufacturing engineers altered the process by testing and then purchasing alloy from a vendor other than that recommended by the laboratories, as well as by slowing some steps in the process. In the case of the Rolling Mills Diagnostic (table 1: C) organizational changes dominated. The process engineers on the staff had neither the skills nor the time for the additional monitoring and analysis tasks required by the new technology. Therefore the plant used funds that had been made available as part of a quality improvement drive, to hire a young engineer, who was trained by the laboratory scientists to perform the necessary diagnosis. He in turn trained process tech& cians, whose performance criteria were revised to provide credit for gaining proficiency with the A~aann~tirc urU&y’” .a-” Not
all
cvctem “J”. -1.
adaptations, large or small, are benign.
1Techiblogy 1 Beneficial
Detrimental
Beneficial
Cancellation (No learning captured) [Solagen]
failure Forward (Learning captured) [XCON]
orid--:eck (Inconsistency in performance criteria or delivery systems; or severe mistiming of adaptation) (XSELat !??3]
s;;a;egic Stiff: (Change in performance criteria at high level; integration of delivery systems) [MRPII]
Truncation (LQSSof functionality) IPolymer-at first]
Adaptation (Creation of value) Foppet]
Accretion (Addltlonal performance criteria; overlay of systems in delivery system without redesign (J.R Circulatnrl
Detrimental
wle Cycle:
Small Cycle:
I
Fig. 4. Forms of adaptstion: benefiti
Just as Ikiiman
and M&au- -
I I I
Job/Task Redesign (Change In performance criteria; aiteration of delivery systems) (XSEL-eventualty]
and detrimental.
[fl
fetid
“&at
school districts adopted innovations without improving performan~ as a strategy for maintaini3g status quo, so in the corporations studied here, adaptations were sometimes barren exercises. As fig. 4 suggests, a large cycle of technical adaptation, i.e, returning to the technology selection or fundamental design choices without being able to transfer any knowledge from the abortive attempt seems a great waste. Although some managers felt that Kodak’s canceled Solagen project did provide some scientific insight into the nature of gelatin, there was relatively little attentpt to capture all that was learned. The person most knowledgeable about the project, a technician, left the company. Detrimental forms of the other kinds of large cycle adaptation are most likely when some of the performance criteria or parts of the delivery system are altered but others remain constant. Strategic change need not originate at the top of the organization (see [12]). However, if technological change brings the performance criteria of different orga..tional levels into conflict, a kind of organizational gridlock can occur, with subunits attempting to drive off in different, inconsistent dkctions~ Thus, for example, if a new flexible manufacturing technology allows a plant to corn-
p&g
On
“siz
by
of
picket
Vanes
~~~
8
“differentiation” strategy; see [51]), but corporate management &ll judges the plant according to unit cost (a “lowest cost producer” strategy), the attempt to redefine rewards for organizational performance at the plant level alone will result only in incorGstency with corporate goals and in disruption, with too little benefit to justify the new technology% cost. Regarding the delivery system, extensive investment is sometimes made in supporting hardware but not enough in training, or in training but not education. In the case study of MRPII, one of the three plants researched invested in the training courses offered by the software vendor so that everyone - from purchasing to receiving to inventory control - knew how to enter the data for which they were responsibie. ziowever, the basic philosophy underlying the precise controls required by ,R.IRPII was not conveyed, with the result that many people lost all sense of meaning in their jobs. With hindsight, it was clear that training did not equate with education and that people needed ‘know-why” as well as Ynowhow.” Incomplete investments in hardware can cause similar breakdowns. In the case of the Jumping
D. Leonard-Barton
/ Impkmentatiom as mutual adbptation
Ring Circulator a melting furnace was rebuilt to accommodate the technological innwation - ari electromechanical pump for s+&rringalumimum scrap down into the molten bath. However, there were not enough trucks to ferry the expected increase in output to the next station in the production process, even if the pump achieved the productivity goals set for it [38]. This pattern eliminating one physical delivery system bottleneck only to encounter another, further removed from the site of the technology introduction - was observed in a number of the cases. Small-scale detriiental adaptation may be less spectacularly obvious than large, but it can still delay or sabotage implementation. Technical . a&pth4tiGa
m2
f,UC &Acfieatd l
if
A.._---cl. cIl~I blIQll&Gb
truncate the technology’s functionality or limit its flexibility. For example, software technicians working on the introduction of the MRPII system discovered in one plant that by tinkering with the source code to customize the program for their use, they made installation of the vendor’s later mm_ -__ ______ versions almost impossillly diEdi. mey were compelled to go back to the program, strip out all their changes and house them instead in separate programs, that could be accessed by the MRPII =-,r=+--= without altering the fundamental operJ, 0bWU1 ations of the original software. Alterations to performance criteria.can be similarly detrimental if job or task rewards are not re-evaluated as a complete set when the technology is introduced, but a new layer of performance criteria is simply added on top of the old in a process of accretion. For example, one of the reasons for the ultimate failure of the electromechanical pump mentioned above was that the furnace operators, who were judged on how much aluminum they pushed through the furnace, had discovered short cuts to producing large amounts of molten aluminum in a short time. Incensed at the increased maintenance required by the new ma.-* tkasr ~~wPP+P.A nlA mcanncal rndbd yU’U&R, ...“. 1” .V1&_ tnC”tha GIL” W1Y LIIN_YU _I”.IIYYnf -1 stirring the metal with a rod attached to a forklift truck-tbat they drove back and forth - creating waves of molten metal that swamped the new pump. The swamping led in a vicious cycle to yet more severe maintenance problems. Similarly, in the case of the CAE system, a new performance criterion was added to the design engineers’ job: use of the system to generate the circuit board design schematic. “Shame sheets”
263
originating in the drafting department showed who had sent that department manually prepa.& +;-@ ;+nmA “I AC S~~~~muu2u~u lusauu
nln.f-+w~~;tvll.r~wvl,,~ ~*-u”a&&~, p-&a-
r\,.+ VU&‘;
CAE system [40]. This negative incentive system caused much resentment. put from the
5. Exampks of adaptation options One of the important distinctions between the framework proposed here and other models of the technology transfer process is the explicit recognition that the same misalignment may be addressed through adjustments to either the technology or the organization. Managerial options for addressing this
section.
5.1. Technology and stage of knowledge of production process Severe misalignments between *he technology and the stage of knowledge of A&eprd_UcAzon process, for instance, might seem to have been caused by an obvious miscalculation by the developers, requiring a return to concept definition, problem-solving or even idea generation. Such misalignments allayrtsuh when the developers are very ignorant of the operations they are attempting to upgrade. However, it is equally likely that such misalignments occur because the underlying operations are in a chaotic state, having grown erratically m a turbulent production atmosplzre in which there was little time for rational planning. In those cases, the corporation may benefit more by pausing to reorganize operations than by adapting the technology. The experience of one of the plants installing the MIWII system illustrates this point. The initial few weeks of attempted introduction so highlighted the misalignment that the system was temporarily pulled out to allow manrlfcartlwina ~I_.uIIYIuI.-AAA~
the .a*” nnnnrtranitv ‘yr-’ ‘--J
tn WY r~~rg&e *_v-
aA_
redesign their pri-cess. At the end of several months, Ml?PII was reintroduced - this time successfully. 5.2. Technology and targeted specifications Similarly, misalignment between the technology as it is when transferred and the technical specifications it was supposed to meet can also be ad-
D. hard-Barton
264
m
/ Implementation as mutual a&ptation
technical or organizational adadaptation Option is O-bviou~, although success is not always attainable: keep working until the specifications are reached, making the accompanying necessary investments in additional resources (including perhaps consulting with experts outside the original development team). ‘Ihe option of organizational adaptation is less obvious - change the specific&ions. This might ,em a highly undesirable move, suggesting lowering of standards. However, in fact the original specifications sometimes turn out to have been inappropriate or irrelevant because benefits other than those originally envisioned have resulted. For instance, rayon was supposed to be “synthetic silk” and was rejected at first because it did not replicate closely enough the properties of l &e material it was supplanting to be accepted 1451. Similarly, in implementing process controls in one corporation’s steel rolling mills (not one of the 12 cases researched here). the original specifications were not met. The software failed to automatically set the steel preshaping function adequately. However, quality of the output increased far more than anticipated. Management in central engineering insisted that the technology was unacceptable because it failed to meet speciflcations, but in the mill, managers redirected their expectations and in effect, adapted the specifications to exclude the automatic preshape function 191.When technology managers in these cases recognized that the specifications inhibited the very innovation sought, they altered the specifications, not the innovations. However, altering specifications can require changing an entire organization’s performance criteria, which would be a large cycle and hence likely a strategic change.
j~tments.
by &her The
t&nical
5.3. Technologyand delivery system h+Iisalignments between technology and its delivery system in the user organization are more likely to result in small cycles of adaptation either investment in new equipment and reneged support systems or a redesign of the technology to match the existing resources, or both. For instance, in the case of XSEL, Digitts sales management was persuaded in 1986 to replace the original computers supporting the program in the field with very much more powerful models. The
leap in accessibility and response time that resulted from this change in the delivery system was experienced by users as a more significant performance improvement in the software program than previous very extensive alterations of the software itself. 5.4. Technology and performance rewards Misalignments in perforti;ance rewards similarly can be addressed throu& ei&cr :echnology or oqpization adaptation. hrlany technogies W in quadrant IV of figi 2 (high significance but negative net impact) or II (p&&e net impact but low significance). Several strategies for adaptat&. would move such technologies towards Quadrant I. The f’irst such adaptation strategy is obvious, if some**=& painful f-or dcvclopers to contemplate: alter the technology to raise benefits and/or to lower costs (including non-financial ones such as learning time, distance to facilities), or both, thus altering the impact so that it becomes net positive. The second adaptation strategy to address reward system misalignments is to alter the criteria by
wb&
j&
performii~f%&
is
judg&
far
one
ox
more user groups. This kind of adaptation alters the context in which impact is judged. That is, an impact that was previously considered significant loses relevance under the new criteria and a previously unheeded impact assumes new value. If this adaptation occurs at the managing operations level (e.g.) the manufacturing or marketing function) and involves a switch in the organization reward criteria, and hence in the critical success factors for that function, then the adaptation implies a switch in operating strategy and the cycle is large. The third strategy for adaptation, and one espe cially appropriate for Quadrant II technologies, is to alter neither technology nor organization, but instead to alter the visibility of certain positive impacts to connect the technology to an already significant factor affecting job performance. That is, managers can move the technology from Quadrant II to I by association. In several of the plants implementing the new Purchase and MRPII systems (table 1:D and E), the new software introduction was successfully tied to a drive for manu facturing excellence or for quality. In contrast, the developers of Smartwave (table 1 :I) missed an opportunity to associate their innovation with a drive for quality in the plant and the program was
D. Leonard-Barton / Implementation as mutual a&put&
seen as an unimportant innovation disrupting op‘A= 2s 9 contribution towards . better quality [63]. eia*2Gns ia+&ei
6. Condusi~ and managerial implications The major point in this paper is that implementation is innovation. Theorists and practitioners have done themselves and other a disservice by conceptually separating the implementation of new technologies from their creation, as if the transfer to operations required merely fulfiiing the original charter. Technology transfer rer*+~< -.L . 3 IQ xgued here, contin=~+~s. ongoing dedid;&iT .:5-t9-b . the process of cimxg2 mci tile ~~~~&cms f4.. _I_,:~-5ent
of mutual
a&@g&cq
k SJse
&e
Gil never exuctliy fit the user environment. There is always a need for a carefully managed “beta site,” i.e., experimental introduction into the user environment with the intent to learn. This is not to say, of course, that developers should not attempt to understand and simulate *fsa **cm* &fi&a&w UUYI~lrvis~~~~nt “II.AAYIIIlI~__.The _-_ better -i--4- the &&i nition process, the less disruptive and costly the adaptation cycles. There are many mechanisms for anticipating and forestalling misalignments, e.g., advisory user design groups or user membership on the technology design team. However, no matter how well conducted the design process, the same conscious management of the creative process is required in the transfer as was necessitated by the original development. The range of managerial options for achieving successful technology transfer includes changes in the user environment as well as in the technology itself and frequently the same misalignment can be addressed either way. Recognition of this fact requires that developers acknowledge some responsibility for identifying adaptation options even after they have, at least in their option, brought the technology to an acceptable stage of development and, on the other hand, that users share some of the uncertainty and risk that new technologies involve. A major proposition implied by this framework is that change in both technology and user environment is more beneficial than holding one constant and changing the other. Large-cycle adaptations can be pathologies or opportunities. The expense of having a technology Rdsm.&a~
265
totally rejected after reaching the point of introduction into user operations, necessitating a return to the drawing boards, is one few organizations care to bear. And the prospect of altering performance criteria from top to bottom in an organixation, in order to exploit a new technology is equally daunting. Yet research on survival in highly competitive industries suggests that the surviving companies are those that are open to advances in process technology - even if the price of that openness is expensive technical experimentation and ~&y
orgao&
s&f&&
[zs].
T%aoa&fi1UGlMUL~)
conscious management of a large cycle of adaptation is in fact strategy formuiation and strategic change: strategy can thus be driven from the ,? 1_Sl_ r&A&er IIIILIUI~ a&+-& sGi.dpj edACL% of mn +cF iaulu siar organization. As recently noted by a m:ufacE ing expert: Changes in equipment and process are powerful engines of change.. . There are few more effective ways of loosening up old ways of organizing -_production. [61] techno!ogies
Small-cycle adaptations can be similarly viewed ~ Pi~~i~~S (?r O??O~U~ti~. when technOlO_~ transfer is regarded as “problem solving involving both source and user in mutually reinforcing collaboration” [46], then the process is one of negotiating toward mutual benefit. Negotiation, researchers have found, is more likely to succeed if reasonable goals are set [4] and if the negotiators frame the issue in terms of expected gains instead of expected costs [5]. Thus the success of technology transfer depends not only on the technology but aI50 on the degree to which both developers and users want to make the transfer succeed. The will to make it succeed is more likely to be present if both sides of the transfer start with the premise that they are co-creating change - change that ti benefit both sides.
Referetwes William Abernathy, The Productiuiry Dilemma (Johns Hopkins Press, Baltimore, 1978). Thomas J. _Al!en, _~~~c?,rrin,p the Flow of Technology: Technology Tra_nsfer and the Dissemination of Technological tnfarmation WithirE the R&D Orggznization (MIT w, Cambridge MA, 1977). Steven Barley, Tech~~ologyas an Occasion for Stmcttig: Evidence from observations of CT Scanners and the SO-
D. Leonard-Barton / Implementation as mutual adaptation
266
& od.3 of Radiology Departments, Administrative Science Quarter& 31 (1986j ‘78-108. [4] Ma H. Barman and Roy J. Lewicki, fvegotiating in Organizations (Sage Pubhcations, London, 1983). [5] Max II. Baaerman,Thomas Magliozzi and Margaret Nede, Integrative &@ning in a Competitive Market, Organizatio& Behavior and Human Decision Processes 35 (1985) 294-313. [6] Paul &man, Thinking about Programmedand Adaptive Implemmt&m: Matching Strategies to Situations, in* H. Ingram and D. Mann (edsj, II%yPolicies Succeed or Fail (Su8e, &verly XUs, CA, 1980). (71 Paul Berman and Milbray W. McLaughlin, An Expioratory Study of Scirooi District Adaptation (Ratid Corpora-
tion R-2010-NE, Santa Monica, CA, 1979). [8] Jan& M. Beyer and Harrison Trite, I,mpIe,menting Change: Alcoholism Policies in Work Organizations (The
Free Press, New York, 1978). [9] Boston Umversity Schooi oi Management, iize SzenaYtai Mill Control: Technology Transfer at Superior Steel (1986). [lo] F. Bradbury, Paul Jervis, Ron Johnston and Alan Pearson, rra@F Fmm&T iE &&cai &mge jSi$h& & Noordhoff, Alphen aan den Rijn, 1978). [ll] Harry Braverman, Labor and Monopoly Capital (Monthly Review Press, New York, 1974). [12] Robert A. Burgehnan and Leonard R Syres, Inside Cornnmtn I..*.“““..“*lnnru,ntinno (The Prmc New YY._I \‘^_ Frpp a a__ w-s”“, - -_-. V&c - ----, 19Rf3 -- --,’
(13) AIok K. Chakrabarti, Some Concepts of Technology Transfer: Adoption of Innovations in Organizational BI-----u-r ?-I\3, /3\ \’ /‘9?3) ??;_?2() Conrex&w&B r..cutqjsr.‘s~“ [14] Kim B. Clark, The Interaction of Design Hierarchies and Market Concepts in Technological Evolution, Research Policy 14 (5) (1985) 235-251. (15) Hirsch Cohen, Seymour KeIIerand Donald Streeter, The Transfer of Technology from Research to Development, Research Management (May 1979) 11-17. [16] Fariborz Damanpour and Wilham Evan, Organizational Innovation and Performance: The Probiem of ‘Organizational Lag’, Administrative Science Quarteriy 29 (1984) 392-409. [17] Brian DeLacey and Dorothy Leonard-Barton, Solagen: Process Improvement in the Manufacture of Gelatin, HarvardBusiness School Case Services # 9-687-020 (1986). (181 Robert Drazin and Andrew Van de Ven, Alternative Forms of Fit in Contingency Theory, Administrative Science QuarterIy 30 (1988) 514-539. (191 John M. Dutton and Annie Thomas, Relating Technological Change and Learning by Domg, unpublished paper, New York University Graduate School of Business Adminiatrafi_nn
(Judy
1_98Jji
Charge qf Manufacturing (JosseyBass, San Francisco, 1988). (211 John E. Ettlie, William P. Bridgesand Robert D. O’Keefe. Organization Strategy and Structural Differences for RadicaIVersus Incremental Innovation, Management Science 39 (6) (June 1984j 682-695. [22] Richard Foster, A CaB for Vision in Managing ‘I’e&nology, Business Week (24 May 1982) 24-33. 1231 Lc& W. Fzy and Deborah A. Smith, Congruence, Con‘hgencj, and Theorf BtuiIding,Academy of Management Review 12 (1) (1987) 117-132.
1201 John
E
Ettlie,
T’ing
1241 DonaId Gerwin, Relationships Between Structure and Technology, in: Paul C. Nystrom and William Starbuck (edsj, Handbook of Organizational Design Vol. 2 (Oxford University Press, Oxford, UK, 1981). [25] Barney J. Glaser and Anselm L. Strauss, The Discovery of Grounded Theory (Aldine: Chicago, 1967) (261 JeraId Hage, Responding to Technological and Competitive Changes: Organization and Industry Factors, in: Donald ‘Davis (ed.j, Managing Technotogicai Innovation (Jossey Bass, San Francisco, 1987). [U] Robert H. Hayes, and Kim B. Clark, Exploring the Sources of Productivity Differences at the Factory Level, in: Clark et al. (edsj, The Uneasy Alliance (Harvard Business School Press, Rmslrch Colioquium, 1985j. (281 Blake Ives and Margrethe H. Olson, User Involvement and MIS SIIFFPCSA Review RecmarCth --a, _MgHqe_ment W-WY . .m _._.a_.. of am_“_Science 30 (5) (1984) 586-603. [29] Ramchandran Jaikumar and Roger E. Bohn, The Development of Int&igent Systems for Indust*&I ‘u’se;A Con_ ceptual Framework, Research on Technological Innovation, Management and Policy 3 (1986) 169-211. [331 P. Jcrvis, Innovation and Techno!ogy Transfer - A Note on the Findings of Project SAPPHO, in: Bradbury et al.
WA. [31] Bonnie Johnson and Ronald Rice, Managing Organizational Innovation: The Evolution from Word Processing to Of&-e h$?wmation S’stems (Columbia University Press.
New York, 1987). [32] Rosabeth M. Kanter, when a Thousand Flowers Bloom: Ctr.rrtrrrnl nnA WY1.W. Cn&nt Pnditinnr fnr -._*..._____._ Innnnntinn “.,UC _“. ~fiIl00tirr0 vv..rr..vr, U,.W _V.._ I........ ,..’ in Organizations, working paper No. 87-018 (October)
(Harvard Business School, Boston, 1986). [33] Robert S. Kaplan, Must CIM be Justified by Faith Alone?, Harvard Business Review (March-April 1986) 87-95. [34] Ralph Katz, The Effects of Group Longevity on Project Communication and Performance, Administrative Science Quarterly 27 (1982) 81-104. j35) Steven Kerr, On the FoBy of Rewarding A, WhiIe Hoping for B, The Academy of Management Journal (December 1975j 769-783. [36] Paul Lawrence and Davis Dyer, Renewing American Industry (Free Press, New York, 1983). [37] Dorothy Leonard-Barton, Implementing Changes in Production Methods: The Case of Structured Computer Software Methodologies, working paper Series #g-784-008 (Harvard Business School, Boston, MA, 1984). [38] Dorothy Leonard-Barton, New Technology at World Aluminum Corporation: The Jumping Ring Circulator, Harvard Business School Case Services, #O-687-050 (1987). [39] Dorothy Leonard-Barton, The Case for Integrative Innovation: An Expert System at Digital, Sloan Management Review 29 (1) (1987) 7-19. [4O] Dorothy Leonard-Barton, A New CAE System for Shield Electronics Engineers, Harvard Business School Case Services #g-687-081 (1987j. [41] Dorothy Leonard-Barton and B&n DeLacey, Skunkworks at Digital Equipment Corporation: The Tale of XCON, Harvard Business School Case Services #g-687051 (1987). [42] Dorothy Leonard-Barton, Implementation Characteristics
D. Leonard-Barton / Implementation as mutt& adrrptation
of Organizational Innovation: Limits and Opportunities for Managerial Strategies, Journai of c’ommunication Research 15 (y) (i988j. [43] K. Lewin and P. Grabbe, Principks of Re-education, in: W.G. Bennis, K.D. Benne and R Chin (Eds), The Planning of Change: Readings in the Applied Behavioral Sciences (Holt, Rinehart and Winston, New York, 1962). [44] Edwin Mansfield et al., The Production and Application of New Industrial Technofogy (Norton, New York, 1977).
[45] M.C. McCarthy, Barriers to Innovation, in: Bradbury et al [lo]. [46] C.S. Morphet, Speculations on a Recipient, Need Oriented -4pproach to TechnoIogy Transfer, in: Bradbury et a!. [lo]. [47] AmaI Nag, Auto Makers Discover ‘Factory of the Future’ is Headache Just Now, Wall Street Journal 207 (93) (14 May 1986) 1. [48] Richard E. Nisbett and Timothy DeCamp Wilson, The Halo Effect: Evidence for Unconscious Alteration of Judgments, Journal of Personality & Social Psychology 35 (4) (1977) 250-256. [49] William A. Pasmore and John J. Sherwood (eds), Sociotechnical Systems: A Sourcebook (University Associates Inc. LaJoIIa,CA, 1978). (501 Donald Pelt and Fred Munson, Originahty Level and the Innovating Process in Organizations, Human Systems Management 3 (1982) 173-187. [Sl] Michael Porter, Competitive Advantage (The Free Press, New York, 1985). [52] Ronald E. Rice and Everett Rogers, Reinvention in the Innovation Process, Knowledge 1 (4) (1980) 499-514. [53] Edward B. Roberts, A Basic Study of Innovators: How to Keep and Capitalize on Their Talents, Research Management 11(4) (1968) 249-265. [54] John F. Rockart, Chief Executives Define Their Chvn Data Needs, Harvard Business Review 57 (2) (March-April
1979) 81-93. [55] Everett M. Rogers, Diffuon of Innovations (Third Edition) (The Free Press, New York, 1982). [56] Richard S. Rosenbloom, Technological Innovations in Firms and Industries: An Assessment of the State of the Art, in: Patrick Kelly and Melvin Kranzberg (eds),
267
Technciogical Dutovation: A Crittca! Review of Current Knowiedge (San Francisco “I%ess. San Francisco, i978).
j57j Roy RotIu&,
Some Pr&em
of T&&&o@ Tiansfei
into Industry: Exampies from the TextiIe Machinery ~ectar, IEEE Transactions on Engineering Management EM25 (1) (February 1977) 15-20. 1581 Denise M. Rousseau and Robert A. Cooke, TechnoIogy and Structure, The Concrete, Abstract, and Activity Systems of Organizations, Joumai of Management 10 (3) (1984) 345-361. [59) AI. Rubenstein et aI., Effectiveness of the Interface Between R&D and Production. Final Report to Productivity Research Program, Engineering Division, of the NationaI Science Foundation (Grant MAE-8207373) (1985). [60] M.A. Saren, A Classification and Review of ModeIs of the Intra-firm Innovation Process, R d D Management 14 (1 j (1984) 11-24. [61) Wickham Skinner, The Productivity Paradox, Harvard Business Review 86 (4) (JuIv-Atqust 1986) 55-59. [62] William E. Souder, Managing New Product Innovation (Lexington Books, Lexington, MA, 1987). [63] John SviokIa and Dorothy Leonard-Barton+Smartwave B: Implementing An Expert System at Digital Equipment Corporation, IIarvzrd Business School Case #9-187-063 (1986). [64] Andrew Van de Ven, Cent&I Problems in the Management of Innovation, Management Science 32 (5) 590-607. ____ _ 1631mc Von HippeL The Dominant Roie of Users in the Scientific Instrument Innovation Process, Research Policy (5) (1976) 212-239 [66] Robert K. Yin, Changing Urban Bureaucracies (Lexington Books, Lexingon, MA, 1979). (671 Robert K. Yin, Case Study Research: Design and Methods. Applied Social Research Methods Series Vol. 5 (Sage Publications Beverly I-BIB,CA, 1984). [68] Gerald Zahman, Robert B. Duncan and JOMY Holbeli Innovations and Organizations (Wiley, New York, (1973). [69] Shoshana Zuboff, New Worlds of Computer-mediated Work, Harvard Business Review (September-October 1982) 142-152.
