Cp

AI

a

ARRAA

KCFSOM

1

adocDytehmeestahlaat

0d

Journal of Operations Management 29 (2011) 65–77

Contents lists available at ScienceDirect

Journal of Operations Management

journa l homepage: www.e lsev ier .com/ locate / jom

ontingency theory of capacity planning: The link between process types andlanning methods

ntti Tenhiälä ∗

E Business School, Calle de María de Molina 12-5, 28006 Madrid, Spain

r t i c l e i n f o

rticle history:eceived 21 March 2009eceived in revised form 23 May 2010ccepted 28 May 2010vailable online 9 June 2010

a b s t r a c t

Although the reliability of production plans is crucial for the performance of manufacturing organizations,most practitioners use considerably simpler planning methods than what is recommended in the opera-tions management literature. This article employs the contingency theory of organizations to explain thegap between the practice and the academic models of production planning. Arguments on the contin-gency effects of process complexity lead to a hypothesis that expects simple capacity planning methods

eywords:omplexityittrong inferencerganization theoryixed-methods research

to be most effective in certain production processes. A strong inference research setting is used to test thecontingency hypothesis against a conventional hypothesis that expects the most sophisticated planningtechniques to always be most effective. Multisource data from the machinery manufacturing industrysupport the contingency hypothesis and reject the universalistic hypothesis. The findings are explainedusing the concepts of task interdependence and bounded rationality. The results have several manage-rial implications, and they elaborate how classic concepts in organization theory can bring practically

tions

relevant insights to opera

. Introduction

In manufacturing organizations, many important decisionsre made in production-planning activities. Production plannersecide when and with what resources organizations produce theirutputs, and the methods that are used to create these plans arerucial to organizational performance (Kanet and Sridharan, 1998;avis and Mabert, 2000; Zwikael and Sadeh, 2007). Poor methodsield plans that are either too loose and result in excessive leadimes or too tight and result in failures to keep promised deliv-ry dates. Consequently, it is not surprising that planning methodsas always been a major research area in the operations manage-ent literature. Different planning techniques have been studied,

specially in analytical and simulation-based research (Kouvelist al., 2005). Those streams of research have produced variousophisticated algorithms that enable the leveling and optimiza-ion of production plans (e.g., Davis and Mabert, 2000; Yang etl., 2002; Deblaere et al., 2007). However, empirical researchersave repeatedly observed that most practitioners use considerably

ess sophisticated planning methods than what is discussed in thecademic literature (Melnyk et al., 1986; Wiers, 1997; McKay etl., 2002). Moreover, empirical evidence indicates that those prac-itioners using advanced planning methods are on average less

∗ Tel.: +34 91 568 9600.E-mail address: antti.tenhiala@ie.edu.

272-6963/$ – see front matter © 2010 Elsevier B.V. All rights reserved.oi:10.1016/j.jom.2010.05.003

management research.© 2010 Elsevier B.V. All rights reserved.

satisfied with their plans than those who use simpler and less accu-rate methods (Jonsson and Mattsson, 2003). The purpose of thisarticle is to explain the gap between the practice and the academicmodels of production planning.

This article employs the logic of strong inference, the con-tingency theory of organizations, and a multisource empiricaldataset to explain the determinants of different planning meth-ods’ effectiveness. The strong-inference logic refers to a researchdesign where theory building is based on tests of competinghypotheses (Platt, 1964). The contingency-theoretical perspectiveof process complexity (e.g., Thompson, 1967) is used to proposethat sometimes the most sophisticated planning methods maybe less effective than the simpler techniques. The contingencyproposition is tested against a hypothesis about the universal supe-riority of the most advanced planning methods. The analyses areconducted with survey data from a sample of machinery manufac-turers. The statistical results are complemented with interviewsthat shed light on the reasons why practitioners end up using cer-tain planning methods.

2. Conceptual framework and hypotheses

2.1. Underlying assumption: importance of planning in complexorganizations

Planning is necessary in all complex organizations where spe-cialized and interrelated resources perform a wide variety of

6 ions M

dacnnscauittw1

iTpai(m2touRam2ac

awcsotficaib1

6 A. Tenhiälä / Journal of Operat

ifferent activities. In the absence of planning, different parts ofn organization could pursue their own objectives, which mayonflict with those of others (March and Simon, 1958). However,ot all organizations are complex; consequently, elaborate plan-ing efforts are not always necessary. In simple settings, wherepecialization, action variety, and task interdependence are low,oordination can be achieved through rules and heuristics (Cyertnd March, 1963). The dichotomy between plans and rules is partic-larly evident in manufacturing, where one of the main paradigms

s characterized by its emphasis on time-phased planning, whereashe other emphasizes rules and simplicity. The central concept ofhe former is material requirements planning (MRP, Orlicky, 1975),hereas the latter is founded on just-in-time (JIT) methods (Ohno,

988).A classic way to pursue simplification in manufacturing is to

solate operations from external uncertainties (Thompson, 1967).he extent of the isolation depends greatly on the order penetrationoint (Olhager, 2003); the earlier the order-specific requirementsre taken into account, the higher the exposure to the environments. Therefore, time-phased planning is important in make-to-orderMTO) manufacturing, and the JIT methods are at their best in

ake-to-stock environments (Karmarkar, 1989; Vollmann et al.,005). Both approaches usually coexist in assemble-to-order sys-ems and other intermediate settings. The postponement of therder penetration point enables the use of JIT methods in thepstream operations of customized manufacturing (Olhager andudberg, 2002). However, the inherent complexity of producingccording to individual orders cannot be eliminated by forcing JITethods upon the MTO parts of the processes (Hopp and Spearman,

004). Consequently, the time-phased planning has remained asvital part of manufacturing management despite the important

ontributions of JIT.Contemporary methods of time-phased production planning

re based on the manufacturing resource planning (MRPII) frame-ork. It was originally developed to complement MRP with

apabilities to check material plans’ feasibility against capacity con-traints (Landvater and Gray, 1989). More advanced applicationsf MRPII have since been developed to allow the feasibility checkso be extended to other factors, such as shipping resources andnancial constraints (Yusuf and Little, 1998). However, the practi-

al implementations of such solutions have remained rare (McKaynd Wiers, 2004). In fact, it has been observed that even the capac-ty planning features of MRPII are far less utilized than what coulde expected on the basis of the academic literature (Halsall et al.,994; Kemppainen, 2007). As the material-planning parts of MRPII

Fig. 1. Alternative methods

anagement 29 (2011) 65–77

are well established (Vollmann et al., 2005), the observation impliesthat companies’ production planning practices can be measuredthrough the methods that they use in capacity planning.

Recent developments in enterprise software deliver a promiseof easily applicable capacity planning tools. While the enterpriseresource planning (ERP) systems are well suited for the simplercapacity checks (Wortmann et al., 1996), the advanced planningand scheduling (APS) systems promote the more sophisticatedmethods (Kreipl and Pinedo, 2004; Stadtler and Kilger, 2005). How-ever, companies’ diligence in applying their enterprise systems’features is known to vary considerably (e.g., Bendoly and Cotteleer,2008). Therefore, variance might also be found in the utilizationof the capacity planning features. This variance enables testingwhether complex organizations that do not put effort into planningsuffer from a lack of coordination (e.g., March and Simon, 1958;Zwikael and Sadeh, 2007). Consequently, the following hypothesisis presented as the underlying assumption of this study:

H1. Efforts in capacity planning are positively associated withperformance.

2.2. Universal effect: advantages of sophisticated planningmethods

It is reasonable to assume that not only the efforts in plan-ning but also the ways of planning matter. Fig. 1 presents themain methods of time-phased production planning according tothe framework of Vollmann et al. (2005). The practical relevanceof the framework is high because dominant ERP software vendorshave structured their production-planning modules in the samefashion (e.g., SAP, 2010a). In addition, most textbooks either referto it directly or provide illustrations that closely resemble it (e.g.,Hill, 2005; Slack et al., 2007; Stevenson, 2004).

The backbone of the planning process is in the material planningactivities, that is, master production scheduling (MPS), MRP, andthe input/output (I/O) control (Vollmann et al., 2005). The optionalactivities are on the side of capacity planning. In Fig. 1, they arenumbered in the order of sophistication. The figure shows thatthe amount of the required data records increases as the methodsbecome more sophisticated. The increase is cumulative because the

records do not fully substitute one another.

Non-systematic capacity planning refers to the inexplicit consid-eration of capacity constraints. At the level of master schedules,it means that planners use their personal experience to evaluatethe feasibility of plans (Proud, 2007). In MRP, the inexplicit capac-

in capacity planning.

ions M

ioaaclK

mcMcRo2oMftos

tCfttfsai1

tfulsaraae

cfimDnctjia

Iptvlrebncs

A. Tenhiälä / Journal of Operat

ty considerations are realized through the lead-time parametersf bills of materials. The processing lead times represent the aver-ges, whereas the variances around the averages are taken intoccount with safety lead times (Vollmann et al., 2005). In the I/Oontrol, priority scheduling rules can be used to level capacity uti-ization without formal planning activities (Green and Appel, 1981;emppainen, 2007).

Rough-cut capacity planning (RCCP) is the simplest systematicethod. It can be done with several techniques that all share the

ommon characteristic of aggregation (Wortmann et al., 1996).aterials are aggregated to end products or product groups and

apacities to production lines or resource groups (Proud, 2007).CCP simplifies planning by ignoring subassembly inventories,perations’ sequences, setups, and batch sizes (Vollmann et al.,005). However, it provides the planners with a systematic meansf supervising how the resource utilization accumulates during thePS activity. This is an advantage if master schedules are updated

requently, MPS items are numerous, or different MPS items loadhe same resources. In these situations, the non-systematic meth-ds are prone to human error and easily result in overloadedchedules.

Capacity requirements planning (CRP) provides a more detailedechnique for ascertaining the feasibility of material plans. TheRP calculations are done not only for the end products but also

or the subassemblies. In addition, routing data enable calculatinghe utilization of individual resources and the effects of opera-ions’ sequences, setups, and batch sizes. Therefore, CRP correctsor the simplifications of RCCP and helps to generate more reliablechedules. However, manual work continues to have a major roles human planners need to make the plans fit the capacity lim-ts. The revisions typically necessitate several iterations (Burcher,992; McKay and Wiers, 2004).

The next step from CRP is to automate the iterations of revisinghe plans. This can be done with finite-loading methods, which areeatured in APS systems (McKay and Wiers, 2004). The process ofsing them is typically the following: first, material plans are down-

oaded from an ERP system. Next, algorithms of the finite-loadingoftware are used to find a solution where capacity constraintsre satisfied with the fewest breaches of due dates. Finally, theevised plans are uploaded back to the ERP system where theyre executed (Stadtler and Kilger, 2005). The obvious benefit fromutomating the capacity leveling is a reduced chance of humanrror.

In addition to capacity leveling, the finite-loading algorithmsan be used to solve more complicated scheduling problems. Thenite-loading tools with optimization may be used, for example, toaximize throughput or to minimize setups or downtimes (e.g.,avis and Mabert, 2000). Such techniques require the most plan-ing parameters, and the accuracy of these parameters is absolutelyrucial to the quality of the plans. However, the maintenance ofhe parameters and the investments in the software may well beustified in some manufacturing environments, such as in capital-ntensive production systems (Kreipl and Pinedo, 2004; Stadtlernd Kilger, 2005).

The planning methods are by no means mutually exclusive.nstead, several methods can be used simultaneously for differenturposes (Meal, 1984). For example, plant managers can use RCCPo evaluate sales plans, master schedulers may use CRP to super-ise their processes, and production planners can perform the finiteoading of critical resources. A concept that brings clarity to this plu-ality is bottom-up re-planning (Fransoo and Wiers, 2008; Vollmann

t al., 2005). In this process, master schedules are updated on theasis of the lower-level planning activities. In a closed-loop plan-ing system, the master schedules are based on the finite loading ofritical resources (Kenat and Sridharan, 1998). In an intermediateolution, the master schedules are revised on the basis of CRP. Con-

anagement 29 (2011) 65–77 67

sequently, the main method of planning can be identified. This is themethod that determines the output to which the manufacturingfunction commits itself.

As all of the advanced planning methods aim to reduce errors inplanning, it can be proposed that they should have a positive effecton operational performance. Some studies have already impliedevidence of such an effect (Sheu and Wacker, 2001; Wacker andSheu, 2006). However, they have not included finite-loading tech-niques, which is a major shortcoming because substantial efforthas been put into their development (Kouvelis et al., 2005). Thedevelopment of progressive algorithms and software can be welljustified if there is evidence on the direct relationship between thesophistication of planning methods and performance. Therefore,the following hypothesis will be tested:

H2. The sophistication of capacity planning methods is positivelyassociated with performance.

2.3. Contingency effect: fit between planning methods andprocess types

Another perspective to different planning methods’ effective-ness is to assume that methods’ suitability would depend on thecontext of their usage. Preliminary support for such an argumentcan be found in the surveys of Jonsson and Mattsson (2002, 2003).They show that practitioners’ satisfaction with different planningtechniques depends on the type of their production processes:managers of job shops are content with RCCP, the most satisfiedusers of CRP work in batch-process plants, and finite-loading meth-ods are most popular in production lines.

The observations are aligned with the review of Sousa and Voss(2008), who found the process type to have an influence on theeffectiveness of many different operations management practices.In the context of planning, the influence of process types can beexplained with two classic contingency-theoretical constructs: therepetitiveness and complexity of the tasks that constitute the pro-cesses (Perrow, 1967; Woodward, 1965). Their combined effectsare as follows:

• RCCP fits with job shops because repetition is minimal in low-volume and high-variety production, and the data records of themore detailed methods are thus difficult to maintain. Moreover,the more detailed methods are not necessary because the com-plexity of the system is limited with general-purpose machineryand a widely skilled workforce (Blackstone and Cox, 2005; Hill,2007).

• CRP fits with batch processes because the more repetitive oper-ations make the maintenance of the data records worthwhile.Furthermore, information about the resource-specific workloadsis necessary because the resources are more specialized anddifferent products are routed through them in different ways(Jonsson and Mattsson, 2003; Wortmann et al., 1996).

• Finite-loading methods fit with batch processes where complex-ity is reduced with bottleneck control (Goldratt and Cox, 1984;Vollmann, 1986). Finite loading works in a batch process if a sta-tionary bottleneck can be identified and all other resources aresubordinated to its schedule. Otherwise, each finite loading ofone resource can make another resource a new bottleneck, andthe iterations of revising the plans may become endless.

• In production lines, complexity is low because all resources aresubordinated to the flow of the line. Therefore, the capacity of the

entire line can be planned as a single resource. Detailed planningis desirable because untimely changeovers are costly in assem-bly lines (Hayes and Wheelwright, 1979; Kreipl and Pinedo, 2004)and cause congestion in manufacturing cells (Venkatesan, 1990;Vandaele et al., 2008). In addition, the repetitiveness of the oper-

6 ions M

oApDsptbawctl1

arMed(pr

aimm

Hifpb

8 A. Tenhiälä / Journal of Operat

ations makes it easier to maintain the parameters of the mostsophisticated methods (Safizadeh and Ritzman, 1997; Stadtlerand Kilger, 2005).

The relationship between the process types and planning meth-ds can also be explained with the concept of task interdependence.ccording to contingency theory, the alternative types of interde-endence are pooled, sequential, and reciprocal (Thompson, 1967;onaldson, 2001). The pooled and the sequential processes are the

implest to coordinate, but they have very different implications forlanning (Barki and Pinsonneault, 2005). The processes with pooledasks are inherently flexible, and this is a capability that should note constrained with planning that is too stringent. A job shop isn archetype of pooled interdependence (Galbraith, 1973). Mean-hile, the sequential processes are suited for efficiency, which is a

apability that can be fostered with detailed planning. In manufac-uring environments, sequential relationships exist in productionines and around the bottlenecks of batch processes (Thompson,967; Woodward, 1965).

The most difficult processes to coordinate are those where tasksre reciprocally interdependent. This is because all actions by anyesource may affect multiple other resources (Galbraith, 1973;onahan and Smunt, 1999). Some specificity in planning is nec-

ssary to prevent undesirable cascade effects, but getting into theetails is difficult because the possible interactions are numerousTushman and Nadler, 1978). Therefore, a moderately sophisticatedlanning method such as CRP is the most suitable option for theeciprocal processes of batch shops (Reeves and Turner, 1972).

In summary, classic contingency-theoretical concepts producemeaningful fit proposition that challenges the hypothesis argu-

ng for the universal superiority of the most sophisticated planningethods. The proposition is illustrated in Fig. 2 and it can be for-ulated as follows:

3. Fitness between capacity planning methods and process typess positively associated with performance; that is, the highest per-ormance is achieved when RCCP is used in job shops, CRP in batchrocesses, and finite-loading methods in bottleneck-controlledatch processes and production lines.

Fig. 2. Link between planning m

anagement 29 (2011) 65–77

3. Methods and data

3.1. Research design

Because Hypothesis 3 challenges the universal performanceeffect of planning methods’ sophistication, it can be viewed as com-peting with Hypothesis 2. This situation calls for a strong inferenceresearch design, which is an inductive approach that builds the-ory based on testing competing hypotheses (Platt, 1964). Stronginference studies must be carefully designed so that the researchsettings do not favor any of the rival hypotheses (MacKenzie andHouse, 1978). Multiple data sources are also necessary: quantita-tive data enable the testing of the hypotheses, whereas qualitativedata provide the understanding that is needed in the developmentof theory (Jick, 1979; Gupta et al., 2006). Although the strong infer-ence research design was originally developed for experimentalstudies (e.g., Nadler et al., 2003), it has been successfully employedin non-experimental empirical research as well (e.g., Shaw et al.,2005).

The sample of this study had to be focused on manufac-turers whose products are sufficiently complex to necessitatetime-phased planning. The manufacturers of simple productswould have obscured the comparison of capacity planning meth-ods because they can rely solely on rate-based material planning(Vollmann et al., 2005). The preliminary screening of a sample,or theoretical sampling, is typical to inductive studies, althoughits disadvantage compared to random sampling is the inability tomake statistical generalizations (Eisenhardt, 1989). However, thetradeoff is aligned with the objective of this study, which is toelaborate the overall relationships between theoretical constructsand not to pursue definitive parameter estimates (see Ketokivi,2006).

When the topic of a study is such a technical issue as capacityplanning, the reliability of informants may create concerns. In thisstudy, it was imperative to ensure that all informants had personal

experience with the everyday routines of production planning.Plant managers and chief operations officers, who are typical infor-mants of large-sample surveys, may not have firsthand knowledgeof the techniques that are used in day-to-day production planning.Therefore, close relationships had to be established with the stud-

ethods and process types.

A. Tenhiälä / Journal of Operations Management 29 (2011) 65–77 69

Table 1Sample overview.

Industry Geographic scope Studied plants Production processes InterviewsProducts

Air defense artilleryCannons, fire-control units, and related electronics Global 8 14 3

Aero-derivative power turbinesTurbines and auxiliary equipment for power generation and

the secondary recovery of oil and gasGlobal 8 20 2

Factory automationFlexible manufacturing systems, robotized production cells,

and loading/deburring/measuring stations for machinetools

Europe and the U.S. 7 21 2

Heavy-capacity industrial cranesProcess cranes for waste management, paper, and steel

industries, gantry and container cranes for shipyards andharbors

Global 6 18 3

Reactive power compensation systemsCapacitor banks, static compensators, and harmonic filters Global 3 3 2

Remote-refrigerated display cabinets for grocery retailersRefrigerators, freezers, combination cabinets, and deli bars Northern and Eastern Europe 5 10 3

U respo

ihactnrows

3

wpoctisEcmaGv9laT

bpcspdvod

Special-purpose elevatorsElevators for skyscrapers and ships and luxury elevator cars Global

nit of analysis: production process; total amount of usable survey responses = 89;

ed organizations so that the best-informed respondents could beandpicked with the aid of the companies’ managers. Furthermore,lthough the differences between the planning methods are con-eptually clear, it does not mean that all practitioners would be ableo distinguish their practices from the alternatives of a question-aire. Site visits and work observations were necessary to ensureespondents’ reliability regarding this issue. Also, for the purposef reliability, the questionnaire was made as simple as possible andas pre-tested with practitioners who were interviewed for this

tudy.

.2. Data collection

As all of the above requirements favored a focused study, dataere collected in a sample of seven machinery manufacturing sup-ly chains, which operated in different industry sectors. The unitsf analysis were the MTO production processes of 40 plants from 5ountries: Canada, Finland, Switzerland, the United Kingdom, andhe United States. A total of 98 different production processes weredentified within the sample, and electronic questionnaires wereent to their production planners during the winter of 2006–2007.ach respondent gave information only about the production pro-ess that he or she was responsible for. The survey form was firstade in Finnish and then translated to English and German to allow

ll informants to answer in their native languages. (The English anderman forms were back-translated to ensure the similarity acrossersions.) The survey yielded 89 usable responses (response rate:1%), so the room for non-respondent bias is limited. Neverthe-

ess, the reason for each non-response was traced to ensure thebsence of any common cause. An overview of the sample is given inable 1.

In addition to the survey, qualitative data were collected toetter interpret the results of the statistical tests. Altogether, 21lanners were interviewed about the rationales of choosing theirapacity planning techniques. The interviews were done before the

urvey and thus enabled the development of contextually appro-riate measures and the testing of the questionnaire. Due to slightifferences in the test questionnaires, the responses of the inter-iewees are not included in hypothesis testing. Site visits and workbservations were conducted in 16 plants, and objective reportingata were collected from 38 production processes.

4 12 6

nse rate = 91%.

3.3. Measures

Fig. 3 shows the operationalization of the hypotheses. Asopposed to the prominent earlier comparisons of planningmethods, this study’s dependent variable represents operationalperformance. Earlier analysts, such as Jonsson and Mattsson (2002,2003), have used planners’ satisfaction as the dependent variable.However, satisfaction is a problematic variable because it dependsnot only on the effectiveness of the solutions but also on respon-dents’ expectations about them (e.g., Churchill and Surprenant,1982). Surveys of planners’ satisfaction with their methods prob-ably favor simpler solutions because the more sophisticatedmethods are likely to carry higher expectations.

Operational performance is a multidimensional construct, andthe most suitable dimension to be used as a dependent variable typ-ically depends on the context of the study (Donaldson, 2001). Themost appropriate dimension for this study is delivery performance.It is the most immediate indicator of successful capacity planningbecause the primary objective of all capacity planning methods isto ensure the feasibility of production plans (Vollmann et al., 2005).Delivery performance is also an equally important competitive pri-ority for all of the studied process types. The results of Safizadehet al. (2000) show that the importance of other commonly useddimensions, such as quality, cost efficiency, and flexibility, tendsto vary significantly between job shops, batch processes, and pro-duction lines. Furthermore, delivery performance is an extremelyimportant competitive priority to all studied processes becausethe sample is collected from the manufacture of capital-intensivemachinery, where even small delays are costly and inconvenient forcustomers (Yeo and Ning, 2002). The importance of delivery per-formance was further emphasized at the time of the study becausethe manufacturing industries were experiencing a strong economicupswing. It underlined the criticality of capacity planning becausethe ineffective planning of highly utilized resources always leads todelivery failures or delivery time promises that are not competitive.

Another factor that influenced the choice of the dependentvariable was feedback from the interviewees who helped in the

development of the questionnaire. Many of them explained thatcost efficiency would be extremely difficult to estimate in compar-ison to competitors because it is an internal measure influencedby many different factors. It would be particularly difficult to esti-mate at the level of individual production processes. Conversely,

70 A. Tenhiälä / Journal of Operations Management 29 (2011) 65–77

les and

cblstp

ttnnibdnr(tsC

aswptta(ntOimsm

Fig. 3. Variab

ompetitors’ delivery performance was said to be well knownecause in the studied industry sectors, delivery times and time-

iness are under constant pressure from customers. Therefore, theuccess rate in bidding and the ability to retain customers gavehe respondents feeling for their processes’ comparative deliveryerformance.

Delivery performance was operationalized as the average ofhree reflective indicators. A multi-item scale is needed to coverhe most important aspects of the construct: conformance, timeli-ess, and lead time. For example, the timeliness aspect alone wouldot be a valid proxy for the entire construct because it can be

nflated with extensively buffered delivery times. Similarly, it maye possible to always deliver exactly as promised if the deliveryates are confirmed much later than the customers had origi-ally requested. However, despite the possible trade-offs, earlieresearch has shown that the items tend to be highly correlatedSzwejczewski et al., 1997). Therefore, many researchers have usedhe scale in empirical analyses (e.g., Swink and Nair, 2007). In thistudy, the indicators can be considered to be fairly reliable, as theirronbach’s alpha is .86 (Nunnally and Bernstein, 1994).

The efforts in capacity planning were operationalized with theverage of two formative indicators that represent the main con-tituents of organizational efforts in planning. The respondentsere first asked whether they consider capacity constraints inroduction planning and, second, whether there is a specific rou-ine for this consideration. This operationalization reflects the ideahat organizational efforts in administrative tasks like planningre typically more effective when they follow a certain routineLillrank, 2003). However, the mere existence of a routine doesot guarantee that individual planners would make any efforts;he first question is thus also necessary (Devaraj and Kohli, 2003).

bviously, the indicators are not necessarily correlated, but this

s not a problem because convergence is not required from for-ative indicators (Bollen, 1984; Shah and Goldstein, 2006). This

imple operationalization was used because the more sophisticatedeasures of planning efforts are typically tied to certain planning

hypotheses.

methods (e.g., Zwikael and Sadeh, 2007). In this study, the effortsand the methods had to be analyzed separately.

The planning methods were operationalized with a “self-typingparagraph approach” (James and Hatten, 1995), in which respon-dents were given brief descriptions of each planning method andasked to choose the one that best describes their own method.Earlier studies have found this kind of an operationalization advan-tageous for various reasons (e.g., King and Teo, 1997; Slater andOlson, 2001; DeSarbo et al., 2005). In the context of this study, itsmain benefit is that it does not assume the respondents to be famil-iar with all alternative planning methods and their textbook labels.In the interviews, the operationalization was also found to be insen-sitive to plant-specific terminologies. The approach is also suitablebecause the planning methods constitute a naturally categoricalvariable due to the bottom-up re-planning procedure (Fransoo andWiers, 2008; Vollmann et al., 2005).

The process types were measured according to their operationaldefinitions (Ketokivi and Schroeder, 2004). The respondents wereasked to select the process type that best describes the processfor which they are responsible (e.g., Das and Narasimhan, 2001;Safizadeh et al., 2000; Swink et al., 2005). The interviewees foundno problem with the forced choice between the categories becausethe unit of analysis was an individual process and not an entireplant.

The fitness between planning methods and process types wasoperationalized dichotomously. This approach to contingencypropositions is called “fit as matching,” and its advantage is thatfitness can be determined in isolation of the dependent variable(Venkatraman, 1989). Each studied process was coded either asfit (1) or unfit (0) on the basis of Hypothesis 3. In addition to thiscoding, two complementary variables were created to distinguish

between the two different kinds of unfitness that occur above andbelow the diagonal of Fig. 2. The first of these variables markedthose processes where the planning method was too rudimentary,while the second one marked the processes with overly sophisti-cated methods.

A. Tenhiälä / Journal of Operations Management 29 (2011) 65–77 71

Table 2Descriptive statistics and inter-correlations of continuous variables.

Variable Scale Range Mean S.D. Correlations

1 2 3

1 Delivery performance Likert scale 1–5 3.10 .732 Organization size+ # of employees 8–710 164 170 −.063 Product complexity % of customization 10–100 72 21 .03 .09

t the

tcKeqq

3

susobketmwad.a

oshptbmtbpuwTpa

(ottoodtecmc

4 Planning effort Likert scale 1–5

+ Logarithmic transformation of the organization’s size is used in the analyses, bu* p < .001.

The number of employees and the extent of product cus-omization were used as control variables because they representomplexities that often hamper delivery performance (Vachon andlassen, 2002). The former was measured as a sum of internal andxternal labor, and the latter was measured as the average betweenuantity and value proportions of customized products. The entireuestionnaire is reproduced in Appendix A.

.4. Formal evaluation of perceptual measures’ reliability

The perceptual measures of survey studies have received theirhare of criticism. Therefore, respondents’ reliability has to be eval-ated in a formal manner (Ketokivi and Schroeder, 2004). In thistudy, inter-rater reliability (IRR) analyses and comparisons withbjective data were used for this purpose. The IRR coefficients coulde calculated for seven production processes that had two equallynowledgeable planners. These production processes representedach of the studied industry sectors, and they were scattered acrosshe different cells of Fig. 2 so that they represented all planning

ethods and both fit and unfit positions. The results of the analysesere quite convincing: intraclass correlation coefficients (Shrout

nd Fleiss, 1979) were significant (p < .001) for each pair of respon-ents and their average was .94. Even the lowest coefficient was

79. There were no significant differences between the consistencynd the absolute agreement definitions of IRR.

The criterion validity of the dependent variable was tested withbjective data collected from 38 production processes during theite visits of this research project. They represent approximatelyalf of the processes that belonged to the visited plants. In thoselants where the managers did not want to share objective data,he main reason was that they did not think that the data woulde sufficiently reliable. The distributions of the perceptual perfor-ance measures were similar between those processes that shared

he data and those that did not, which is evidence against a possi-le self-selection bias (e.g., only the managers of best-performingrocesses choosing to share their data). The plant managers werenanimous in suggesting the use of on-time delivery performance,hich corresponds to the second item of the perceptual scale.

hey considered it the most reliable and valid indicator of deliveryerformance in their reporting systems. The most recent annualverage was used in the test.

The comparison of the measures yielded a correlation of .61p < .001). The result can be considered satisfactory because thebjective measures are not unproblematic, either. The organiza-ions had varying definitions for timeliness; some calculated it onhe level of individual items, while others aggregated it to deliveriesr orders. Similarly, the baseline could be the original promised dater some adjusted date. The baselines also ranged from ex-worksates to commissioning dates. Moreover, the objective measure has

he disadvantage of not capturing the lateness of the delayed deliv-ries. Meanwhile, perceptual measures tend to be less sensitive toontext-specific definitions. This applies especially to performanceeasures that can be formulated to be relative to the principal

ompetition (e.g., Swink and Nair, 2007).

3.49 1.00 .36* .07 −.05

mean and the standard deviation are shown untransformed to ease interpretation.

4. Results

4.1. Descriptive statistics

Table 2 presents the descriptive statistics and inter-correlationsof the continuous variables. Table 3 shows the average values of thecontinuous variables in each pair of categorical variables. The num-bers appear reasonable, given that users of finite-loading methodsput the most effort into planning, while non-systematic plannersmake the least effort.

4.2. Hypothesis testing

The hypotheses are tested with hierarchical regression analy-sis. The results are shown in Table 4. The first step of the analysisincludes all control variables. The dummy variables for the stud-ied industry sectors have considerable explanatory power becausesome of the studied processes share the same competitors andperformance standards due to the focused research design. There-fore, controlling for this effect is crucial, but the coefficients arenot theoretically interesting and are thus omitted from the table.The variables for the size of the organization and the complexity ofproducts turn out to be insignificant. The only significant differencebetween process types can be found between job shops and batchprocesses.

The second step adds the planning effort into the equation. It hasa positive effect, as predicted in Hypothesis 1. This result is impor-tant because it shows that the sample is valid for the comparisonof the planning methods. If the sample had included a lot of simpleproduction processes where time-phased planning is unnecessary,then the hypothesis would not have been supported, and the com-parison of methods would not have been meaningful.

The third step adds the planning methods into the analysis andenables the comparison of different planning techniques’ effective-ness. None of the methods has a significant effect, which means thatHypothesis 2 does not receive any support from the data.

The fourth step of the analysis is done in two ways. Step 4ashows the average difference between the fitting and the unfit-ting planning methods of Fig. 2. Step 4b shows the effects of usingunfitting methods and distinguishes between methods that are lesssophisticated and those that are more sophisticated than proposed.However, the difference between the negative effects of the overlysophisticated and too rudimentary methods is not statistically sig-nificant (z = .77, p = .44; for details of the comparison method, seePaternoster et al., 1998). In any case, the results give fairly strongsupport to Hypothesis 3 because all three variables have signifi-cant coefficients and explain a considerable proportion of variancein performance.

4.3. Interviews

One important observation from the statistics is the wide uti-lization of the non-systematic planning method and methods thatdo not fit the processes where they are used. This raises the question

72 A. Tenhiälä / Journal of Operations Management 29 (2011) 65–77

Table 3Averages of continuous variables in each pair of categorical variables.

Planning method Job shop Batch process Batch processwithbottleneckcontrol

Production line

Non-systematic capacity planning nOrganization sizeProduct complexityPlanning effortDelivery performance

3295642.503.23

3229902.253.06

3283742.503.06

11126691.962.74

Rough-cut capacity planning nOrganization sizeProduct complexityPlanning effortDelivery performance

12204733.793.98

9126833.722.51

351483.332.54

2127933.752.31

Capacity requirements planning nOrganization sizeProduct complexityPlanning effortDelivery performance

5115713.402.75

10221693.903.36

1350865.003.17

265794.003.37

Finite loading with capacity leveling nOrganization sizeProduct complexityPlanning effortDelivery performance

4153823.882.91

1240864.002.64

564714.103.23

5191544.503.43

Finite loading with optimization nOrganization sizeProduct complexity

331364

150079

17686

520250

D

owaamdwt

TR

DTs

Planning effortDelivery performance

4.003.02

elivery performance scores are adjusted with industry averages.

f how practitioners choose their planning methods. The questionas addressed in the interviews. The most illustrative quotations

re presented in Table 5. The interviewees’ opinions had consider-

ble similarities; for example, planners who used non-systematicethods or RCCP in unfit contexts shared a feeling that the more

etailed techniques would be overwhelmingly complicated. Mean-hile, the planners who used RCCP in fitting contexts explained

hat “fancier” techniques would probably exist but that they had

able 4egression results.

Variable Step 1: Controlvariables

Step 2: Planningeffort

Constant 2.91* (.48) 2.12* (.51)(Industry dummy variables)Organization size .06 (.07) .05 (.07)Product complexity −.06 (.05) −.04 (.05)(Job shop)Batch process −.40‡ (.20) −.48† (.19)Batch process with bottleneck control −.16 (.25) −.26 (.24)Production line −.28 (.22) −.13 (.21)Planning effort .28* (.09)(Non-systematic planning)Rough-cut capacity planningCapacity requirements planningFinite loading with capacity levelingFinite loading with optimizationFitting methodsToo rudimentary methodsOverly sophisticated methodsR2|R̄2 .53|.43 .61|.52�R2 .53 .08�F for �R2 5.3* 9.6*

ependent variable: Delivery performance. Regression coefficients are unstandardized. Stanhey are only shown for the last steps because that is where they get the lowest values.erve as the baseline of the process types. Non-systematic planning is the baseline of the

* p < .01.† p < .05.‡ p < .10.

4.502.38

4.003.10

4.203.62

not explored them because they were satisfied with the outcomesof their current practices.

In both the fitting and the unfitting contexts, the rationale for

using CRP was that it was part of the companies’ ERP systems. Noneof the interviewees knew whether their ERP systems featured anyalternative methods. In cases where CRP should not have been used,the planners blamed the unreliability of their plans on the poorusability of their ERP systems. However, even those planners who

Step 3: Planningmethods

Step 4a: Fittingmethods

Step 4b: Unfittingmethods

2.08* (.56) 2.30* (.39) 3.13* (.43)

.05 (.07) .04 (.05) [.66] .03 (.05) [.64]−.04 (.05) .02 (.04) [.74] .02 (.04) [.74]

−.52† (.20) −.57* (.14) [.57] −.65* (.20) [.30]−.31 (.27) −.18 (.19) [.57] −.32 (.29) [.26]−.12 (.25) −.28 (.18) [.33] −.41 (.27) [.15].31† (.13) .15 (.09) [.30] .14 (.10) [.28]

−.13 (.28) −.33 (.20) [.31] −.31 (.20) [.31].09 (.31) −.26 (.23) [.23] −.14 (.30) [.14].02 (.37) −.41 (.27) [.26] −.23 (.39) [.13]−.18 (.37) −.29 (.26) [.31] −.10 (.39) [.14]

.85* (.13) [.61]−.72* (.24) [.17]−.98* (.24) [.30]

.62|.49 .81|.74 .81|.74

.01 .19 .19

.3 45.5* 22.7*

dard errors are shown in the parentheses. Tolerances are shown in square brackets.Coefficients of the industry dummy variables are omitted to save space. Job shopsplanning methods.

A. Tenhiälä / Journal of Operations Management 29 (2011) 65–77 73

Table 5Rationales for selecting planning methods.

Job shop Batch process Batch process with bottleneckcontrol or production line

Non-systematic capacity planning n1 = 3; n2 = 1“[Formal planning methods]are not worth the trouble”

n1 = 3; n2 = 3“We have had badexperiences fromsystematic techniques”“Our trials with planningtools have failed”

n1 = 14; n2 = 2“We do not use any planningsoftware because they wouldonly make things complicated”

Rough-cut capacity planning (RCCP) n1 = 12; n2 = 2 (Fit)“This is a sufficiently robustmethod for our needs”“Fancier solutions probablyexist but we do not need them”

n1 = 9; n2 = 2“More detailed methodscould be beneficial but theytend to incur more workand make things difficult”

n1 = 5; n2 = 1“The advantage of a simplemethod is that everyone canunderstand how the plans arederived”

Capacity requirements planning (CRP) n1 = 5; n2 = 2“We have done capacityplanning in this way since theimplementation of our ERPsystem”

n1 = 10; n2 = 2 (Fit)“This is how capacityplanning is done in our ERPsystem”

n1 = 3; n2 = 1“This is the only way to docapacity checks in our [ERP]system”

Finite loading with capacity leveling or optimization n1 = 7; n2 = 1“We implemented [thefinite-loading software]because it was recommendedby our consultants”

n1 = 2; n2 = 1“This software is used forthe planning of allprocesses in our plant”

n1 = 16; n2 = 3 (Fit)“This tool was implementedbecause the ERP systemrequired too much manualwork”“This tool is needed to ensure

n total =

uwam

tsfac

5

5

“uaappaeeopgto

“trpthnm

1 is the frequency in the survey (total = 89); n2 is the frequency in the interviews (

sed CRP in the correct contexts reported that capacity planningas a particularly challenging part of their work. This notion is

ligned with the earlier discussions about the batch shop being theost complex process type for capacity planning.The most typical reason to adopt finite-loading methods was

hat someone in the organization had come across a convincingoftware tool. In situations where these methods were too detailedor the process, users admitted the existence of problems butttributed them to the incorrect use of the software. The method’sontextual unsuitability was not suspected.

. Discussion

.1. Organization-theoretical perspectives to capacity planning

The results indicate that there is a time and place for suchimprecise” planning methods as RCCP and CRP, whose widespreadtilization has astonished academics (Halsall et al., 1994; Jonssonnd Mattsson, 2003). It seems that if finite-loading techniquesre used in job shops, they encourage making tight schedules forrocesses that are not sufficiently stabile for them. The job-shoprocess design is based on preferring flexibility to efficiency (Hayesnd Wheelwright, 1979; Safizadeh et al., 1996). Therefore, comput-rized capacity leveling or optimization, which specifically aims forfficiency, is out of place in those environments. In the terminol-gy of contingency theory, the resources of job shops are said to beooled (Thompson, 1967). If the resources are by definition aggre-ated, then it is not surprising that the planners, who use detailedechniques, complain that their plans are not robust enough, asbserved by Wiers (1997).

In batch processes, the challenge of the detailed planning isshifting bottlenecks” (Monahan and Smunt, 1999). Finite-loadingechniques do not seem to work, despite the fact that many algo-ithms and software tools have been developed to tackle the

roblem (e.g., Kouvelis et al., 2005; SAP, 2010b). Instead of blaminghe tools or their users, it can be asked whether the failures couldave more fundamental causes. A contingency-theoretical expla-ation is that planning itself becomes a less effective coordinationechanism in the reciprocal processes of batch shops (Galbraith,

the feasibility of our plans”

21).

1973; Tushman and Nadler, 1978; Barki and Pinsonneault, 2005).Thus, instead of striving for more detailed planning, the managers ofbatch shops would be better off by investing in capabilities to solveexceptions in the execution of the plans (Perrow, 1967; Reeves andTurner, 1972).

Finite-loading techniques work in bottleneck-controlled batchshops and production lines because the complexity of these pro-cesses is reduced by the fact that the tasks to be planned aresequentially interdependent (Thompson, 1967). Revising the plansis simple because changes in the schedule of one resource onlyinfluence the resources in the downstream of the process.

In addition to contingency theory, the limited applicability ofthe advanced planning methods is also aligned with the con-cept of bounded rationality (March and Simon, 1958). It holds thatin the complex reality of organizations, it is usually sufficientto satisfy some level of performance, and only in special occa-sions, it is possible to try to optimize the results (Simon, 1978). Incapacity planning, the special occasions take place when schedul-ing problems can be narrowed down to fairly static formulae,status information from the processes is complete, and the pro-cesses can be isolated from external uncertainties. Such conditionshold badly in typical job shops and batch processes (Reeves andTurner, 1972). In most cases, the scheduling problems depend onwhat products are loaded onto the processes, the real-time col-lection of precise status information is not economically viable,and the processes cannot be completely sealed from their envi-ronments.

In summary, several classic organization-theoretical conceptsgive reasons to suspect the universal applicability, let alone supe-riority, of the most sophisticated capacity planning methods.However, practitioners appear to be uninformed about the impor-tance of matching planning methods with their processes. It alsoseems that the issue is not discussed in the existing literature.Therefore, the results of this study elaborate the benefits of tak-

ing theoretical perspectives to operations management topics,which have been traditionally viewed from a problem-solvingperspective (Schroeder, 2008). Specifically, the findings demon-strate the practical utility of contingency theory (Sousa and Voss,2008).

7 ions M

5

mwiidclt(Tt2cleocdaoi

rts(maao2bhe

5

dmelstaHp1st

mmpdw

vcficbk

4 A. Tenhiälä / Journal of Operat

.2. Possible extensions

Capacity planning is by no means the only area of operationsanagement where organization theory could help us understandhy practitioners often rely on much simpler methods than what

s promoted in the literature. One potential area of contributions demand forecasting, where researchers have constantly intro-uced new and ever-more sophisticated variants of time-series,ausal, genetic, and neural network models. Over the years, ana-ytical comparisons of different techniques have repeatedly shownhat the more sophisticated methods outperform the simpler onesHogarth and Makridakis, 1981; Hill et al., 1996; Alon et al., 2001;aylor, 2008). Yet, the practical implementations of the advancedechniques are lagging (Klassen and Flores, 2001; Syntetos et al.,009), and some of them have reportedly led to disastrous out-omes (e.g., Worthen, 2003). Reviews show that despite the vastiterature on the advanced forecasting methods, most practition-rs still rely on human judgment (Lawrence et al., 2006). Perhapsrganization theory could also inform the research on demand fore-asting and offer a basis for propositions on how the effectiveness ofifferent techniques is contingent on the forecasting environmentnd factors other than the characteristics of the input data and thebjectives of the forecasting task, which have already been studiedn the existing literature.

Another similar topic is contracting. Decades of analyticalesearch have produced numerous models for incentive alignmenthrough revenue sharing, yet practitioners tend to prefer muchimpler and mathematically inefficient wholesale price contractse.g., Cachon, 2003). Also in this area, empirical evidence is largely

ixed regarding the benefits of using the sophisticated models (Limnd Teck-Hua, 2007; Katok and Wu, 2009). Recent research haslready sought to explain this phenomenon with different aspectsf bounded rationality (e.g., Ho and Zhang, 2008; Loch and Wu,008). Perhaps this stream of behaviorally oriented research coulde complemented with contingency-theoretical propositions onow the task environments of the contracting parties influence theffectiveness of different contract types.

.3. Limitations

One limitation of this study is that the same informants providedata for both the dependent and the independent variables. Thisay have introduced common method bias and especially influ-

nced the test of Hypothesis 1, which was based on an analysis of ainear relationship between two continuous variables. In such tests,ingle-source data can lead to inflated regression coefficients andhe false identification of significant effects. However, two issueslleviate this concern. First, all continuous measurement items passarman’s single-factor test; that is, they load on different com-onents in a principal component analysis (Podsakoff and Organ,986). Second, the high correlation between the perceptual mea-ure of timeliness and the objective data indicates that the percep-ual measures cannot be entirely driven by common method bias.

When it comes to Hypothesis 2, one could argue that commonethod bias would operate so that those respondents who use theost sophisticated planning methods feel pressure to report higher

erformance levels than their processes actually achieve. However,espite this possible positive bias, no support for the hypothesisas found.

In the testing of Hypothesis 3, the construction of the fitnessariable makes it difficult to imagine how common method bias

ould have influenced the results. In order for the regression coef-cients to be inflated, the respondents should have known, eitheronsciously or subconsciously, the contingency argument that waseing tested. The interviews provided strong evidence against suchnowledge existing among practitioners.

anagement 29 (2011) 65–77

Another limitation is the reliance on delivery performance in theanalyses. It may have understated the contributions of optimiza-tion tools because they can be used to minimize costs instead ofmaximizing schedule adherence (Stadtler and Kilger, 2005). Hence,the job shops and batch processes whose delivery performance hadsuffered from the finite loading may have benefited in terms of effi-ciency. This possibility is left as a topic of further research becauseit would necessitate a different research design. Possible efficiencyadvantages can hardly be measured in a cross-sectional surveywhere respondents are asked to evaluate their production costsrelative to competition. The possible benefits occur more likely asincreased productivity, which is difficult to evaluate in compar-ison to competitors. Consequently, a more appropriate researchdesign would be a longitudinal study on an implementation of afinite-loading tool.

A third limitation results from the theoretical sampling of thisstudy. It is that the regression coefficients are not generalizableto any specific population. Therefore, the most important result ofthis study is not the estimated effects but the overall support for thecontingency proposition of Fig. 2. This kind of outcome is alignedwith the inductive logic of strong inference research because itspurpose is not to make generalizations from samples to populationsbut rather to make generalizations from observations to theory(Stuart et al., 2002). However, it also means that replications areneeded if more accurate estimates of the effect sizes are desired.

6. Conclusions

The results of this study give tentative support for a contin-gency theory of capacity planning. It proposes that the complexityof process types determines the applicability of different capac-ity planning methods. As illustrated in Fig. 2, the theory proposesthat there are two possible ways to misalign planning methodswith production processes: the methods can be either too sim-ple or too sophisticated. The theory offers a new answer to thequestion of why so many practitioners use less sophisticated plan-ning methods than what is discussed in the literature (Jonsson andMattsson, 2003; McKay et al., 2002). As the results indicate that theless sophisticated methods are more effective in some processes,it is not appropriate to attribute the gap to practitioners’ lack ofmathematical skills or insufficient training (Hopp et al., 2007).

The results of this study have several practical implications.First, most practitioners seemed to be unaware of the various alter-native methods in capacity planning and the limitations of theirapplicability. As the unsuitable methods turned out to be verycommon, it can be proposed that exploring the options would bebeneficial for many organizations. Second, if a planning tool doesnot appear to work, then the culprit is not necessarily the softwareor its users. Instead, the entire planning method may be unfit forthe process. In that case, a more appropriate method should be cho-sen on the bases of Fig. 2. Third, although ERP systems provide finetools for CRP (e.g., McKay and Wiers, 2004), it does not mean thatCRP is the “best practice” for everyone. The users of ERP systemsshould consider whether to use CRP or something else. In job shops,it is sufficient to use rough-cut methods, which are usually also fea-tured in ERP systems (e.g., SAP, 2010a). However, users may not beaware of them because they may be less promoted due to theirrudimentary image. On the other hand, if the production processis sequential, then finite-loading techniques are more suitable andthe organization should consider an investment in some add-on

software.

Overall, the findings indicate that there is still work to bedone with such a seemingly mature topic as capacity planning. Animportant lesson lies in the wide use of non-systematic planningmethods. It may suggest that contemporary operations manage-

ions M

mptimdnpdop

ampWrtw

A. Tenhiälä / Journal of Operat

ent curricula often fail to provide appropriate skills in capacitylanning. Perhaps the capacity planning exercises of typical opera-ions management courses are too simplified and delude studentsnto favoring non-systematic methods. Alternatively, some courses

ay overly emphasize the most advanced methods, as many aca-emics probably like them better than RCCP and CRP, which areot mathematically challenging. This would explain why so manyractitioners presumed systematic capacity planning to be veryifficult. In any case, the findings suggest that the educators ofperations management could generally do a better job in teachingractically useful capacity planning techniques.

The observed effectiveness of the simpler planning methodslso questions the practical utility of focusing research on the opti-ization techniques. The findings thus support the calls for more

ragmatic research in operations management (Guide and Van

assenhove, 2007; Hopp et al., 2007). A pragmatic approach to the

esearch on production planning would acknowledge that a singleechnique or practice can seldom prevail in all environments. Itould also acknowledge that optimization is not always desirable

Operational performance

Name the three most important competitors for the products of your production p

Evaluate the performance of your production process in comparison to theabove-named competitors [1: much worse, 2: somewhat worse, 3: about similar,4: somewhat better, 5: much better]The ability to confirm deliveries for the first requested datesThe ability to deliver on the confirmed delivery dateThe average lead time from order acquisition to delivery

Capacity planning

How do the following statements describe the planning practices within yourproduction process? [1: very poorly, 2: somewhat poorly, 3:moderately, 4: quitewell, 5: very well]“We consider capacity constraints in our production planning”“We have a specific routine for our capacity planning efforts”

What is the main capacity planning method in your production process? (“The mainmethod” determines the output to which your production process commits itself.)[Check the closest option]

Non-systematic methods: the feasibility of production plans is evaluated on the basesof production planners’ personal experience or spreadsheet calculations

Rough-cut capacity planning: the feasibility of production plans is evaluated bycalculating the required capacity utilization at the level of the entire productionprocess

Capacity requirements planning: the feasibility of production plans is evaluated bycalculating the required capacity utilization at the level of individual resources orwork centers

Capacity leveling: the feasibility of production plans is ensured with a software toolthat is used to rearrange the schedules so that critical capacity constraints are notbreached

Optimization: the feasibility of production plans is ensured with a software tool that isused to optimize the utilization of resources

Process characteristics

What is the type of your production process? [Check the closest option]� Job shop � Batch process � Bo

How many people are directly employed in your production process?Total amount of permanent employees: Aver

What proportion of products requires order-specific customization in your producPercentage of product quantities: Perc

anagement 29 (2011) 65–77 75

in complex real-world planning situations. The wider adoption oforganization-theoretical concepts, such as contingency effects andbounded rationality, in the research at the technical level of every-day operations management could be helpful in the developmentof a more pragmatic discipline.

Acknowledgements

I would like to thank Mikko Ketokivi, Kari Tanskanen, GopeshAnand, Suzanne de Treville, and Kenneth K. Boyer as well as threeanonymous reviewers and the Associate Editor for their commentsand suggestions that greatly helped me to improve this paper. Iam also thankful for the financial support from Tekes—the Finnishfunding agency for technology and innovation. In particular, I amgrateful to the studied companies and their employees whose con-tributions made this research possible.

Appendix A. Questionnaire

rocess

1 2 3 4 51 2 3 4 51 2 3 4 5

1 2 3 4 51 2 3 4 5

�

�

�

�

�

ttleneck-controlled batch process � Production line

age amount of contract or temporary workers:

tion process?entage of products’ value:

7 ions M

R

A

B

B

B

B

B

C

C

C

D

D

D

D

D

D

E

F

G

G

G

G

G

H

H

H

H

H

H

H

H

H

J

J

J

J

K

6 A. Tenhiälä / Journal of Operat

eferences

lon, I., Qi, M., Sadowski, R.J., 2001. Forecasting aggregate retail sales: a comparisonof artificial neural networks and traditional methods. Journal of Retailing andConsumer Services 8 (3), 147–156.

arki, H., Pinsonneault, A., 2005. A model of organizational integration, implemen-tation effort, and performance. Organization Science 16 (2), 165–179.

endoly, E., Cotteleer, M.J., 2008. Understanding behavioral sources of processvariation following enterprise system deployment. Journal of Operations Man-agement 26 (1), 23–44.

lackstone Jr., J.H., Cox III, J.F. (Eds.), 2005. APICS Dictionary, 11th ed. APICS – TheAssociation for Operations Management, Alexandria, VA.

ollen, K.A., 1984. Multiple indicators: internal consistency or no necessary rela-tionship? Quality and Quantity 18 (4), 377–385.

urcher, P.G., 1992. Effective capacity planning. Management Services 36 (10),22–25.

achon, G.P., 2003. Supply chain coordination with contracts. In: de Kok, A.G., Graves,S.C. (Eds.), Supply Chain Management: Design, Coordination and Operation.Elsevier, Amsterdam, Netherlands, pp. 229–340.

yert, R.M., March, J.G., 1963. A Behavioral Theory of the Firm. Prentice-Hall, Engle-wood Cliffs, NJ.

hurchill Jr., G.A., Surprenant, C., 1982. An investigation into the determinants ofcustomer satisfaction. Journal of Marketing Research 19 (4), 491–504.

as, A., Narasimhan, R., 2001. Process-technology fit and its implications for manu-facturing performance. Journal of Operations Management 19 (5), 521–540.

avis, D.J., Mabert, V.A., 2000. Order dispatching and labor assignment in cellularmanufacturing systems. Decision Sciences 31 (4), 745–771.

eblaere, F., Demeulemeester, E., Herroelen, W., Van de Vonder, S., 2007. Robustresource allocation decisions in resource-constrained projects. Decision Sci-ences 38 (1), 5–37.

eSarbo, W.S., Di Benedetto, C.A., Song, M., Sinha, I., 2005. Revisiting the Milesand Snow strategic framework: uncovering interrelationships between strategictypes, capabilities, environmental uncertainty, and firm performance. StrategicManagement Journal 26 (1), 47–74.

evaraj, S., Kohli, R., 2003. Performance impacts of information technology: is actualusage the missing link? Management Science 49 (3), 273–289.

onaldson, L., 2001. The Contingency Theory of Organizations. Sage Publications,Thousand Oaks, CA.

isenhardt, K.M., 1989. Building theories from case study research. Academy ofManagement Review 14 (4), 532–550.

ransoo, J.C., Wiers, V.C.S., 2008. An empirical investigation of the neglect of MRPinformation by production planners. Production Planning & Control 19 (8),781–787.

albraith, J.R., 1973. Designing Complex Organizations. Addison-Wesley, Reading,MA.

oldratt, E.M., Cox, J., 1984. The Goal: Excellence in Manufacturing. North RiverPress, New York, NY.

reen, G.I., Appel, L.B., 1981. An empirical analysis of job shop dispatch rule selection.Journal of Operations Management 1 (4), 197–203.

uide Jr., V.D.R., Van Wassenhove, L.N., 2007. Dancing with the devil: partneringwith industry but publishing in academia. Decision Sciences 38 (4), 531–546.

upta, S., Verma, R., Victorino, L., 2006. Empirical research published in Productionand Operations Management (1992–2005): trends and future research direc-tions. Production and Operations Management 15 (3), 432–448.

alsall, D.N., Muhlemann, A.P., Price, D.H.R., 1994. A review of production plan-ning and scheduling in smaller manufacturing companies in the UK. ProductionPlanning & Control 5 (5), 485–493.

ayes, R.H., Wheelwright, S.C., 1979. Link manufacturing process and product lifecycles. Harvard Business Review 57 (1), 133–140.

ill, A.V., 2007. The Encyclopedia of Operations Management. Clamshell Beach Press,Eden Prairie, MN.

ill, T., 2005. Operations Management, 2nd ed. Palgrave MacMillan, Basingstoke,UK.

ill, T., O’Connor, M., Remus, W., 1996. Neural network models for time series fore-casts. Management Science 42 (7), 1082–1092.

o, T.-H., Zhang, J., 2008. Designing pricing contracts for boundedly rational cus-tomers: does the framing of the fixed fee matter? Management Science 54 (4),686–700.

ogarth, R.M., Makridakis, S., 1981. Forecasting and planning: an evaluation. Man-agement Science 27 (2), 115–138.

opp, W.J., Iravani, S.M.R., Shou, B., 2007. A diagnostic tree for improving productionline performance. Production and Operations Management 16 (1), 77–92.

opp, W.J., Spearman, M.L., 2004. To pull or not to pull: what is the question?Manufacturing & Service Operations Management 6 (2), 133–148.

ames, W.L., Hatten, K.J., 1995. Further evidence on the validity of the self typingparagraph approach: Miles and Snow strategic archetypes in banking. StrategicManagement Journal 16 (2), 161–168.

ick, T.D., 1979. Mixing qualitative and quantitative methods: triangulation in action.Administrative Science Quarterly 24 (4), 602–611.

onsson, P., Mattsson, S.-A., 2002. Use and applicability of capacity planning methods.

Production and Inventory Management Journal 43 (3/4), 89–95.

onsson, P., Mattsson, S.-A., 2003. The implications of fit between planning environ-ments and manufacturing planning and control methods. International Journalof Operations & Production Management 23 (8), 872–900.

anet, J.J., Sridharan, V., 1998. The value of using scheduling information in planningmaterial requirements. Decision Sciences 29 (2), 479–496.

anagement 29 (2011) 65–77

Karmarkar, U., 1989. Getting control of just-in-time. Harvard Business Review 67(5), 122–131.

Katok, E., Wu, D.Y., 2009. Contracting in supply chains: a laboratory investigation.Management Science 55 (12), 1953–1968.

Kemppainen, K., 2007. Plans and rules: a study of order management and operationsscheduling in manufacturing companies. In: De Koster, R., Delfmann, W. (Eds.),Managing Supply Chains: Challenges and Opportunities. Copenhagen BusinessSchool Press, Copenhagen, Denmark, pp. 175–200.

Kenat, J.J., Sridharan, V., 1998. The value of using scheduling information in planningmaterial requirements. Decision Sciences 29 (2), 479–497.

Ketokivi, M., 2006. Elaborating the contingency theory of organizations: the caseof manufacturing flexibility strategies. Production and Operations Management15 (2), 215–228.

Ketokivi, M., Schroeder, R.G., 2004. Perceptual measures of performance: fact orfiction? Journal of Operations Management 22 (3), 247–264.

King, W.R., Teo, T.S.H., 1997. Integration between business planning and informa-tion systems planning: validating a stage hypothesis. Decision Sciences 28 (2),279–308.

Klassen, R.D., Flores, B.E., 2001. Forecasting practices of Canadian firms: surveyresults and comparisons. International Journal of Production Economics 70 (2),163–174.

Kouvelis, P., Chambers, C., Yu, D.Z., 2005. Manufacturing operations manuscriptspublished in the first 52 issues of POM: review, trends, and opportunities. Pro-duction and Operations Management 14 (4), 450–467.

Kreipl, S., Pinedo, M., 2004. Planning and scheduling in supply chains: an overviewof issues in practice. Production and Operations Management 13 (1), 77–92.

Landvater, D.V., Gray, C.D., 1989. MRP II Standard System: A Handbook for Manu-facturing Software Survival. John Wiley & Sons, New York, NY.

Lawrence, M., Goodwin, P., O’Connor, M., Önkal, D., 2006. Judgmental forecasting:a review of progress over the last 25 years. International Journal of Forecasting22 (3), 493–518.

Lillrank, P., 2003. The quality of standard, routine and nonroutine processes. Orga-nization Studies 24 (2), 215–233.

Lim, N., Teck-Hua, H., 2007. Designing price contracts for boundedly rational cus-tomers: does the number of blocks matter? Marketing Science 26 (3), 312–326.

Loch, C.H., Wu, Y., 2008. Social preferences and supply chain performance: an exper-imental study. Management Science 54 (11), 1835–1849.

MacKenzie, K.D., House, R., 1978. Paradigm development in the social sciences: aproposed research strategy. Academy of Management Review 3 (1), 7–23.

March, J.G., Simon, H.A., 1958. Organizations. John Wiley, New York, NY.McKay, K., Pinedo, M., Webster, S., 2002. Practice-focused research issues for

scheduling systems. Production and Operations Management 11 (2), 249–258.McKay, K.N., Wiers, V.C.S., 2004. Practical Production Control: A Survival Guide for

Planners and Schedulers. J. Ross Publishing, Boca Raton, FL.Meal, H.C., 1984. Putting production decisions where they belong. Harvard Business

Review 62 (2), 102–111.Melnyk, S.A., Vickery, S.K., Carter, P.L., 1986. Scheduling, sequencing, and dispatch-

ing: alternative perspectives. Production and Inventory Management Journal 27(2), 58–68.

Monahan, G.E., Smunt, T.L., 1999. Processes with nearly-sequential routings: a com-parative analysis. Journal of Operations Management 17 (4), 449–466.

Nadler, J., Thompson, L., Van Boven, L., 2003. Learning negotiation skills: four modelsof knowledge creation and transfer. Management Science 49 (4), 529–540.

Nunnally, J.C., Bernstein, I.H., 1994. Psychometric Theory, 3rd ed. McGraw-Hill, NewYork, NY.

Ohno, T., 1988. Toyota Production System: Beyond Large-Scale Production. Produc-tivity Press, Cambridge, MA.

Olhager, J., 2003. Strategic positioning of the order penetration point. InternationalJournal of Production Economics 85 (3), 319–329.

Olhager, J., Rudberg, M., 2002. Linking manufacturing strategy decisions on processchoice with manufacturing planning and control systems. International Journalof Production Research 40 (10), 2335–2351.

Orlicky, J., 1975. Material Requirements Planning: The New Way of Life in Productionand Inventory Management. McGraw-Hill, New York, NY.

Paternoster, R., Brame, R., Mazerolle, P., Piquero, A., 1998. Using the correct statisticaltest for the equality of regression coefficients. Criminology 36 (4), 859–866.

Perrow, C., 1967. A framework for the comparative analysis of organizations. Amer-ican Sociological Review 32 (2), 194–208.

Platt, J.R., 1964. Strong inference. Science 146 (3642), 347–353.Podsakoff, P.M., Organ, D.W., 1986. Self-reports in organizational research: problems

and prospects. Journal of Management 12 (4), 531.Proud, J.F., 2007. Master Scheduling: A Practical Guide to Competitive Manufactur-

ing, 3rd ed. John Wiley & Sons, Hoboken, NJ.Reeves, T.K., Turner, B.A., 1972. A theory of organization and behavior in batch

production factories. Administrative Science Quarterly 17 (1), 81–98.Safizadeh, M.H., Ritzman, L.P., 1997. Linking performance drivers in production plan-

ning and inventory control to process choice. Journal of Operations Management15 (4), 389–403.

Safizadeh, M.H., Ritzman, L.P., Mallick, D., 2000. Revisiting alternative theoreticalparadigms in manufacturing strategy. Production and Operations Management

9 (2), 111–127.

Safizadeh, M.H., Ritzman, L.P., Sharma, D., Wood, C., 1996. An empirical analysis ofthe product-process matrix. Management Science 42 (11), 1576–1591.

SAP, 2010a. SAP Library: mySAP ERP 6.0: Production Planning and Control,http://help.sap.com/saphelp erp60 sp/helpdata/en/ba/df293581dc1f79e10000009b38f889/frameset.htm (last viewed on February 10, 2010).

ions M

S

S

S

S

S

S

S

S

S

S

S

SS

S

S

S

S

T

A. Tenhiälä / Journal of Operat

AP, 2010b. SAP Library: mySAP SCM 5.0: Production Planning andDetailed Scheduling, http://help.sap.com/saphelp scm50/helpdata/en/c1/147a375f0dbc7fe10000009b38f8cf/frameset.htm (last viewed on February 10,2010).

chroeder, R.G., 2008. Introduction to the special issue on theory developmentin operations management. Production and Operations Management 17 (3),354–356.

hah, R., Goldstein, S.M., 2006. Use of structural equation modeling in operationsmanagement research: looking back and forward. Journal of Operations Man-agement 24 (2), 148–169.

haw, J.D., Gupta, N., Delery, J.E., 2005. Alternative conceptualizations of the rela-tionship between voluntary turnover and organizational performance. Academyof Management Journal 48 (1), 50–68.

heu, C., Wacker, J.G., 2001. Effectiveness of planning and control systems: an empir-ical study of US and Japanese firms. International Journal of Production Research39 (5), 887–905.

hrout, P.E., Fleiss, J.L., 1979. Intraclass correlations: uses in assessing rater reliabil-ity. Psychological Bulletin 86 (2), 420–428.

imon, H.A., 1978. Rationality as process and as product of thought. The AmericanEconomic Review 68 (2), 1–16.

lack, N., Chambers, S., Johnston, R., 2007. Operations Management, 5th ed. PearsonEducation, Harlow, UK.

later, S.F., Olson, E.M., 2001. Marketing’s contribution to the implementation ofbusiness strategy: an empirical analysis. Strategic Management Journal 22 (11),1055–1067.

ousa, R., Voss, C.A., 2008. Contingency research in operations management prac-tices. Journal of Operations Management 26 (6), 697–713.

tadtler, H., Kilger, C. (Eds.), 2005. Supply Chain Management and Advanced Plan-ning, 3rd ed. Springer, Berlin, Germany.

tevenson, W.J., 2004. Operations Management, 8th ed. McGraw-Hill, Boston, MA.tuart, I., McCutcheon, D., Handfield, R., McLachlin, R., Samson, D., 2002. Effec-

tive case research in operations management: a process perspective. Journalof Operations Management 20 (5), 419–433.

wink, M., Nair, A., 2007. Capturing the competitive advantages of AMT: design-manufacturing integration as a complementary asset. Journal of OperationsManagement 25 (3), 736–754.

wink, M., Narasimhan, R., Kim, S.W., 2005. Manufacturing practices and strategyintegration: effects on cost efficiency, flexibility, and market-based perfor-mance. Decision Sciences 36 (3), 427–457.

yntetos, A.A., Boylan, J.E., Disney, S.M., 2009. Forecasting for inventory planning: a50-year review. Journal of the Operational Research Society 60 (S1), S149–S160.

zwejczewski, M., Mapes, J., New, C., 1997. Delivery and trade-offs. InternationalJournal of Production Economics 53 (3), 323–330.

aylor, J.W., 2008. A comparison of univariate time series methods for forecastingintraday arrivals at a call center. Management Science 54 (2), 253–265.

anagement 29 (2011) 65–77 77

Thompson, J.D., 1967. Organizations in Action: Social Science Bases of AdministrativeTheory. McGraw-Hill, New York, NY.

Tushman, M.L., Nadler, D.A., 1978. Information processing as an integrating conceptin organizational design. Academy of Management Review 3 (3), 613–624.

Vachon, S., Klassen, R.D., 2002. An exploratory investigation of the effects of supplychain complexity on delivery performance. IEEE Transactions on EngineeringManagement 49 (3), 218–230.

Vandaele, N., Van Nieuwenhuyse, I., Claerhout, D., Cremmery, R., 2008. Load-basedPOLCA: an integrated material control system for multiproduct, multima-chine job shops. Manufacturing Service Operations Management 10 (2),181–197.

Venkatesan, R., 1990. Cummins engine flexes its factory. Harvard Business Review68 (2), 120–127.

Venkatraman, N., 1989. The concept of fit in strategy research: toward ver-bal and statistical correspondence. Academy of Management Review 14 (3),423–444.

Vollmann, T.E., 1986. OPT as an enhancement to MRP II. Production and InventoryManagement Journal 27 (2), 38–49.

Vollmann, T.E., Berry, W.L., Whybark, D.C., Jacobs, F.R., 2005. Manufacturing Planningand Control for Supply Chain Management, 5th ed. McGraw-Hill Irwin, NewYork, NY.

Wacker, J.G., Sheu, C., 2006. Effectiveness of manufacturing planning and control sys-tems on manufacturing competitiveness: evidence from global manufacturingdata. International Journal of Production Research 44 (5), 1015–1036.

Wiers, V.C.S., 1997. A review of the applicability of OR and AI scheduling techniquesin practice. Omega – The International Journal of Management Science 25 (2),145–153.

Woodward, J., 1965. Industrial Organization: Theory and Practice. Oxford UniversityPress, London, UK.

Worthen, B., 2003. Future results not guaranteed; contrary to what vendors tell you,computer systems alone are incapable of producing accurate forecasts. CIO 16(19), 1.

Wortmann, J.C., Euwe, M.J., Taal, M., Wiers, V.C.S., 1996. A review of capacity planningtechniques within standard software packages. Production Planning & Control7 (2), 117–128.

Yang, K.-K., Webster, S., Ruben, R.A., 2002. An evaluation of flexible workday policiesin job shops. Decision Sciences 33 (2), 223–249.

Yeo, K.T., Ning, J.H., 2002. Integrating supply chain and critical chain concepts inengineer-procure-construct (EPC) projects. International Journal of Project Man-

agement 20 (4), 253–262.

Yusuf, Y.Y., Little, D., 1998. An empirical investigation of enterprise-wide integrationof MRPII. International Journal of Operations & Production Management 18 (1),66–86.

Zwikael, O., Sadeh, A., 2007. Planning effort as an effective risk management tool.Journal of Operations Management 25 (4), 755–767.