si

tie 1

ds

s m

ul

ser

y-b

esu

gs

ion

1. Introduction

ying redvanta997; Wproduommofferingenessxposeslt chain spe

acknowledged in the literature, to the best of our knowledge, none

of the earlier studies have disclosed what factors determine the

s areeci-udiesthreeas in

literature on the management of change orders. Then we present

Contents lists available at SciVerse ScienceDirect

w.e

Int. J. Productio

Int. J. Production Economics 135 (2012) 420429research and practical implementations.1 Tel.: 358 50 300 5750.research questions that aim to address the gaps identied in theliterature. This is followed by a description of our cost analysismethodology and research design. Finally, we present the resultsof the analyses and conclude with propositions for further

0925-5273/$ - see front matter & 2011 Elsevier B.V. All rights reserved.

doi:10.1016/j.ijpe.2011.08.016

n Corresponding author. Tel.: 34 91 568 9600.E-mail addresses: jukka.uskonen@aalto. (J. Uskonen),

antti.tenhiala@ie.edu (A. Tenhiala).advance, and in most cases, it is difcult to charge the customersfor these costs (Riley et al., 2005). Although this dilemma has been

other studies in different kinds of manufacturing environments.The paper is structured as follows: rst, we review the existingspecications poses a dilemma for many make-to-order (MTO)manufacturers. Most companies consider it to be an important partof their customer service and a natural extension of their pursuit ofresponsiveness (Danese and Romano, 2004). However, changeorders are known to incur costs that are very hard to estimate in

The ndings are discussed in general terms, and the costmeasured in relative values in order to avoid the context-spcity of the results, which is often a concern in single-case st(Yin, 2003). The conclusions of this article are formulated asgeneralized propositions that can be tested in practice as wellfulllment of their orders (Partanen and Haapasalo, 2004). Suchrequests are understandable because customer requirements mayvary over time, and there is always some lead time in whichchanges can occur when products are made or assembled to order.

The question of whether to accept mid-process changes in order

triangulation between different kinds of data (Jick, 1979); qualita-tive data such as interviews, process mapping, and work observa-tions were necessary in identifying the cost elements. Meanwhile,objective data from the companys information systems werenecessary to quantify and model the costs of individual changes.Responsiveness to customers vara crucial source of competitive amanufacturing companies (Fisher, 1to increased demand for customizeddeveloped tools and practices to accrequirements into their product o2002). While this kind of responsivto charge premium prices, it also euncertainties. A particularly difcucustomers who request changesquirements has becomege for many modernong et al., 2006). Due

cts, manufacturers havedate customers uniques (Forza and Salvador,enables manufacturersthem to new kinds ofllenge is coping withcications during the

costs of change orders, or how those costs could be best mitigated.Studies in the construction industry have shown that the overallcosts are often considerable (OBrien, 1997), but detailed analysesin the manufacturing industries have been lacking.

This article aims to develop an understanding of the behavior ofchange orders costs by identifying their constituents and accrualmechanisms. For that purpose, this study employs the methods ofactivity-based costing (ABC) and a multisource dataset from an in-depth case study. The case company is a mid-size machinerymanufacturer that produces customized refrigeration cabinets forgrocery retailers. The explorative analyses of this study necessitateThe price of responsiveness: Cost analyin make-to-order manufacturing

Jukka Uskonen a,1, Antti Tenhiala b,n

a Aalto University, Department of Industrial Engineering and Management, Otaniemenb IE Business School, Calle de Mara de Molina 12-5, 28006 Madrid, Spain

a r t i c l e i n f o

Article history:

Received 9 September 2008

Accepted 20 August 2011Available online 26 August 2011

Keywords:

Customization

Uncertainty

Activity-based costing

a b s t r a c t

The ability to produce goo

advantage, but it expose

specications during the f

in the name of customer

empirical data and activit

change orders costs. The r

opportunities for cost savin

improvements in informat

journal homepage: wws of change orders

7, 02150 Espoo, Finland

according to customers specications may be an important competitive

anufacturers to the risk of customers requesting changes in their

llment of their orders. Manufacturers often accept these change orders

vice despite the fact that they incur additional costs. This study uses

ased costing to explore the real values and the accrual mechanisms of

lts show that the total costs are considerable, but the analyses also reveal

through the categorization of change orders, time fencing techniques, and

processing.

& 2011 Elsevier B.V. All rights reserved.

lsevier.com/locate/ijpe

n Economics

the lead time of order fulllment so that changes have less time to

we designed an empirical inquiry to shed light on how the costs

J. Uskonen, A. Tenhiala / Int. J. Production Economics 135 (2012) 420429 4212. Literature review

2.1. Responsiveness as an ability to adopt change orders

The concept of responsiveness has a number of different mean-ings in the literature of manufacturing management. In additionto the capability of satisfying varying customer requirements(Fisher, 1997; Wong et al., 2006), the concept has also beendened as the speed of fullling customized orders (McCutcheonet al., 1994; Salvador and Forza, 2004). It has also been viewed inmore general terms as the overall capability to seize businessopportunities (Kritchanchai and MacCarthy, 1999) and has beendiscussed in more specic terms as the ability to swiftly conrmthe specications and delivery dates of customers orders(Pibernik, 2005). Yet another stream of research has focused ondening the different constituents of responsiveness (Holweg,2005; Reichhart and Holweg, 2007). While all of the varyingperspectives have their merits, in this study, responsiveness willbe considered as the capability to satisfy wide-ranging andfrequently changing customer requirements. The acid test of thatkind of responsiveness is the ability to adopt customers changerequests that occur after the initial entry of their orders.

2.2. Importance of adopting change orders

Most MTO manufacturers consider change orders to be aninherent part of their business. Manufacturers typically sym-pathize with the fact that the conditions in customers ownenvironments may change during the time that is needed tofulll the orders (Danese and Romano, 2004). This is especiallythe case in the manufacturing of capital goods. For example, inthe production of construction materials, the turbulent conditionsof construction sites translate to frequent changes in customersplans and schedules (OBrien, 1997). Such changes often necessi-tate modications to the technical specications and the deliverydates of the manufacturers products (Vrijhoef and Koskela, 2000).

The difculty in preventing change orders from occurringarises from the fact that it is often hard to determine whetherthe changes result from customers behavior or from the manu-facturer itself. Manufacturers usually acknowledge the difcultyof eliciting customers real needs at the time of the initial orderentry (Huffman and Kahn, 1998). Miscommunications or mistakesoccurring at that point may result in a need for modication atsome later phase in the order fulllment process (Hegde et al.,2005). It is normally preferable to execute such modicationsthan to deliver products that are not suitable for the customersintended use. Furthermore, many of the customer-originatedchanges are quite similar to the ordinary engineering changeorders, which the manufacturers face anyway (Tavcar andDuhovnik, 2005).

One particularly good motivation to accept change orders isgiven in a study by Hendricks and Singhal (2003). Its results showthat if disagreements about the contents of customers orders hitthe news, then the market values of publicly traded manufactur-ing companies will immediately drop by about 13%. Moreover,this plunge in the share value is accompanied with a considerablereduction of sales (Hendricks and Singhal, 2005). Hence, it is oftenthe safest choice to try to accommodate customers changerequests as soon as they are received.

2.3. Costs of adopting change orders

While the importance of adopting change orders appears to bewidely acknowledged, the costs of the modications have seldombeen studied in detail. In the classic model of the hidden factory,

the change transactions are counted as non-value-addingof change orders are composed and how they behave over time.This objective is served by seeking answers to the followingresearch questions:

RQ1: What are the critical factors that determine the costs ofchange orders?

RQ2: How does each critical factor inuence the costs ofchange orders?

RQ3: What are the annual total costs of being responsive tocustomers change orders?

RQ4: How the costs of change orders could be reduced withoutdeteriorating the responsiveness perceived by the customers?

4. Activity-based costing of change orders

4.1. Consumption of resources by activities

In this study, the method of analyzing the costs of changeorders follows the principles of activity-based costing (ABC). TheABC methodology was originally developed to replace traditionalcosting systems, whose ability to deal with indirect costs wereclaimed to be inadequate (Cooper, 1987, 1988). As the nameimplies, the logic of ABC is focused on activities instead of the costcenters, which are the main components in the traditional costingsystems. When building an ABC model, the rst step is to identifythe activities that consume the resources of the organization(Kaplan and Cooper, 1998). The most typical activities in manu-facturing organizations are the operations of the production,sales, and procurement processes. The most typical resourcesinclude materials, labor, support functions, and all necessaryoccur (Partanen and Haapasalo, 2004). Effective solutions includeat least the following: the use of standardized components,modular products, and product platforms (Salvador and Forza,2004; Hoover et al., 2001). In customized production, thesesolutions enable the postponement of product differentiation,which reduces order fulllment lead times (Krajewski et al.,2005), and also gives customers more time to nalize theirspecications, thus reducing the need for subsequent changeorders (Forza et al., 2008). In order to evaluate the protabilityof investments in these solutions, it would be valuable to have aclear understanding about the real nature of change orders costs.

3. Research questions

Due to the paucity of research on the costs of change orders,activities, and they are estimated to contribute up to 40% ofmanufacturers overhead costs (Miller and Vollmann, 1985). Morerecent gures can be found from studies in construction industry,in which cost and time overruns are relatively common, and one oftheir main causes are mid-process changes in the project plans(Williams et al., 2003). The primary drivers of these unexpectedcosts are the additional activities and the overtime work that arenecessary to execute the changes (Hanna et al., 2004). The totalcosts of the change orders vary from 5% to 15%, depending on thesize and the type of the project (Riley et al., 2005). The overallannual costs of change orders in the U.S. construction industryhave been estimated as 1326 billion dollars (Gunduz and Hanna,2005).

Despite the fact that only rough gures about the costs ofchange orders have been offered in the literature, many compa-nies have made signicant investments to reduce the amount ofchanges. The most intuitive response for that purpose is to reduceoverhead, such as the heating and lighting of facilities.

J. Uskonen, A. Tenhiala / Int. J. Production Economics 135 (2012) 4204294225. Data and methods

5.1. Research design

This research was conducted as an in-depth case study.Although single-case studies are sometimes considered proble-matic in terms of generalizability, the research design has someclear advantages. For the purposes of this investigation, thesingle-case approach was particularly suitable because the costanalyses necessitate detailed corporate data that would be dif-cult to acquire in a multiple-case design. The downsides of themethod are not that pressing because the topic is of exploratorynature (Glaser and Strauss, 1967). Thus, it is not necessary toprovide exact generalizations, but rather to identify the mostimportant factors and describe how they are related to oneanother (Handeld and Melnyk, 1998). Such tentative resultscan be elaborated in further studies. Following the inductive logicof the relationship-mapping research, we will try to facilitate thefurther studies by formulating our main ndings as generalizablepropositions (Eisenhardt, 1989).

The possibility of triangulating between different sources ofdata is an important advantage of case studies. In this investiga-tion, process mapping, interviews, work observations, and processdata were used to identify the main elements of the ABC model.Later, we conducted additional interviews and arranged cross-functional workshops to validate the ABC model and the resultsthat it yielded.

5.2. Case description

The case company is a typical MTO manufacturer in themachinery manufacturing industry. It produces refrigerationWhen the ABC methods are applied to the costs of changeorders, the activities refer to the work that is needed to executethe requested changes in the original specications. Such activ-ities may include revising production plans, rearranging workorders, reworking on some components, reassembling products,and reordering or expediting some externally procured rawmaterials. The resources that are consumed by the changesinclude materials and the labor hours of production planners,buyers, and shop-oor personnel.

4.2. Utilization of activities by cost objects

In the logic of the ABC methodology, activities consumeresources. Similarly, the goods and services that are being producedutilize the activities. They are the nal receivers of all costs andthus they are called the cost objects. The allocation of costs to thecost objects is done in two steps: rst, resource drivers quantifyhow much each activity consumes resources; then, activity driversdene how much each cost object utilizes different activities.

In normal ABC systems, the cost objects are the products,services, or customers of an organization. In this study, theconcept is extended to cover change orders. The ABC methodol-ogy is particularly suitable for the costing of change ordersbecause the methodology was originally developed to improvethe understanding of indirect costs (Cooper, 1987, 1988). Asdiscussed in the literature review, the earlier research consideredchange orders as a source of indirect costs and thus as a part ofthe hidden factory whose costs are difcult to analyze andmanage (Miller and Vollmann, 1985). Once the hidden costs arerevealed, it is possible to start managing and reducing them in asystematic manner (Kaplan and Cooper, 1998; Turney, 1996).machinery and remotely refrigerated display cabinets for groceryretailers. Customization is needed because grocery stores andsupermarkets vary greatly in their dimensions, layouts, andinterior design. The products must also be painted and accessor-ized according to the specications of the retail chains. Moreover,different materials and refrigerants must be used in differentcountries based on the local regulations.

5.3. Development of the ABC model

The rst step of the analysis was to map the change ordermanagement process and develop an ABC model of executing ageneric change order. The process was as follows: rst, a salesperson receives a change request from a customer. If the requestedchange relates to customized features or technical specications ofthe product, the sales person rst consults a product designerabout the feasibility of the request. Once the feasibility is ensured,the subsequent treatment depends on the amount of time remain-ing to the nal assembly of the product. This is because thecompany tried to enforce a frozen period of four weeks in theirmaster schedules. The period was based on the normal lead timesof order-specic raw material purchases (one to three weeks) andthe throughput time of the internal production activities (one weekon average). Thus, the changes that occurred within this timeframe possibly necessitated changes in the procurement andproduction plans. Such requests were formally documented aschange orders and passed on to the production planners. If thechanged customer order was already very close to its completion,the sales person would need a permission of the plant manager toissue a change order. On the other hand, if the request is related toa customer order outside the frozen period, the sales person wouldsimply change the contents of the order and no change orderwould be needed to call the attention of the production planners.

When the production planners get the change order, they mustnd out the status of the changed customer order and assess theimplications of revising the production schedules accordingly.When doing this, they need to consult the supervisors of differentparts of the production process. At minimum, they must informeveryone to halt the progress of the changed order until they havegured out how to execute the change. Typically the plannersneed to consult the warehouse personnel about the availability ofbasic raw materials and the buyers about the status of possibleorder-specic raw materials. On the basis of these inquiries, theproduction planners decide on what would be the least disruptiveway of executing the change. After reaching a decision, theydistribute the revised plans to the relevant personnel in theproduction, warehousing, and purchasing functions.

Finally, the personnel of the shop oor, warehouse, and purchas-ing must react to the revised production plans. The activities of theshop-oor personnel depend on where they work in the productionprocess. The relevant work centers in this process are painting,evaporator assembly, polyurethane casting, automatics assembly,nal assembly, and accessorizing. Painting must be completedbefore casting, whereas casting needs to be nished together withthe evaporator and the automatics, before the nal assembly andthe subsequent accessorizing may begin. The warehouse activitiesrelate to moving and storing goods, while the buyers activitiesconsist of revising procurement plans and purchase orders as well asnegotiating expedited deliveries from the suppliers.

After mapping the change order management process, weidentied the labor, material, and inventory resources that wereconsumed by the activities. Next, we determined how eachactivity consumed the resources (i.e., the resource drivers). Lastly,we determined how different kinds of change orders triggereddifferent activities (i.e., the activity drivers). This necessitatedidentifying the critical factors that differentiated change orders in

terms of their costs and thus provided the answer to Research

chan

J. Uskonen, A. Tenhiala / Int. J. Production Economics 135 (2012) 420429 423Question 1. Together, the elements constituted the ABC model, assummarized in Fig. 1.

The identication of the activities, resources, activity drivers,and the cost determinants was based on 30 days of workobservation, 16 interviews, and a number of informal discussionswith people from different business functions. The resourcedrivers were dened by analyzing data from the enterpriseresource planning (ERP) system of the case company. The datarecords of interest included standard costs of materials, bills ofmaterials (BOMs), and routings. The unit costs of standard andovertime labor as well as the inventory holding costs wereacquired from the nance department.

Fig. 1. ABC model of executing5.4. Implementation and validation of the ABC model

In order to use the ABC model for cost calculations, we createda large spreadsheet database for the data on the resources,activities, and resource drivers. One of the spreadsheets servedas an interface in which the user could input values for the costdeterminants identied in the rst step of the analysis. The ruleson how different combinations of determinants triggered differ-ent activities (i.e., the activity drivers) were stored as formulaeand macros. This tool enabled us to calculate the cost of any givenchange order and thus gave the answer to Research Question 2.During this phase of the study, we also carried out 12 additionalinterviews to ensure the completeness and validity of the model.

5.5. Using the ABC model with real data

To test the ABC model, we entered a large amount of realchange order data into the user interface of the spreadsheetdatabase. We used the ERP system of the case company to collectthe information about all changes to customer orders that hadoccurred during the year 2006. The sum of these calculations gavethe answer to Research Question 3. The overall sum and theresults on the costs of individual changes were discussed in threeworkshops with the personnel of the case company. In the rsttwo workshops, we presented the results to altogether 12 middlemanagers, sales persons, and production planners who all hadexperience in processing change orders and thus had good sense5.6. Simulations with the ABC model

Similarly as the ABC model enabled calculating the costs ofchanges that had already occurred, it enabled calculating costs ofimaginary change orders. One could, for example, take the informa-tion on any change request received from a customer and enter it tofor how much the costs could approximately be. On the basis oftheir experience, the workshop participants conrmed that theresults were realistic. In the third workshop, we presented thendings to the executives of the company. This workshop focusedon the policy changes based on the results.

ge orders in the case company.the user interface of the spreadsheet database. The model wouldthen calculate the costs of accepting that change. This way themodel could be put into everyday use at the company. However, italso served our research interest because we were able to use themodel to simulate how the total costs derived in the previous step ofthe analysis would have changed if the changes of the year 2006 hadbeen a little different. We focused the simulations on the effects ofwhen the change orders were received by the production planners.Using the spreadsheet database, we could easily calculate how thetotal costs would have changed, had the change orders arrivedearlier than what they did in reality. We rerun the total costanalyses with datasets that were otherwise the same as in theprevious step but the date of receipt was one to ve days earlier.These simulations gave the answer to Research Question 4.

6. Results

6.1. RQ1: determinants of change orders costs

The determinants of the change orders costs were identied inthe rst step of the analysis. All detailed results and absolute costsare omitted from this article due to their condentiality. How-ever, the most important ndings can be discussed in generalterms and elaborated with values that are relative to the standardcosts of the products. As for the drivers of the change orders

costs, three main factors kept arising in the interviews:

The type of the change. The product that is subject to the change. The time when production planners receive the change order.

The type of the change refers to the nature of the modicationthat the customer has requested. Although the possibilities fordifferent modications appeared to be almost innite at rst, itwas eventually possible to categorize them in ten generic types.The categories are listed in Table 1. Some of the changes affecthow the product is built and thus necessitate changes in both thework content (W) and the production schedules (S). Meanwhile,some of the changes are related only to the delivery dates anditem quantities and thus only affect the production schedules (S).

The second determinant of the costs is the type of the productthat is being changed. In the studied case, there were seven clearlydistinguishable product groups with different cost implications.The differences were negatively related to the average standardcosts of the product groups. This was because many of the changeactivities consumed a lot of labor resources, which had fairlyconstant costs. Consequently, the relative cost effects tended tobe higher for the less expensive products. The products in thelower end of the companys portfolio were called multi-shelverefrigerators (Multi-Shelve 1, Multi-Shelve 2, and Multi-Shelve 3).They were basic coolers with congurable shelves. The higher endof the portfolio consisted of serve-over cabinets (Serve-Over),freezer cases (Case), glass door cabinets (Glass-Door), and combi-nation cabinets (Combi). Those product groups included morecustomizable options than the lower-end products. Thus, theywere more prone to mid-process modications but the relativecosts of the changes were generally lesser. This was because

the labor intensity of their production made their standard costshigher. Also, the customizable options contained materials andparts that were relatively expensive but often packaged in modularcomponents and thus quicker to change.

The third determinant of the costs represents the time dimen-sion of the changes. It is calculated as the number of daysremaining until the planned completion of the customer orderat the time when the change order is received by the productionplanners. This is, of course, relevant because the necessaryactivities depend on how far the procurement and productionprocesses have reached at the time of receiving the change order.

6.2. RQ2: dynamics of the change orders costs

J. Uskonen, A. Tenhiala / Int. J. Production Economics 135 (2012) 420429424Table 1Different types of changes in the manufacturing of refrigeration cabinets.

Type of change Description Impactn

Colorexterior The exterior color is changed W&S

Colorinterior The interior color is changed W&S

Itemadded A product is added to the order S

Itemremoved A product is removed from the order S

Delivery

dateearlier

The delivery date is brought earlier S

Delivery datelater The delivery date is postponed S

Dimensionsheight The height is changed (higher or lower) W&S

Dimensionslength The length is changed (longer or shorter) W&S

Refrigerant The cooling uids are changed W&S

Controls The design of the automatic controller is

altered

W&S

n W: work content (materials, technical specications, and work instructions);

S: production schedules.Fig. 2. Average costs of diIn the second step of the analysis, we used the ABC model tocalculate the costs of different kinds of change orders. The resultsdemonstrated how the three different cost determinants inu-enced the costs of any change order. We observed considerabledifferences between the different types of changes. The costs rangedfrom nearly nothing to over 110% of the products standard costs.Fig. 2 shows a common pattern in making different types ofchanges to a given customer order at a given point of time.

The costs differ between the types of changes because thescope of the additional activities varies considerably. For example,changes in the controlling logic or in the refrigerant affect onlythe installation of the controllers and the evaporators. Meanwhile,changes in the colors or the dimensions of the product may causeadditional work in a number of different work centers. They alsoinvolve relatively expensive resources of the painting, polyur-ethane casting, and nal assembly work centers. Another expen-sive type of change, the addition of a completely new item,necessitates additional work in all work centers.

Fig. 3 shows the cost differences in the types of changes for allof the product groups. The relative differences between thedifferent types of products are quite small. The only majordifference is trivial: the changes in the interior colors are notpossible for the products that have standardized interiors. Other-wise the differences result from standard costs, as the relativecosts are higher for the cheaper products, as well as from productmodularity; the costs are higher for the less modular products.

In Figs. 2 and 3, the time factor of the costs has been xed. Ifthe timing of the change orders is allowed to vary in the costingmodel, then the result is rather intuitive: the costs are normallyhigher the later the modications are made. Fig. 4 illustrates thetime-based behavior of costs in the case of an example product.

There are thresholds in the accumulation of the costs because themodications get more expensive as more work centers have beguntheir operations. Another reason for the increasing time-basedcosts is the suppliers surcharges for expedited deliveries. Theynaturally increase as the requested lead times get shorter.fferent change types.

J. Uskonen, A. Tenhiala / Int. J. Production Economics 135 (2012) 420429 425In certain situations, however, the costs of the changes may decreasewhen the original delivery date draws closer. The costs of postpon-ing an order decrease when all of the production phases have beennished. This is because the production planners and the shop-oorpersonnel do not have to take any action at that point, and thusthere is no additional consumption of labor resources. The additionalexpenses are only manifested as inventory costs. In these cases, it isbetter not to react to the changes at all. Thus, it is critical that therelevant decision makers are aware of this phenomenon and refrainfrom passing the information about the changes to the shop oor. Ifthe shop-oor personnel see from their information systems that anitem is postponed or deleted, they will instinctively adapt to thechange, and higher coordination costs will ensue.

6.3. RQ3: total costs of change orders

The total costs of the change orders were calculated on the basis ofthe records from the ERP system. The data consisted of 1696 changeorders that were received and executed during the year 2006.

Fig. 3. Costs of different chan

Fig. 4. Example on the behaThe records contained information about the type of the change,the product that was subject to the change, and the relative time ofreceiving the change order (i.e., the amount of days before the duedate of the order). This information made it possible to enter eachchange order into the user interface of the ABC model. The modelthen calculated the costs, which we summed up to get the totalannual cost. Table 2 presents the result and its distribution among thedifferent types of changes. Although the total cost represents less thanone percent of the case companys revenues, the managers whoparticipated in our workshops thought that it was quite signicantconsidering the relatively low margins of the majority of theircustomer orders.

The costs are graphed cumulatively in Fig. 5. It shows that thecosts of change orders follow a Pareto distribution: 7% of thechange orders caused half of the total costs in 2006. Thesechanges were mainly modications to product dimensions, addeditems, and advancements of delivery dates. Meanwhile, postpone-ments of delivery dates were the most frequent modications(68%), but they only caused 35% of the costs.

ges in different products.

vior of costs over time.

6.4. RQ4: potential savings in change order costs

The ABC model and the data from the ERP system made itpossible to simulate how the costs would behave if the costdeterminants were altered from what they actually were in theyear 2006. This makes sense because not all of the determinantsare necessarily xed in reality. Although the type of the changeand the product that is subject to the change are typically givenby the customer, the manufacturer may well inuence the timingof the change orders. That is because the time when the produc-tion planners receive the change order depends greatly on howearly customers changed needs are recognized in the companysfront end and how swiftly they are processed thereafter. Conse-quently, we focused the analysis of the potential savings on thetiming of when the production planners receive the changeorders. We did this by recalculating the total annual costs withdatasets that were otherwise the same as the data of year 2006

but where each change order was manipulated to have arrivedone to ve days earlier. Fig. 6 shows the results.

The results show that a reduction of only one day in theprocessing of the change orders would have decreased the totalcosts by 10%. This reduction was considered realistic and achiev-able by the personnel of the case company because they knewthat the processing of the change requests often took several daysin the front end of the company. In fact, the personnel thoughtthat it would be possible to reduce at least a couple of days fromthe administrative process.

6.5. Summary of results and implications for the case company

In summary, the analyses disclosed (1) the main factors thatinuence the costs of change orders, (2) the ways in which thosefactors drive the costs, (3) the total costs that can be attributed tothe change orders, and (4) the importance of fast processing ofchange orders if the total costs are to be reduced. Apart fromanswering to our research questions, the ndings triggeredseveral policy changes at the case company.

First, improving the management of change orders was made apriority in the overhaul of the companys ERP system, which wasscheduled to take place in the following two years. In the revisedERP system conguration, the change order management processwas to be formalized and streamlined in order to reap the costsavings of the reduced processing times. Also, the vendor of theERP system was inspired to develop automated procedures forchange order management. Second, the ndings were used to

Table 2Total costs of change orders in 2006.

Type of changea Number ofchanges

Share of allchanges (%)

Totalcost (eur)

Share of allcosts (%)

Colorexterior 33 2 5893 3

Colorinterior 71 4 13004 6

Item(s)added 106 6 38913 17

Item(s)removed 177 10 1806 o1Delivery

dateearlier

48 3 25138 11

Delivery Datelater 1160 68 80442 35

Dimensionsheight 8 o1 11531 5Dimensionslength 35 2 53949 23

Refrigerant 58 3 600 o1

Total 1696 100 231276 100

a Data were not available about the change orders for the automatic con-

trollers.

of al

J. Uskonen, A. Tenhiala / Int. J. Production Economics 135 (2012) 420429426Fig. 5. Cumulative costsFig. 6. Simulated savings from swimotivate changes in the product portfolio. All existing initiativesto increase product modularity, which were initially consideredvery expensive, got support from the estimated savings in thechange order costs of modular products. Consequently, manychanges were implemented to minimize the proportion of non-modular products in the companys portfolio. Third, the costanalysis also gave additional input to the cost-benet analyses

l change orders in 2006.fter handling of change orders.

is that the more modular the production processes are the less

J. Uskonen, A. Tenhiala / Int. J. Production Economics 135 (2012) 420429 427costly it is to execute changes to the technical specications andfeatures of the products. That is because in a modular process,individual featureswhich are potential subjects for changeordersare added to products in their entirety in specic partsof the process (e.g., Pine, 1993; Baldwin and Clark, 2000; Tu et al.,2004). This effectively reduces the amount of resources that areinuenced by the change orders in comparison to non-modularprocesses where individual features may be worked on in manyof several suggested process reengineering projects that wereaimed to reduce the throughput time of the manufacturingoperations. Also these initiatives were generally very expensiveand thus deemed very risky in the absence of reliable estimatesabout their benets. Not only did the cost analysis provide suchestimates, but the information on how the costs depend on thetiming, the product type, and the type of the change helped toprioritize between different initiatives.

7. Discussion

This article reports a preliminary exploration to the costs ofchange orders. The results are not suggested to apply as such toother organizations. Instead, they serve as an example of how therelative costs of change orders are composed and how they behavein one mid-size machinery manufacturing company. The resultsalso demonstrate that the ABC methodology can be successfullyused to decompose the costs of change orders so that theirpractical implications can be better understood and managed. Inthe following sections, we will summarize the most importantndings as propositions that can be tested in further studies.

7.1. Determinants of change order costs

First of all, the ndings indicate that the costs of change orderscan be large enough to deserve managerial attention. By identify-ing the most critical determinants of change orders costs,companies can improve their nancial performance. The assess-ment of the determinants can be done on the basis of the threefactors that were identied in this study: the time of the change,the type of the change, and the product that is being changed. Inthe following, we formulate propositions to summarize the effectof each of them.

Proposition 1a. The costs of change orders are nonlinearlyassociated with the timing of when the production plannersreceive them: the costs are generally the higher the later thechange orders are received, but due to the reduced need forcoordination activities, the costs begin to reduce for postpone-ments when the delivery date draws closer.

This proposition emphasizes the importance of discretion bythe sales personnel and production planners who are the rstpeople to deal with the change orders received from the custo-mers. When a change order is received very late, the coordinationcosts of executing it may exceed the costs of nishing a productthat is not needed as early as originally expected. Of course, onemust have conducted the kind of analysis presented in this paperin order to make the right decisions on when to execute and whennot to execute the change.

Proposition 1b. The costs of change orders are related to the typeof the change so that the costs are generally the higher the moreresources are involved in the execution of the change.

This proposition relates the ndings to the literature onprocess modularity. In essence, one corollary of the propositiondifferent parts of the process.Proposition 1c. The costs of change orders are related to theproduct that is subject to the change so that the costs are thelower the more modular the product is.

This proposition emphasizes the importance of product mod-ularity. This is not necessarily surprising but adds to the currentbody of knowledge in at least two ways. First, there has beensome disagreement whether product modularity decreases orincreases costs in manufacturing (Agrawal et al., 2001; Zipkin,2001). Our ndings support the former view. Second, the earlierresearchers, who have found support for the cost advantages,have identied certain mechanisms through which the savingscan occur (Jacobs et al., 2007), but they have not identied thereduced change order costs as one of these mechanisms.

7.2. Cost categories of change orders

The second observation of general nature is that the types ofchange orders can be grouped on the basis of their cost effects.According to the total cost analysis, 7% of the changes made upabout 50% of the costs in the case company. Analyzing the costs ofthe different types of changes can help in categorizing betweencostly and inexpensive change orders. Such a division may enablethe creation of specic time fences for the different types ofchange orders. Such a practice would enable more efcient butstill mainly responsive operations. Therefore, we formulate thefollowing proposition:

Proposition 2. Enforcing a common frozen period for all changeorders reduces equally the total costs of change orders and thecompanys perceived responsiveness to change orders, but usingdifferent frozen periods for different types of change ordersenables proportionally larger reductions in the total costs thanin the perceived responsiveness.

This proposition contributes to the body of knowledge on timefencing practices. In the traditional practice, only one frozenperiod is applied to each end product regardless of what kind ofchanges are made to the production plans (e.g., Vollmann et al.,2005; Proud, 2007). Our result suggests that it could be benecialto have multiple different frozen periods for each product. In suchapproach, the changes that incur costs earlier would have thelongest frozen periods, whereas the changes that start incurringcosts only later in the fulllment process could have much shorterfrozen periods. In the case company, all changes to the dimen-sions of the product belong to the former group while the changesin the colors or in the refrigerant belong to the latter (see Fig. 4).

The idea of using multiple frozen periods for individual endproducts is not entirely new. For example, Yeung et al. (2003)have proposed that setting up different frozen periods for differ-ent BOM levels would help coping with demand variations.Similarly, Trentin et al. (2011) have proposed that using multiplefrozen periods would facilitate the postponement of productdifferentiation without costly reengineering interventions toproduction processes. Our proposition adds to these ndings bydescribing yet another way of using multiple frozen periods andby suggesting that it could help coping with change orders in amanner that is simultaneously responsive and cost efcient.Furthermore, our proposition applies to MTO manufacturingwhereas the earlier contributions have been made in the contextof make-to-stock manufacturing.

7.3. Value of time in the processing of change orders

The third main nding is related to the importance of time. Itis a crucial part of the equation because it is not necessarily as

rigidly xed as the type of the change and the changed product.

tions are supposed to hold. We propose that in this study, the

was no order fulllment lead time, then the propositions wouldnot be relevant. Also, it should be noticed that the tness to the

J. Uskonen, A. Tenhiala / Int. J. Production Economics 135 (2012) 420429428boundary conditions are dened by the existence of the followingfour factors: order-specic customer requirements, uncertaintyregarding the customer requirements, non-zero order fulllmentlead time, and the specialization of resources.

Without order-specic customer requirements there wouldobviously be no change orders from customers. Although therecould be other kinds of modications, such as engineering changeorders, the propositions would not hold as such. For example,when it comes to the engineering change orders, the speed ofprocessing is much less relevant than, for example, the timing ofthe changes. However, it should be noted that this boundarycondition does not necessitate the products to be customized.Instead, changes to the order-specic customer requirements canbe manifested merely as changes to the requested delivery dates.Such changes can occur in a wide variety of business environ-ments, and we suggest that our propositions will hold in thoseenvironments. For example, it could be benecial for a companyto establish different frozen periods for postponements andadvancements of delivery dates. Alternatively, different frozenperiods could be applied to the changes in the day of the deliveryand the hour of the delivery.

Another boundary condition is that there should be someuncertainty regarding the order-specic customer requirements.This is not always the case. If a company makes tailor-madeshoes, for instance, the likelihood of the preferred size to changeduring the order fulllment process is very low. This may alsoapply to the color, style, and the delivery date of the product.Meanwhile, in the operating environment of the studied com-Often only the time of receiving the initial change request fromthe customer is xed, and it is up to the manufacturers ownprocesses how much time is spent before a change order is issuedto the production planners. The process of administering changeorders can be sped up by recognizing potential modicationsearly and by facilitating a awless and fast information channel toensure swift and reliable messaging between the front-endpersonnel who receive the change orders and the productionplanners who execute them. This leads to the followingproposition:

Proposition 3. The faster is the processing of change orders thelower are the total costs of change orders.

This proposition complements the work of Trentin and Forza(2010), who have discussed a similar effect on delivery perfor-mance. Moreover, our results quantify the value of swift informa-tion processing. Specically, our simulations with the ABC modelshowed that, on average, one wasted day in the administration ofa change order increases its costs by approximately 10%. There-fore, the cost of slowness can be calculated for informationows just as it can be calculated for slow-moving materials.Earlier studies have demonstrated that this analogy betweenslow-moving information and slow-moving goods holds in math-ematical models and laboratory settings (e.g., Chen, 1999; Crosonand Donohue, 2006). Our results demonstrate the phenomenon inthe real world. The ABC model of this paper serves as an exampleof how the costs of slow information can be calculated in practice.

7.4. Boundaries of generalizability

One way to assess the extent and the limits of the externalvalidity of results from a single-case study is to identify thecritical factors that made the results possible (Dubin, 1978). As apart of the process of theoretical generalization (Yin, 2003), theseboundary conditions specify the domain within which the proposi-pany, both the specications and the delivery dates are highlyboundary conditions is a continuousnot a binaryattribute. Inother words, the question is not whether or not a given companyts within the conditions but to what extent the company tswithin them. For example, the longer the lead times are the morerelevant the propositions should be. Although the boundaryconditions are relatively tight and many contemporary businesstrends, such as form postponement, component commonality,and time-based competition, reduce the relevance of the proposi-tions, there are still many relevant industries where the condi-tions hold very well. They include at least all capital goodsmanufacturing, the construction industry in its entirety, and theproduction of luxury craft products.

8. Conclusions

Many make-to-order manufacturers strive to be responsive tocustomers changing requirements even when the requirementsare modied after the initial placement of the order. This studydemonstrated that the practice of accepting change orders mayincur signicant costs to manufacturing companies. However, theresults also showed that different kinds of changes may have verydifferent cost implications, and only a relatively small portion ofchanges can make up the majority of change orders total costs.Therefore, attention should be paid to the management of changeorders, and in order to do so, one should rst get an under-standing of how the costs of change orders behave. The ABCmethodology described in this article provides a tool for carryingout the necessary analyses.

The results from using the ABC model in the case companyshowed that change orders can be categorized according to theircost implications. Making such categorization may open oppor-tunities for cost savings without jeopardizing the overall respon-siveness of the company. Once the most costly changes have beenidentied, management efforts can be focused on reducing themor gaining control over them. The ways to mitigate the costs ofchange orders include establishing separate frozen periods fordifferent kinds of modications and imposing additional chargessensitive to the vast uncertainties of the construction sites towhich the products are typically ordered.

The third boundary condition is that there should be some leadtime between the specication of the customer requirements andthe delivery. This is not always the case either. For example, whenproduct differentiation takes place very late in the fulllmentprocess, the customer-specic requirements can be satisedsimply by adding accessories or by conguring the software ofthe product. Sometimes such customizations can be done at thepoint of sales, and thus the time for the customers to change theirminds is effectively eliminated.

Lastly, the resources used in the production must be specia-lized to some extent. The more a company utilizes commoncomponents, general-purpose machinery, and multi-skilled work-force, the less there is need to coordinate responses to changeorders. The reduced need for replanning production schedulesand expediting the purchases of order-specic raw materialsreduces the costs of the change orders and thus makes thepropositions less relevant.

We propose that the applicability of the results in otherenvironments than the studied company depends on how wellthe other environments t within the boundary conditions. It isnoteworthy that all of the conditions are critical. For example, ifin some environment, the other three conditions existed but therefor the most costly changes.

Finally, the simulations with the costing model demonstrated

performance: a typological theory. International Journal of Operations &

Hendricks, K.B., Singhal, V.R., 2003. The effect of supply chain glitches on share-holder wealth. Journal of Operations Management 21 (5), 501522.

Hoover, W.E., Eloranta, E., Holmstrom, J., Huttunen, K., 2001. Managing theDemand-Supply Chain: Value Innovations for Customer Satisfaction. Wiley,New York, NY.

Huffman, C., Kahn, B.E., 1998. Variety for sale: mass customization or massconfusion? Journal of Retailing 74 (4), 491513.

J. Uskonen, A. Tenhiala / Int. J. Production Economics 135 (2012) 420429 429Hendricks, K.B., Singhal, V.R., 2005. Association between supply chain glitches andoperating performance. Management Science 51 (5), 695711.

Holweg, M., 2005. The three dimensions of responsiveness. International Journal ofOperations & Production Management 25 (7), 603622.Production Management 28 (11), 10671094.Glaser, B.G., Strauss, A.L., 1967. The Discovery of Grounded Theory: Strategies for

Qualitative Research. Wiedenfeld and Nicolson, London, UK.Gunduz, M., Hanna, A.S., 2005. Benchmarking change order impacts on produc-

tivity for electrical and mechanical projects. Building and Environment 40 (8),10681075.

Handeld, R.B., Melnyk, S.A., 1998. The scientic theory-building process: a primerusing the case of TQM. Journal of Operations Management 16 (4), 321339.

Hanna, A.S., Camlic, R., Peterson, P.A., Lee, M.-J., 2004. Cumulative effect of projectchanges for electrical and mechanical construction. Journal of ConstructionEngineering and Management 130 (6), 762771.

Hegde, V.G., Kekre, S., Rajiv, S., Tadikamalla, P.R., 2005. Customization: impact onproduct and process performance. Production and Operations Management 14(4), 388399.be understood in companies that strive for responsiveness. If thepolicy of a company is to accept customers change orders, then allchange requests should always be processed as quickly as possible.

Acknowledgements

We thank the anonymous reviewers whose comments andsuggestions were very helpful and constructive. We are alsograteful to Eero Eloranta and Kari Tanskanen for their feedbackon the earlier drafts of the manuscript. We further gratefullyacknowledge the nancial support from Tekesthe Finnish Fund-ing Agency for Technology and Innovation. Finally, we thank ArtoKoskelainen and the other employees of the case company whomade this research possible in the rst place.

References

Agrawal, M., Kumaresh, T.V., Mercer, G.A., 2001. The false promise of masscustomization. McKinsey Quarterly 2001 (3), 6271.

Baldwin, C.Y., Clark, K.B., 2000. Design Rules: The Power of Modularity. The MITPress, Cambridge, MA.

Cooper, R., 1987. Does your company need a new costing system. Journal of CostManagement 1 (1), 4549.

Cooper, R., 1988. The rise of activity-based costingpart one: what is an activity-based costing system. Journal of Cost Management 2 (2), 4554.

Croson, R., Donohue, K., 2006. Behavioral causes of the bullwhip effect and theobserved value of inventory information. Management Science 52 (3),323336.

Chen, F., 1999. Decentralized supply chains subject to information delays.Management Science 45 (8), 10761090.

Danese, P., Romano, P., 2004. Improving inter-functional coordination to face highproduct variety and frequent modications. International Journal of Opera-tions & Production Management 24 (9), 863885.

Dubin, R. (Ed.), 1978. Theory Building, Revised ed., Free Press, New York, NY.Eisenhardt, K.M., 1989. Building theories from case study research. Academy of

Management Review 14 (4), 532550.Fisher, M.L., 1997. What is the right supply chain for your product? Harvard

Business Review 75 (2), 105116.Forza, C., Salvador, F., 2002. Managing for variety in the order acquisition and

fullment process: the contribution of product conguration systems. Inter-national Journal of Production Economics 76 (1), 8798.

Forza, C., Salvador, F., Trentin, A., 2008. Form postponement effects on operationalcompetitive performance: do integration strategies mediate the relationship?International Journal of Operations & Production Management 27 (10),10461068.

Jick, T.D., 1979. Mixing qualitative and quantitative methods: triangulation inaction. Administrative Science Quarterly 24 (4), 602611.

Kaplan, R.S., Cooper, R., 1998. Cost & Effect: Using Integrated Cost Systems to DriveProtability and Performance. Harvard Business School Press, Boston, MA.

Krajewski, L., Wei, J.C., Tang, L.L., 2005. Responding schedule changes in build-to-order supply chains. Journal of Operations Management 23 (5), 452469.

Kritchanchai, D., MacCarthy, B.L., 1999. Responsiveness of the order fullmentprocess. International Journal of Operations & Production Management 19 (8),812833.

McCutcheon, D.M., Raturi, A.S., Meredith, J.R., 1994. The customization-respon-siveness squeeze. Sloan Management Review 35 (2), 8999.

Miller, J.G., Vollmann, T.E., 1985. The hidden factory. Harvard Business Review 63(5), 142150.

OBrien, W.J., 1997. Construction supply-chains: case study, integrated cost andperformance analysis. In: Alarcon, L. (Ed.), Lean Construction. A.A. BalkemaPublishers, Rotterdam, Netherlands, pp. 187222.

Partanen, J., Haapasalo, H., 2004. Fast production for order fulllment: implement-ing mass customization in electronics industry. International Journal ofProduction Economics 90 (2), 213222.

Pibernik, R., 2005. Advanced available-to-promise: classication, selected methodsand requirements for operations and inventory management. InternationalJournal of Production Economics 93, 239252.

Pine II, B.J., 1993. Mass Customization: The New Frontier in Business Competition.Harvard Business School Press, Boston, MA..

Proud, J.F., 2007. Master Scheduling: A Practical Guide to Competitive Manufac-turing, 3rd. ed. John Wiley & Sons, Hoboken, NJ.

Reichhart, A., Holweg, M., 2007. Creating the customer-responsive supply chain: areconciliation of concepts. International Journal of Operations & ProductionManagement 27 (11), 11441172.

Riley, D.R., Diller, B.E., Kerr, D., 2005. Effects of delivery systems on change ordersize and frequency in mechanical construction. Journal of ConstructionEngineering and Management 131 (9), 953962.

Salvador, F., Forza, C., 2004. Conguring products to address the customization-responsiveness squeeze: a survey of management issues and opportunities.International Journal of Production Economics 91 (3), 273291.

Tavcar, J., Duhovnik, J., 2005. Engineering change management in individual andmass production. Robotics and Computer-Integrated Manufacturing 21 (3),205215.

Trentin, A., Forza, C., 2010. Design for form postponement: do not overlookorganization design. International Journal of Operations & Production Manage-ment 30 (4), 338364.

Trentin, A., Salvador, F., Forza, C., Rungtusanatham, M.J., 2011. Operationalisingform postponement from a decision-making perspective. International Journalof Production Research 49 (7), 19771999.

Tu, Q., Vonderembse, M.A., Ragu-Nathan, T.S., Ragu-Nathan, B., 2004. Measuringmodularity-based manufacturing practices and their impact on mass custo-mization capability: a customer-driven perspective. Decision Sciences 35 (2),147168.

Turney, P.B.B., 1996. Activity-Based Costing: The Performance Breakthrough.Kogan Page, London, UK.

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

Vrijhoef, R., Koskela, L., 2000. The four roles of supply chain management inconstruction. European Journal of Purchasing & Supply Management 6 (34),169178.

Williams, T., Ackermann, F., Eden, C., 2003. Structuring a delay and disruptionclaim: an application of cause-mapping and system dynamics. EuropeanJournal of Operational Research 148 (1), 192204.

Wong, C.Y., Arlbjrn, J.S., Hvolby, H.-H., Johansen, J., 2006. Assessing responsive-ness of a volatile and seasonal supply chain: a case study. International Journalof Production Economics 104 (2), 709721.

Yeung, J.H.Y., Ma, L., Law, J.S., 2003. The impact of using multiple freeze fences oncosts in MRP systems. International Journal of Production Research 41 (10),22012228.

Yin, R.K., 2003. Case Study Research: Design and Methods, 3rd ed. Sage Publica-tions, Thousand Oaks, CA.

Zipkin, P., 2001. The limits of mass customization. Sloan Management Review 42(3), 8187.by approximately 10%. The costs of slow information ows should Jacobs, M., Vickery, S.K., Droge, C., 2007. The effects of product modularity onthat swift information ows play a crucial role in the efcientprocessing of change orders. Wasting a single day in the adminis-tration of customers change requests appears to increase the costs

The price of responsiveness: Cost analysis of change orders in make-to-order manufacturingIntroductionLiterature reviewResponsiveness as an ability to adopt change ordersImportance of adopting change ordersCosts of adopting change orders

Research questionsActivity-based costing of change ordersConsumption of resources by activitiesUtilization of activities by cost objects

Data and methodsResearch designCase descriptionDevelopment of the ABC modelImplementation and validation of the ABC modelUsing the ABC model with real dataSimulations with the ABC model

ResultsRQ1: determinants of change orders' costsRQ2: dynamics of the change orders' costsRQ3: total costs of change ordersRQ4: potential savings in change order costsSummary of results and implications for the case company

DiscussionDeterminants of change order costsCost categories of change ordersValue of time in the processing of change ordersBoundaries of generalizability

ConclusionsAcknowledgementsReferences