Texturing, Spackling, and Jointing: Strategies for Helping Coordinate Product, Process, and Supply Chain Design

Transcription

1 Texturing, Spackling, an Jointing: Strategies for Helping Coorinate Prouct, Process, an Supply Chain Design Kyle Cattani The Kenan-Flagler Business School UNC Chapel Hill Chapel Hill, NC (919) Ely Dahan Anerson School at UCLA 110 Westwoo Plaza, B-514 Los Angeles, CA (310) Glen M. Schmit The McDonough School of Business Georgetown University Washington, DC (202)

2 Texturing, Spackling, an Jointing: Strategies for Helping Coorinate Prouct, Process, an Supply Chain Design Abstract We evelop a framework for etermining whether the prouction process shoul be esigne as make-to-stock (MTS) an/or make-to-orer (MTO), given the firm s assessment of market preferences for stanar prouct esigns an custom prouct esigns an the firm s prouction capabilities in terms of efficient versus flexible prouction. These process an prouct esign ecisions can in turn impact supply chain esign since stanar proucts might typically involve retail outlets, while custom proucts can often be sol irectly. If the firm chooses to offer only stanar proucts we suggest consiering a process strategy calle texturing, if it offers both stanar an custom esigns we suggest consiering a process strategy calle spackling, an if it offers custom proucts we suggest consiering a process strategy calle jointing. All three strategies involve both MTS an MTO. 1

3 Texturing, Spackling, an Jointing: Strategies for Helping Coorinate Prouct, Process, an Supply Chain Design 1. Introuction The avent of the Internet has heightene interest in the use of irect sales channels an in the esign an prouction of custom, make-to-orer (MTO) proucts. The Internet gives manufacturers irect access to customers an in turn gives customers irect access to information about proucts, creating a convenient means through which customers can configure an orer their preferre proucts. In some cases manufacturers foresee using this irect channel to complement the customer s alternative of buying a stanar make-to-stock (MTS) prouct from a traitional retailer, while in other cases the irect channel may fully isplace the traitional channel. In yet other situations, the irect channel may be of no value the manufacturer may choose to continue selling strictly through retail outlets to customers who prefer immeiate elivery an visual inspection of the prouct. Avances in prouction technologies also seem to be influencing methos by which proucts can be esigne, prouce, an istribute. For example, flexible manufacturing processes are becoming less costly an more capable. Supply chain management initiates are reucing lea-times an inventory requirements. An tracking technologies are facilitating the smooth an efficient flow of goos. Thus manufacturers are face with a changing lanscape creating a picture of both opportunity an challenge. The firm must make some very basic, but strategic, ecisions about its prouct esigns, process esigns, an supply chain esigns. Shoul the firm s prouct esigns inclue custom proucts, or stanar proucts, or both? Shoul the firm s processes be esigne as make-toorer, make-to-stock, or some combination of these alternatives? An shoul the supply chain 2

4 be esigne to eliver proucts irect, or though a traitional retail channel, or both? We begin to aress these relatively high-level ecisions in a very general framework. We focus on the issues of prouct esign type (stanar versus custom) an process esign configuration (MTO versus MTS), noting that irect supply chains seem quite amenable to the elivery of custom proucts while retail supply chains eal efficiently with stanar proucts where quick elivery an fast inventory turnover are paramount. The question of how best to match prouct esign with process esign is not a new one. Hayes an Wheelwright (1979a) evelope the notion of the prouct/process spectrum to aress a funamental trae-off that firms face, between efficiency an flexibility. Flow shops typically mass-prouce stanar items in a steay, make-to-stock (MTS) fashion using eicate resources to achieve efficient, low-cost prouction. However, a ownsie of flow-shop prouction is that output may not irectly match customer eman in a couple of significant ways. First, the stanar proucts may not exactly match each customer s nees. Secon, achieving a steay flow of efficient output is ifficult uner stochastic eman; the manufacturing firm likely will prouce too many items or too few an experience overage or unerage costs. At the other en of the spectrum from the flow shop is the job shop. A job shop can prouce customize make-to-orer (MTO) proucts using flexible resources an thereby exactly match both customer esign preferences an eman volume. A possible supply chain esign avantage is that the MTO manufacturer may be able to sell custom proucts irectly to en users more easily than a traitional retailer coul, an thus have a more streamline supply chain. The ownsie of custom MTO items is that these proucts generally require less efficient, but flexible job-shop type resources. Further, the customer may experience a elay cost 3

5 in not getting the prouct immeiately many stanar proucts are sol through retails outlets where the consumer takes immeiate elivery of the prouct rather than incurring any elay. The firm esires a process esign base on flow-shop efficiency but also sees possible avantages from prouct esigns that feature customizable MTO items an from streamline supply chains that sell irect. How can the firm best manage these trae-offs? The traitional answer to this management challenge has been to match the process esign strategy with the prouct esign strategy, as suggeste by Hayes an Wheelwright (1979b). That is, if the prouct esign is one of stanar proucts prouce in high volume, then the process esign shoul be base on efficient flow-shop resources proucing in MTS fashion (a process typically referre to as mass prouction), while if the esign strategy involves one-ofa-kin custom proucts then the process esign strategy shoul incorporate less efficient but flexible resources an a job-shop MTO environment. While recognizing that there are possible intermeiate strategies of batch flow an line flow, this traitional response aligns prouct esign for stanar proucts with an MTS mass prouction process esign, an prouct esign for custom proucts with a flexible MTO process esign. We consier the possibility of further refining the firm s response to this challenge by using hybri (i.e., ual) strategies that simultaneously take avantage of both MTS an MTO in non-traitional ways. In this paper we first evelop a marketing moel that etermines whether the firm might be best serve by offering custom proucts in aition to its stanar proucts, or only custom proucts, or only stanar items. We then present separate operations moels for each of these three cases, to tie the results of the marketing analysis to three possible ual MTS/MTO strategies: the first is applicable to the case where the market preference is only for stanar proucts (we call this a texturing strategy), the secon shoul be consiere when the 4

6 market preference is for both stanar an custom proucts (we call this a spackling strategy), an the thir might prove avantageous when the market preference is only for custom proucts (a jointing strategy). We show the conitions uner which each of these hybri strategies is preferre, as compare to more traitional strategies. Our framework is epicte in Figure 1, an is comprise of a simple two-stage analysis. The marketing moel first etermines which row of Figure 1 the firm fins itself. That is, it establishes whether the firm shoul offer only stanar proucts in retail channels, only customize proucts in a streamline supply chain involving a irect marketing channel, or a mix of both, an establishes expecte eman volumes. Figure 1. Strategy is Determine by Market Preferences an Prouction Capabilities Prouction Capabilities Premium for Flexible Prouction vs. Efficient Prouction High Low Market Preferences Stanar Proucts Both St. & Custom Proucts Custom Proucts Mass Prouction Make st. proucts via MTS using efficient capacity Focus Make st. pro. via MTS using eff. cap.; Make custom pro. via MTO using flex. cap. Postponement Start custom pro. as MTS using eff. cap., finish via MTO using flex. cap. (most steps the former) Texturing Make st. pro. via 2 resources; MTS using eff. cap. & MTO using flex. cap. Spackling Make st. pro. via MTS using flex. cap.; Make custom pro. via MTO using the same flex. cap. Avancement Start custom pro. as MTS using eff. cap., finish via MTO using flex. cap. (most steps the latter) Using this marketing information, along with information regaring eman variability, the firm then uses an operations moel to etermine whether it is locate in the left or right column of Figure 1. More specifically, the operations moel etermines the type, an level, of capacity to acquire, an etermines whether to operate in MTS or MTO fashion. (We use the 5

7 term MTS to mean that prouction procees in anticipation of eman, with output being ae to inventory, while MTO procees only after eman realization.) Capacity comes in two forms, efficient capacity, that can only prouce stanar proucts in MTS fashion, an flexible capacity, that can prouce either stanar or customize proucts, an can operate in either MTS or MTO fashion. That is, efficient capacity is associate with a relatively steay prouction scheule for a limite set of prouct configurations, where output is ae to inventory. Flexible capacity, on the other han, is associate with resources that can prouce a wier configuration of proucts on relatively short orer, within the capacity constraint, but at a higher cost. First, consier the case where the market eman is primarily for stanar proucts (the top row of Figure 1). Traitionally, a firm might view this as a situation calling for efficient capacity an MTS, i.e., mass prouction. 1 Inee, we fin that to be the case, assuming there is a significant cost avantage for such prouction. However, if the penalty associate with flexible capacity is less imposing, then the firm may want to consier an approach we call texturing. With the texturing strategy, the firm s primary prouction comes from efficient capacity couple with MTS prouction, but since MTS has the isavantage of not exactly matching eman in a given perio, the firm restricts its efficient capacity (its MTS output) to avoi builing too much stock prouct. Concurrently, it hols some separate flexible capacity use in MTO fashion only if it realizes eman in that perio exceeing MTS capacity. The result is that efficient capacity is utilize fully, while flexible capacity is use only if there is sufficient 1 Technically, the term mass prouction may be appropriate only if the items are sol in large volume. However, even if volumes are small, proucing stanar items via MTS can likely be one in a relatively more efficient manner than making such items via MTO. 6

8 eman. If this scenario is repeate in a series of inepenent an ientical perios, then varying levels of flexible capacity (MTO prouction) woul be use from perio to perio base on each perio s specific eman realization. This yiels a total prouction level that is bumpy (it varies from perio to perio), but that closely matches the bumpy eman (see Figure 2). Our analogy here is to a texture (bumpy) wall or ceiling surface. Note that with texturing, the firm effectively operates two ifferent prouction facilities, one equippe with efficient resources an the other with flexible resources. Figure 2. Texturing Yiels a Bumpy Stream of Stanar Proucts, from Two Facilities. Prouction Volume Total output closely tracks eman Stanar proucts mae via MTO using flexible capacity Stanar proucts mae via MTS using efficient capacity Time An example of texturing involves a well-known semiconuctor manufacturer. The firm prouces a steay volume of stanar computer chips in an efficient manufacturing fab, an supplements that capacity with flexible but expensive capacity that is generally use for R&D purposes. The R&D capacity is use for prouction purposes only if eman excees the fab capacity. See Cattani, et al. (2003a) for further iscussion of the texturing strategy. Next, consier the case where significant market eman exists for both stanar an customize proucts (the mile row of Figure 1). In this situation, a firm might traitionally establish a focuse strategy: one efficient facility (or prouction line) focuses on prouction of 7

9 stanar proucts manage in MTS fashion, an a separate flexible capacity focuses on custom proucts operating as MTO. We inee fin this to be optimal if the cost premium for flexible capacity is high. However, in many situations the firm may also want to consier another ual MTS/MTO approach, which in this case we call spackling. With spackling, the firm acquires only flexible capacity. Its first priority in any given perio is to prouce custom proucts via MTO. However, orer patterns for custom proucts are bumpy (uncertain), yieling an unesirable prouction profile compare to smooth scheules that allow for higher capacity utilization (see Figure 3). Thus the firm uses the same (flexible) prouction capacity to prouce stanar proucts in MTS fashion, to smooth the prouction scheule. The result is that MTO prouction closely tracks eman for customize proucts in each perio, while MTS output tracks eman for stanar proucts, not over the short-run but rather over the longer-run, with some inventory buil-up. In the spackling case, total prouction output from perio-to-perio is relatively smooth, or spackle. 2 An example of where spackling might be consiere is in the auto inustry, where managers have been talking for a ecae or more about the five-ay car. With this concept, a portion of cars woul be prouce only after en-user orers are receive, shipping within five ays of the orer. Manufacturers might eal with ay-to-ay eman uncertainty proucing cars for stocking at ealers in the same plant where they make the five-ay cars, using a spackling technique. The custom proucts woul be prouce first each ay, an the remaining capacity woul be use to prouce for ealer stocks. Our moel can help the firm etermine the optimal 2 The American Heritage Dictionary of the English Language: Fourth Eition 2000 efines spackle as A traemark use for a paste esigne to fill cracks an holes in plaster before painting or papering. This traemark often occurs in lowercase an as a verb 8

10 capacity level an pricing strategy. The spackling strategy is iscusse in more etail in Cattani, et al. (2003b). Figure 3. Spackling Yiels a Relatively Steay Total Output Stream, from One Prouction Facility. Total output is close to capacity Prouction Volume Stanar proucts mae via MTS using flexible capacity Custom proucts mae via MTO using flexible capacity Time Lastly, consier the case where eman calls primarily for customize proucts (the bottom row of Figure 1). While each en item is customize to the iniviual customer s tastes, the firm might still be able to make effective use of both efficient an flexible resources through a strategy we call jointing. To see how jointing works, consier Figure 4. In the setup of Frame A, twenty-six customers (A through Z) each buy a unique prouct esign (configurations A through Z). For simplicity, assume prouction of each configuration requires exactly two process steps. Further assume that all configurations use the same raw materials. However, steps one an two are specific to each configuration after each step, each configuration is ifferent from all other configurations. Thus the prouct fans-out into its iniviual custom configurations at the point that raw materials are consume we refer to this as the prouct fanout point. The push-pull point in Frame A of Figure 4 is positione after step two. This point is so name because an item is pushe through prouction up to this point an then hel in inventory 9

11 (as inicate by the triangle) until a customer purchases that specific item, figuratively pulling the item through the rest of the process (in this case, prouction is alreay complete at the pushpull point). Thus the approach epicte in Frame A is one where the firm prouces all configurations (A through Z) an then hols them in inventory until consumer purchase. Figure 4. Jointing Involves Postponing the Fan-Out Point an/or Avancing the Push-Pull Point. Frame A: Original Setup Frame B: After Prouct/Process Re-Design Raw Mtl. Fan-Out point Step 1 Config A Config B Config Z Step 2 Config A Config B Config Z Config A Config B Config Z Push-Pull point Cust. A Cust. B Cust. Z Raw Mtl. Postponement Fan-Out point Step 1 Generic Generic WIP WIP Push-Pull point Step 2 Cust. Config A A Cust. Config B B Config Cust. Z Z Avancement Jointing involves merging of the fan-out an push-pull points at a common joint in the process, as illustrate in Frame B of Figure 4. In this example, the prouct is re-esigne such that the first step is ientical for all configurations. That is, a partially-complete prouct is generic through step one in that it can be finishe into any one of the possible twenty-six configurations. The fan-out point now falls between process steps one an two. The push-pull point in Frame B has in turn been avance one step to also fall between process steps one an two items are now hel in an unfinishe state as generic work-in-process (WIP) until a specific 10

12 orer is receive, at which time process step two (specific to the configuration orere) is execute an the prouct is elivere to the customer. Thus jointing can be achieve through some combination of postponement (of the fan-out point) an/or avancement (of the push-pull point). One avantage of jointing is that only the generic prouct is inventorie, reucing unerage an overage costs through pooling benefits. A secon avantage of jointing is that efficient resources can still be use to prouce the generic prouct in MTS fashion up to the joint where the fan-out an push-pull points meet (beyon that joint the prouct is finishe in MTO fashion using flexible resources). If the cost premium for flexible resources is high, then jointing might best be pursue via postponement, while if the cost premium is low then avancement might be the preferre technique. An example of jointing in the service inustry is the walk-up rental of a room at an upscale hotel. The hotel room is cleane an hel in an almost-reay state up to the point where the walk-up customer arrives. Assuming this is a repeat customer, previous information regaring that customer is then accesse to customize the room to meet the preferences of that iniviual customer. For example, the staff may know that the customer prefers non-feather pillows, an can equip the room with these items at the last minute. In this paper we evelop a simple marketing moel to assess market preferences. Using the results of the marketing moel we synthesize an integrate into a common framework the three previously evelope ual strategies of texturing, spackling, an jointing. In this way we show how these strategies can be use to help coorinate prouct, process, an supply chain esign. We fin that these ual strategies become more attractive as the cost premium for flexible prouction over that of efficient prouction ecreases. We anticipate that this cost 11

13 premium will continue to iminish ue to avances in operations management, increasing the viability an attractiveness of the texturing, spackling, an jointing strategies. In 2 we lay out a marketing methoology for etermining whether eman is for stanar, custom, or both types of proucts. In 3, 4, an 5 we provie more etaile iscussions of texturing, spackling, an jointing, respectively. We conclue in Determining Whether Deman is for Stanar, Custom, or Both Types of Proucts We aress prouct an supply chain esign ecisions by moeling the firm s ecision to market stanar MTS proucts, an/or custom MTO proucts. We assume the firm has alreay mae relate ecisions regaring the exact prouct features to offer (i.e., what features to incorporate in the stanar proucts an in the base custom proucts, an what options to offer in the custom proucts). The effect of having two prouct types (stanar an custom) is to segment the market. A secon segment can generate incremental eman, but may also cannibalize eman for the first market segment. In other wors, some customers are willing to switch between prouct types. We refer to those customers that woul have bought the stanar prouct if it was the only prouct offere but will switch to buying a customize prouct if both types are offere as s-to-c switchers. Similarly, we refer to customers as c-to-s switchers if they woul buy a custom prouct if that was the only type offere but will switch to a stanar prouct if offere. The sizes of the s-to-c an c-to-s switcher categories are a function of customer preferences along with the prices for the stanar an customize proucts. Pricing plays a crucial role because lowering the price for one moel increases its sales by expaning the market but it also cannibalizes aitional sales of the other moel by increasing the number of switchers. 12

14 To etermine the optimal prices, we begin by aopting the usual efinition of a customer s reservation price, interpreting it as the most she is willing to pay for a prouct, an assume her reservation price for a customize prouct takes into account any elay cost that she might incur in having to wait for prouct elivery. Reservation price may be assesse, for example, through conjoint analysis an user esign as escribe by Dahan an Hauser (2002). Our focus in on the customer choice between a custom an stanar prouct; in this paper we take as a given the feature levels the firm incorporates into its stanar prouct an the menu of customizable features it offers. 3 Let c j enote prouct j s cost an p j its price, such that m j p j c j enotes the firm s ollar markup on prouct j. Here, prouct j can be the stanar prouct or any one of the custom proucts. Let r i j enote customer i s expecte ollar-equivalent utility for prouct j, excluing price: we refer to r i j as customer i s reservation price for prouct j. If viewe eterministically, then given the choice of buying prouct j or buying nothing, customer i buys at a price below r i j. Define i j r i j c j as the iscriminating markup; it is the markup the firm coul achieve in selling prouct j to customer i if it coul perfectly price iscriminate at the iniviual customer level (to the first egree), uner eterministic customer choice. Customer i s net utility (surplus) from buying prouct j at price p j is s i j = (r i j p j ) = ( i j m j ), since i j r i j c j an m j p j c j. From here on, we consier only one custom prouct for each customer, her most preferre custom prouct. Thus for any given customer i we can restrict our consieration of prouct j to be either stanar (s) or custom (c). Further, we assume the firm prices such that the absolute ollar markups for all configurations of the custom prouct are of constant magnitue 3 It is beyon the scope of this paper to etermine the optimal set of prouct features to inclue in the stanar proucts or to offer in custom proucts, but of course these ecisions also impact profitability. 13

15 m C. (This is accomplishe by pricing all margin into a base configuration an then aing or eleting options at cost.) This assumption an its rationale are iscusse in epth in Cattani, et al. (2003b), an have the effect of allowing us to compare the preference of a stanar prouct to that of a custom prouct across customers without neeing to know which specific configuration each customer has chosen. We assume that customers have preferences such that these iscriminating markups are istribute uniformly over the intervals [ min s, max s ] an [ min c, max c ]. See Figure 5. We assume the customer buys at most one prouct from the firm; the prouct offering her the most surplus. See Schmit an Porteus (2000) for similar methoology. Figure 5: Market Segmentation Base on Preference for Stanar versus Customize Proucts max c Discriminating Markup for Custom Prouct m c min c Customizers Nonbuyers s-to-c switchers c-to-s switchers Stanarists ( max max s s +m m, C m S ) (0, 0) min s m S Discriminating Markup for Stanar Prouct As inicate in Figure 5, we categorize customers into several types: a segment of potential customers who on t buy (non-buyers), a segment who buys the customize prouct but who wouln t buy the stanarize prouct (customizers), a segment who buys the stanar 14 max s

16 prouct but wouln t buy the customize prouct (stanarists), a segment who woul buy the stanar if it ha no choice but switches to the custom if a choice is offere (the s-to-c switchers), an a segment with the reverse preference (the c-to-s switchers). The size of each region epens on the prices (the markups) the firm charges for each prouct type. For example, increasing the markup of the stanar prouct increases the number of non-buyers while reucing the number of stanarists, an reucing the number of c-to-s switchers. Since each customer s ecision epens on markups, prouct pricing becomes a profit maximization calculation for the firm. We present below the solution for the case where ( max s m s ) < ( max c m c ). The solution for the case where ( max c m c ) < ( max s m s ) is foun similarly. The firm s profit π is: π = (markup on stanar) x (stanarists + s-to-c switchers) + (markup on custom) x (customizers + c-to-s switchers) max max min 1 max min max 1 max ( c c)( s s ) 2 ( s ) ( c )( s ) 2 ( s ) 2 2 = m c m m S + m S mc ms + m S. (1) We fin the optimal markups by eriving the first orer conitions associate with the above profit function an solving simultaneously for m S * an m C *. The Hessian matrix can be checke for the negative efinite property to etermine whether optimality is assure. The general solution is not analytically elegant so we forego its presentation here. Given a specific set of minimum an maximum iscriminating markups the solution is straightforwar. To gain basic insights we consier the special case where the minimum iscriminating markups are zero, an the firm sets equal markups for stanar an custom proucts, m S * = m C *. In this case, the solution reuces to: max max c s m s* = an the profit at s * 3 max max 2 m is ( c s ) 3 2 π * =

17 The above equations suggest prices an profits increase as max s an max c increase, while higher unit costs rive prices up an profits own, as one woul expect. Increasing the maximum reservation price for one prouct increases the prices of both proucts. We have assume that all costs are variable. In reality there likely are fixe costs associate with offering each type, such that the firm may choose to offer a type only if the projecte incremental profit from offering that type (as calculate above) excees some threshol value. For example, Figure 6 illustrates the cases where the firm might choose to offer only stanar or only custom proucts (the case where it likely chooses to offer both types was previously illustrate in Figure 5). In this paper we o not ientify that threshol value but simply categorize the outcome of the marketing moel into one the following three cases: 1) the firm chooses to offer only stanar proucts since virtually all eman is for stanar proucts; 2) the firm chooses to offer only custom proucts since virtually all eman is for custom; or 3) the firm chooses to offer both types since there is substantial eman for both. We procee in 3, 4, an 5 to evelop an operations moel to aress each of these three cases. Figure 6: When it May Be Desirable To Offer Only Stanar or Only Custom Proucts. The firm chooses to offer only custom proucts Discriminating Markup for Custom Prouct The firm chooses to offer only stanar proucts max c Deman for custom m c Deman Non- for buyers stanar min c (0, 0) min s m m S Discriminating Markup for Stanar Prouct max s Discriminating Markup for Custom Prouct m max c c min c (0, 0) Deman for custom proucts Deman for stanar proucts Non-buyers min s m S max s Discriminating Markup for Stanar Prouct 16

18 3. Process Design when Deman is Primarily for Stanar Proucts (Texturing) The operations moel for the case where eman exists only for stanar proucts follows that of Cattani, et al. (2003a), which we summarize in this section. To motivate the analysis, consier a semiconuctor manufacturer who prouces a computer chip in an efficient manufacturing fabrication plant, or fab. The chip is a stanar item but will become obsolete quickly, such that the efficient plant effectively gets just one opportunity to set the prouction volume. To hege against the uncertain eman volume, the firm also consiers acquiring some relatively expensive, but more flexible capacity that can be use if eman excees the efficient capacity. Effectively, such flexible prouction woul be make-to-orer (MTO), in that prouction woul commence only after eman is realize, even though the MTO prouce chips woul not be customize in any way. Shoul the firm actually acquire this flexible capacity? More specifically, how much efficient capacity an how much flexible capacity shoul it acquire? We employ a simple single-perio moel to aress these questions. Deman is characterize as a non-negative, continuous ranom variable X with istribution F( ) an probability ensity f( ), an represents eman for one perio (e.g., the season). The firm maximizes expecte profit by setting its efficient an flexible capacities, enote by K E an K F, respectively. Total capacity is enote by K = K E + K F. The firm prouces K E units before observing eman (given our setup, the firm will fully use any efficient capacity that it acquires). If eman excees K E, the firm fills this aitional eman with output from the flexible capacity, up to its capacity level, K F. Variable (per unit) costs are assume to be constant over the volume of interest for each prouction type. The variable costs of proucing one unit using efficient an flexible capacities are c E an c F, respectively. The selling price is p. We assume 17

19 lost sales, an at the en of the perio any leftover units are salvage at s. Each type of capacity incurs a fixe cost of θ i per unit per perio with iœ{e,f} for efficient an flexible capacities. The cost of each type of capacity per perio is thus θ i K i. 4 We make the following assumptions: (A 1) cf, ce, θ F, θ E > 0. All prouction costs are greater than zero. (A 2) c > s, c + θ > s. Salvage value is less than the cost of prouction. There is no F E E incentive to prouce solely for the salvage market. (A 3) cf + θ F < p an c E + θ E < p: We avoi trivial cases where the firm woul never invest in MTS or MTO capacity, respectively. The firm s optimal capacities K E * an K F * epen on the cost premium for flexible prouction, ((c F + θ F ) (c E + θ E )). If MTO prouction (always associate with flexible capacity) is no more expensive than the efficient type, then the firm exclusively uses MTO. If MTO is more costly but not too expensive, then the firm aopts the texturing strategy. By not too expensive we mean that it must be below, efine as follows: ( p ( cf θf) )( cf s) / ( p cf) +. (2) If the cost premium for flexible prouction excees, then the firm prouces only via MTS, which in this moel is always associate with efficient capacity. We refer to the strict use of MTS using efficient capacity as a mass prouction strategy. More specifically, the optimal efficient capacity K * E an the optimal flexible capacity K F * are: 4 Since efficient prouction is prouce to capacity once capacity is ecie, each unit of capacity will incur both the fixe cost θ E as well as the variable cost c E, so these costs coul be combine. 18

20 Case A (strict MTO prouction): If (c F + q F ) (c E + θ E ) 0, then K E * = 0 an K * 1 F = F ( p c θ ) / ( p c ) F F F. (3) Case B (texture prouction): If 0 < (c F + q F ) (c E +θ E ) <, then Case C (mass prouction): If (c F + q F ) (c E +θ E ), then K E * = F 1 [((c F + θ F ) ( c E + θ E )) / (c F s)], an K * F = K * K * 1 E = F ( p c θ )/ ( p c ) K * 1 E = F ( p c θ ) / ( p s) K F * = 0. E F F F K E *. E an (4) (5) (6) In each case we fin a newsvenor-type critical fractile solution. For a etaile interpretation of the results, along with comparative statics, see Cattani, et al. (2003a), where the moel is also extene to the case of two proucts. 4. Process Design When Deman is for Both Stanar an Custom (Spackling) Our approach for the case where eman exists for both stanar an custom proucts follows that of Cattani, et al. (2003b), which we summarize in this section. To motivate the analysis, consier Timbuk2, who makes messenger bags. The firm introuce an Internet site where a customer can orer a customize MTO bag. This irect channel streamlines the supply chain, while complementing the firm s traitional retail channel where pre-configure (stanar) MTS bags are sol. Timbuk2 consiere two alternative process esign strategies: a focus strategy where an efficient offshore plant makes stanar MTS bags at minimal cost while a flexible omestic plant makes custom MTO bags in a timely manner; an a spackling strategy where the firm makes both types of proucts (stanar an custom) omestically in the flexible plant. Senior 19

21 management wonere what overall effect each strategy woul have on manufacturing cost an ultimately, of course, on profitability. We employ an operations moel that maximizes profits by setting levels of flexible an efficient capacity, given the eman parameters for stanar an custom proucts. We consier a single-perio where eman is realize over T subperios, with T being the lea time offere to retailers for elivery of stanar proucts. The firm receives its orers for custom proucts at the beginning of each subperio an must eliver them by the en of that subperio. Separately, at the beginning of the first subperio, the firm receives all orers for stanar proucts, ue at the en of the perio. We o not account for holing costs. Subperio emans for custom proucts are assume to be inepenently an ientically 2 istribute C ~ ( C, C ) F N µ σ. Here we present the simplifie moel that assumes emans for stanar proucts are eterministic, with exactly T µ S units ue at the en of the perio. Total perio eman (the sum of emans for custom an stanar proucts) is enote by 2 ( µ µ ) σ F ~ N( T +, T ). T S C C The firm implements either a strategy of focus or spackling. This choice is effectively one of setting prouction capacities for efficient an flexible resources, enote by K E an K F, respectively (one unit of capacity can make exactly one unit of one prouct per sub-perio). A focus strategy involves positive amounts of both capacity types, while spackling requires only flexible capacity. Let θ E enote the fixe cost per subperio for a unit of efficient capacity (use only with focus to make stanar units) an θ F be the fixe cost of flexible capacity (use with focus to make custom units, an with spackling to make both stanar an custom units), 20

22 an let c E an c F enote their respective variable prouction costs per unit. Each unit sells for price p. We assume that we can view a custom prouct to be some base prouct plus or minus a set of options an that each custom an stanar prouct requires one unit of capacity. Without loss of generality, efine such base prouct to be the stanar prouct. We assume that we price to achieve the same markup on all custom proucts, such that we effectively sell all options at cost an simply set the markup on the base prouct. See Cattani, et al. (2003b) for rationalization of this assumption. Consistent with this marketing approach, here we assume θ F an c F apply to the base (stanar) prouct. At the beginning of the perio (the first subperio) the firm sets capacities K E an K F (if the firm chooses K E = 0, i.e., to spackle, then flexible capacity K). Let X Ci enote the realization of eman for custom units in subperio i. In the case of spackling, K units are prouce in total in each of the initial subperios: X Ci units of customize proucts an K X Ci units of stanar proucts. If eman over the T subperios for customize proucts is less than TK T µ S, the firm makes T µ S stanar units an then stops proucing stanar proucts for the remaining subperios. If perio eman for customize proucts is greater thantk T µ, there are lost sale of stanar units. We assume p, cp, cf, ce, θ F, θ E > 0 (all prouction costs are greater S than zero), c F + θ < p (we avoi the uninteresting case where flexible capacity is zero), an for F spackling we assume µ C << K (the firm has sufficient capacity each subperio to meet customize orers, X i < K i). Figure 1 suggeste spackling is preferre if the cost premium for flexible prouction is low. Here we fin low is anything below V: 21

23 σ V ( ) * * ( T T ) zθ + ( p c ) L( z ) C F F F µ T * 1 where z Φ ( c c θ ) ( c c ) * an ( i ) S ( ) =, the stanarize variate for the given fractile i {E,F}, i P i i P i L z is the stanar normal loss function. More formally, optimal capacities are: Case A (Focus): If ( θ ) ( θ ) c + c + V, then F F E E (7) K E * = µ S an * * K F µ C zf σ C = +. Case B (Spackling): If ( θ ) ( θ ) c + c + < V, then K * E = 0 an F F E E (8) (9) K*=K * 1 F = µ S + µ C + σc. T From (7) we gain the following managerial insights: spackling is most compelling when eman for custom proucts is more variable ( σ is large), the lea-time for prouction of stanar units is long (T is large), eman for stanar proucts is low ( µ S is small), the penalty for lost sales is high (p is large) an, of course, when the premium for flexible prouction is small (when ( c θ ) ( c θ ) + + is small). F F E E C (10) 5. Process Design When Deman is Primarily for Custom Proucts (Jointing) The escription below follows Schmit (2003). While this section covers the case where there is eman for (only) custom proucts, the principles iscusse apply also to the prouction of custom proucts when offere in conjunction with stanar proucts. In this paper we o not etermine the optimal positioning of the fan-out an push-pull points, or the optimal positioning of their common joint shoul the jointing strategy be followe, but offer a qualitative iscussion of some of the issues. 22

24 The iscussion surrouning Figure 4 provie efinition of terms such as the prouct fan-out point an the push-pull point. Consier a firm such as Hona, that offers an assortment of options for its moel S2000. These options cater to customers who iffer in their preferences for color, soun system, an other features, an in their price sensitivity. However, from an operations perspective, offering prouct variety is generally more costly, since proucing multiple configurations requires more expensive flexible resources, an since it may be ifficult to match output with eman. For example, Hona may make more re cars an fewer silver cars than are emane, or vice versa. How can Hona ameliorate this prouct variety ilemma? One possibility is to implement a postponement strategy, as illustrate via the classic case of the Hewlett-Packar (HP) inkjet printer (see Lee, et al. (1993)). HP originally prouce printers in make-to-stock (MTS) fashion (by this we mean printers were sol out of stock inventory), shipping numerous versions from its Vancouver plant to meet the nees of various geographic regions. HP reesigne the process to a the power supply locally, at multiple istribution centers, rather than at the factory (i.e., HP postpone the point at which the prouct fanne out into ifferent configurations), thus reucing safety-stock requirements. Lee an Tang (1997) iscuss some of the benefits of postponement in a more general setting. If Hona ecie to similarly implement a postponement strategy of some type, how far shoul it postpone the fan-out point (i.e., when shoul a generic S2000 fan out into its various configurations)? Shoul Hona have the istribution center paint the car, for example? Or shoul they postpone some of the prouction steps all the way own to the ealer level, by offering ealer-installe options? In our terminology, if the firm ships the prouct to istributors before painting, for example, the firm elays or postpones the point at which the prouct fans out into its various colors. Thus 23

25 the egree to which postponement is practice is effectively etermine by the positioning of the prouct fan-out point where shoul the fan-out point be positione? Alternately (or in combination with a postponement strategy), Hona might consier a strategy we call avancement. With avancement, the firm may initiate prouction but then hols the prouct in some unfinishe state until the customer places an orer, at which time prouction is complete. The auto inustry has long iscusse the iea of selling the bulk of its cars via MTO (currently most of the cars in the U.S. are bought off-the-lot). It has mentione a 10-ay car (meaning the car woul be elivere 10 ays after the orer is receive), or a 3-ay car, or an x-ay car. (See Jost (1994).) If Hona offere an x-ay car, how much assembly shoul it initiate before receiving a customer orer? At a minimum, it woul probably have the engine, transmission, an other components alreay assemble. Shoul it initiate any final assembly before receiving iniviual orers? Similar to the relationship between postponement an the fan-out point, there is a relationship between avancement an the push-pull point. The push-pull point is the point at which prouction is halte until receipt of an orer the item is pushe through prouction in make-to-stock (MTS) fashion, then hel in inventory until a customer orer pulls it through the rest of the way as a make-to-orer (MTO) unit. A ownsie of more extensive forms of avancement is that lea-times become longer rather than selecting a car off-the-lot, Hona s customers woul have to wait for its prouction an elivery. If Hona chose to pursue avancement, how far shoul the push-pull point be avance? Avancement is introuce here as a new term, but it is not a new concept. What the term is intene to convey is the notion that MTO prouction might in some cases mean starting from scratch in making the prouct, an in other cases it might mean simply flipping a switch 24

26 before shipping the prouct. In the first case, the push-pull point is at the start of the process, an in the latter it is at the en. Specifically, where shoul the push-pull point be? An where shoul it be relative to the fan-out point? Conceptually, prouction can be ivie into three stages, recognizing that the fan-out an push-pull points may not be elineate as cleanly as we imply in the current iscussion. We assume here that the fan-out point precees the push-pull point, as this seems likely to be the preferre sequence in most cases. Prior to the fan-out an push-pull points (stage 1), the firm makes generic items in continuous fashion. This can be accomplishe using efficient resources (approaching those of a flow shop, for example), such that, qualitatively, the cost for completing any given process step is low (as compare to what it woul cost to complete that step if it followe the fan-out an/or push-pull points). Between the fan-out an push-pull points (stage 2), the firm nees more flexible resources, as it is ealing with an assortment of proucts, but it can still make these in steay make-to-stock fashion, such that assembly cost is expecte to fall within an intermeiate range, relative to what it woul cost to perform that step in stage 1, or in stage 3. Subsequent to both the fan-out an push-pull points (stage 3), the firm nees flexible resources, an nees to prouce items on eman rather than at the firm s preferre rate of output. This implies poorer capacity utilization, such that the cost to perform a given process step is likely to be relatively more expensive. In aition to these cost issues, the firm shoul consier how esign costs, inventory costs an prouct obsolescence costs are affecte by avancement an postponement. An, very importantly, the firm shoul consier how customer preferences for variety an for immeiate versus elaye elivery impact the firm s ecisions. For example, if customers are sensitive to time elays, then postponement seems to be a more attractive strategy because higher levels of avancement imply longer lea-times. With 25

27 postponement, the majority of the prouction steps can take place before eman is realize. If customers are seeking wier variety an customization of proucts then avancement might be more attractive, since it may be har to achieve a wie variety of en-proucts from one generic nearly-finishe assembly. On the prouction sie, if the manufacturer can reuce lea-times through other means such as efficient scheuling, then avancement becomes more attractive while if commonality in esign can be reaily achieve then postponement seems esirable. We further suggest that it may be attractive to merge the prouct fan-out point with the push-pull point, a strategy we call jointing. When using the jointing strategy, the firm hols inventory of a generic prouct at the push-pull point, instea of holing inventory of specialize proucts further ownstream. Inventory costs shoul be lower, ue to pooling benefits (the generic sub-assembly can be use to meet eman for any en unit), an there woul not be any leftover units of the en prouct. However, there remains the question of where the joint (the merge point) shoul be locate: if a prouct has n steps in its prouction process, there are n + 1 possible merge points (from before the first step to after the last). That is, shoul the firm primarily rely on postponement (positioning the joint far ownstream) or avancement (putting the joint well upstream)? We expect that avancement is preferre when elivery elay can be minimize or tolerate, an when flexible prouction capacity is inexpensive. This intuition nees to be sharpene an quantifie in future work. Also, in practice the process steps an process flow may be significantly more complex than suggeste by our examples. The impact of this complexity nees to be assesse. Further, while merging the fan-out an push-pull points is attractive, it may not always, or even generally, be optimal the firm may want to stop short of jointing. Or, maybe there are 26

28 instances where the push-pull point shoul actually precee the fan-out point. Further work is neee to better unerstan these issues. 6. Discussion an Conclusion In this paper we have evelope a framework that links together certain elements of process esign (namely, whether the firm implements MTS an/or MTO), prouct esign (whether the firm offers stanar an/or custom proucts), an supply chain esign (whether the firm stocks proucts at retailers an/or sells irect). At its core, our framework is aime at aressing a common trae-off: firms esire an MTS approach because it minimizes prouction costs but simultaneously esire an MTO approach because it eliminates overage an unerage costs an can prouce customize proucts. To ameliorate the ilemma, we have outline three possible ual strategies that make simultaneous use of both types of process esign (both MTS an MTO). Each of these three strategies is matche with a specific prouct/supply chain esign. When customer preferences call for stanar prouct esigns stocke at retailers, texturing is the ual strategy that may offer benefits, if customers eman both stanar an customize esigns then spackling is the ual strategy to be consiere, an if customize proucts are offere then jointing may prove useful. It is instructive to compare an contrast the strategies of texturing, spackling, an jointing. With texturing, the core (i.e., primary) process type is MTS, with MTO ae as a texture to insure that prouction more closely matches the bumpy eman observe over multiple perios. Thus our strategy is similar to that suggeste by the prouct/process spectrum (i.e., high-volume stanar proucts call for flow shops that emphasize manufacturing efficiency an low cost) with the ifference that even with stanar proucts, an element of MTO may prove useful to reuce unerage/overage costs. 27

29 With spackling, MTO prouction might be consiere the core process esign, since priority in every sub-perio is given to the custom MTO proucts to insure timely elivery of items that cater to customers specific nees. MTS prouction is layere on top to smooth out the prouction stream an thereby increase manufacturing efficiency. Again, our strategy is similar to that suggeste by the prouct/process spectrum (i.e., specialize custom proucts call for job-shops that emphasize flexibility) with the ifference that an element of MTS may prove useful to improve efficiency. At the same time, we note that the spackling approach as outline herein is base on an assumption that eman for the two types of proucts is somewhat balance. If the eman for stanar proucts is significantly higher than that for custom proucts, then the firm may want to aopt a multi-layere spackling strategy involving two plants; an efficient plant proucing the bulk of the stanar proucts in MTS fashion, with the secon flexible plant proucing all the custom proucts an the remainer of the stanar proucts, in spackling fashion. With jointing, the process esign that preominates, either MTS an MTO, epens on prouct esign an supply chain esign opportunities. For example, the prouct esign may be conucive to postponement, or the supply chain may favor postponement because ownstream supply chain members have the capabilities to perform certain manufacturing steps. On the other han, avancement may be preferre if the prouction process is capable of being esigne to achieve relatively shorter lea times. In each of the texturing, spackling, an jointing strategies, our work points to the intertwine nature of prouct esign, process esign, an supply chain esign. 28

30 References Cattani, K., E. Dahan an G. Schmit. 2003a. Texturing: Using make-to-orer prouction to hege against uncertainty. Uner Review at Journal of Operations Management. Cattani, K. D., E. Dahan an G. Schmit. 2003b. Spackling: Smoothing make-to-orer prouction of custom proucts with make-to-stock prouction of stanar items. Uner Review at Management Science. Dahan, E. an J. R. Hauser The virtual customer. Journal of Prouct Innovation Management. 19 (5) Hayes, R. an S. Wheelwright. 1979a. Link manufacturing process an prouct life cycles. Harvar Business Review. Hayes, R. an S. Wheelwright. 1979b. The ynamics of process -- prouct life cycles. Harvar Business Review. Jost, K The three-ay car an manufacturing control systems. Automotive Engineering Lee, H., C. Billington an B. Carter Hewlett packar gains control of inventory an service through esign for localization. Interfaces. 23 (4) Lee, H. L. an C. S. Tang Moeling the costs an benefits of elaye prouct ifferentiation. Management Science. 43 (1) Schmit, G. an E. L. Porteus The impact of an integrate marketing an manufacturing innovation. Manufacturing an Service Operations Management. 2 (4) Schmit, G Mass customization. Operations Management Eucation Review. 29