Incentive alignment at the manufacturing–marketing interface: Design and validation of a management game

Beate Zöbeley; Stefan Minner; Christoph Kilger

doi:10.1007/s12159-011-0048-7

Incentive alignment at the manufacturing–marketing interface: Design and validation of a management game

Zöbeley, Beate; Minner, Stefan; Kilger, Christoph 2011-03-12 00:00:00 Logist. Res. (2011) 3:89–100 DOI 10.1007/s12159-011-0048-7 OR IGINAL PAPER Incentive alignment at the manufacturing–marketing interface: Design and validation of a management game • • Beate Zo ¨ beley Stefan Minner Christoph Kilger Received: 21 July 2010 / Accepted: 24 February 2011 / Published online: 12 March 2011 Springer-Verlag 2011 Abstract Supply chain coordination problems are fre- 1 Introduction quently found at the manufacturing–marketing interface. Inspired by a case study from the food industry, we ‘‘Can marketing and manufacturing coexist?’’ This ques- designed and validated a management game that focuses tion which Shapiro posed in 1977 has frequently been on potential conﬂicts between sales order acceptance and quoted. ‘‘Can marketing and manufacturing afford to not manufacturing utilization. We discuss how individual coexist?’’ was one answer [14]. Aligning the two functional behavior under distributed decision making can be areas has a signiﬁcant impact on company performance, improved to comply with overall company objectives if but looking at manufacturing and marketing from a system awareness is increased, incentive systems are resource-based and a market-based point of view visualizes carefully aligned, and cross-functional communication that these two essential functional areas are often clear protocols are improved. An empirical investigation in a opponents [8]. In many companies, interaction at the controlled laboratory experiment with university students interface of sales and manufacturing is coined by funda- shows the game’s effectiveness to achieve the key learning mental conﬂicts, lack of mutual understanding and com- objectives. The results show that both an aligned bonus munication, perturbing company efﬁciency [3, 18]. While scheme and information and communication increase sales is typically rewarded based on revenues, manufac- overall performance and decrease frictions between the two turing is rewarded for achieving high operational efﬁciency functional areas. As a further result from the experiment, and low production cost. With signiﬁcant changeover times we ﬁnd that an improved bonus scheme has a larger impact and a variety of customer-speciﬁc orders, this is a chal- than improved communication and information. lenging task. These diverging interests of sales and man- ufacturing naturally lead to conﬂicts [6]. The different tasks Keywords Marketing–operations interface Incentives and objectives of both areas are often reﬂected in the Management game Laboratory study compensation and mindset of the people involved, often resulting in suboptimal system performance [7]. Recent literature has proposed mechanisms to mitigate the adverse B. Zo ¨ beley effect of local incentives and private information, mostly Roche Diagnostics GmbH, GX, Sandhofer Straße 116, through various contractual arrangements providing 68305 Mannheim, Germany e-mail: beate.zoebeley@roche.com incentives for all players involved to make decisions that serve the entire system best [2]. However, the complexity S. Minner (&) of business reality does not allow perfect guidance of Logistics and Supply Chain Management, University of Vienna, decisions by incentives. Bru ¨ nner Straße 72, 1210 Vienna, Austria e-mail: stefan.minner@univie.ac.at By integrating compensation, transparency, and atti- tude into a single framework, we adopt a broad view on C. Kilger aligning individual behavior with total system’s objec- J&M Management Consulting AG, Willy-Brandt-Platz 5, tives. Based on this uniﬁed perspective, we develop a 68161 Mannheim, Germany business game focusing on coordination problems at the e-mail: christoph.kilger@jnm.com 123 90 Logist. Res. (2011) 3:89–100 manufacturing–marketing interface. A case study of a real At the case company, sales and manufacturing are company served as a starting point for our investigation and coordinated by an order management function. Ideally, all provided the basis for developing the game. Information incoming orders are processed centrally. However, desired was gathered by expert interviews and assessment of communication structures are often not adhered to, e.g., company data. The case company belongs to the food order placement and due date negotiations often take place industry and offers a wide and increasing variety of product between a customer and the responsible sales person. Order speciﬁcations. About 50% of the products are made to order acceptance and scheduling decisions can lead to waiting (MTO) being the focus of this study. Demand is sensitive to times including planned waiting times (e.g., the customer is the length and reliability of quoted lead times. Standard quoted a later due date than desired, a sales decision) and order lead times are on the order of 5 days. However, the unplanned waiting times (e.g., due to frictions in schedul- company receives a considerable amount of rush orders ing, on ﬁrst sight a manufacturing problem). In the man- which, upon order acceptance, have to be produced within agement game, these waiting times are combined into a 1–2 days. The manufacturing process is characterized by single queue. substantial changeover times, mainly due to cleaning pro- Optimizing a complex MTO production system with cesses. Therefore, producing rushed customer orders considerable changeover times and limited capacities fac- requires a careful trade-off between manufacturing and ing time-sensitive demand is a challenge on its own. Due revenue concerns. Despite a potentially negative impact, date management investigates how to optimize such a management has observed the acceptance of an increasing system, but most of the existing literature ignores both number of rush orders. pricing decisions and the impact of prices and lead times on Figure 1 shows a simpliﬁcation of the as-is processes. A demand [11]. Scheduling research (e.g., [16]) and revenue lack of coordination between decisions, as well as a lack of management [19] contribute solution approaches to MTO transparency and missing system awareness, leads to systems, but changeover times are scarcely included. Joint inefﬁciencies. Speciﬁcally, sales accept a variety of orders, consideration of order acceptance, due date determination, which often include speciﬁc features and are on short and scheduling is rare [10]. notice, without considering their negative effects on supply To achieve a coordinated pursuit of company objectives, chain costs. A reward system providing sales with incen- the behavior of decentralized decision makers has to be tives for achieving high sales volumes and manufacturing aligned. Monetary incentives constitute a central pillar. for achieving high operational efﬁciency, i.e., misaligned Implemented by means of compensation such as bonus incentives, is a major cause for conﬂicts and inefﬁciencies payments, they provide the core element of inﬂuencing and as local incentives reinforce local optimization. Compen- aligning behavior. This is the essence of agency theory as the theoretical basis to monetary incentive provision in dis- sation based on revenue naturally provides sales with incentives to accept as many orders as possible, irrespec- tributed decision making systems. Porteus and Whang [17] tive of their cost implications. Compensation based on use a principal-agent framework to investigate the coordi- individual revenues fosters competition between sales nation of the manufacturing–marketing interface. The people for scarce capacities and ampliﬁes the problem, company owner as the principal creates an internal market in especially in the presence of many customer enquiries for which manufacturing and marketing managers as agents rush orders. Lastly, non-monetary rewards such as nomi- operate in. Kouvelis and Lariviere [13] present a general- nating the ‘‘employee of the month’’ based on sales vol- ization of the internal market mechanism. They show that a umes reinforces the problem. system can be decentralized efﬁciently by distributing Fig. 1 System processes, SALES MANUFACTURING characteristics, and key Incoming “Pre Order Processing Queue” Fixed Production Lead- “Delay Queue” Finished Orders Time: Standard Goods decisions Capacity ... ... planned: customer is informed Lookahead not planned: customer is informed before placing an order after order has been placed Fixed Production Lead- Time: Accelerated Decisions: Communication? (1) Order acceptance (3) Due date assignment / (2) Due date determination scheduling Information on Workload? 123 Logist. Res. (2011) 3:89–100 91 decision control to a number of agents, while implementing learning objectives are (A) the effect of aligned incentives, suitable incentive mechanisms that align each agent’s indi- communication and information at the interface of sales, vidual goals with overall company objectives. Quantitative and manufacturing on company performance and (B) the performance measures can be complemented by qualitative potentially negative impact of rush orders. measures. However, the effectiveness of system coordina- The game aims at improving system awareness in gen- tion solely by means of monetary incentives is challenged by eral. Taking a process-oriented bird’s eye view on the several factors. The complexity of reality causes contracts to company, it allows recognizing the interdependencies be nearly unavoidably incomplete. Additionally, motivation between time-sensitive demand, order acceptance, and due theories suggest that decision makers are not solely pursuing date determination under capacity constraints and consid- monetary objectives; the decision maker’s attitude, e.g., erable changeover times. Due to the simpliﬁcations, cause- entrepreneurial thinking, can inﬂuence his actions. Lastly, and-effect relationships can be seen, and participants can the transparency of the system, including communication experience how their local actions affect the system as a and information, can have an impact on decisions. whole. Not only awareness for the processes is ameliorated, The manuscript is organized as follows. In Sect. 2,we but also comprehension of the behavior of all game par- discuss the development of the management game with its ticipants is enhanced [4, 15]. learning objectives. Section 3 reports on the design and The key learning is conveyed by contrasting a situation results of a game validation and further includes a cen- with misaligned incentives to one with aligned incentives tralized stochastic dynamic optimization model to bench- in two separate game rounds. With misaligned incentives, mark the game performance. Conclusions and discussion the participants should have the feeling that it is best for are given in Sect. 4. them if they decide independently. This, however, leads to lower the company performance. In the second round, they should feel that it is best to cooperate. This leads to 2 Management game development improved company performance and to a higher bonus of the participants, who thus learn that misaligned incentives We developed a management game for guiding behavior in can be a major cause for inefﬁciencies. Also, participants decentralized systems, speciﬁcally addressing the coordi- see the impact of inappropriate performance measures nation issues at the manufacturing–marketing interface. when, e.g., not utilization per se, but the number of orders delivered in time is the decisive criterion instead. Lastly, 2.1 Management games as interactive learning tools they experience that an effective bonus design can not only make the company better-off but also can increase the reward of each employee. Management games are an increasingly popular method of active learning. The term ‘‘Management Game’’ involves Rush orders often have a negative impact on supply the interaction between some models of a company, which chain costs and operating efﬁciency. However, this impact determines the impact of decision making and human of an incoming order can often not be easily assessed but behavioral elements [5] and the participants. Although less depends on the complex interplay between order charac- information is conveyed, active learning methods excel by teristics and the production system’s state. Regarding the the depth of learning which is increased by personal case study, one core problem is that sales people are not involvement [5]. One key beneﬁt of management games aware of the negative impact that accepting rush orders that are widely used as risk-free learning environments is to might have. In a production setting with restricted capac- understand the interaction between different functional ities and considerable changeover times, understanding areas. Knolmayer et al. [12] provide an overview on freely opportunity cost of changeovers is a vital component: not accessible, interactive learning objects in the area of the changeover induces cost, but the products that could logistics. have been produced and sold had the changeover not been We developed a partial model board game as an inter- carried out. Furthermore, the production of a rush order can active learning tool to be employed during company cause a delay of standard orders. The associated delay cost trainings following the idea of the ‘‘Beer Distribution can outweigh the margin the rush order would contribute to Game’’ as a role model, customized for tackling the case company proﬁts. In the game, optimizing system perfor- company’s problems of suboptimal system performance— mance involves the rejection of some, yet not all, rush misaligned incentives, lack of transparency, and lack of orders. The core learning is that a rush order can naturally system awareness. A board-based game was chosen to still be accepted even if causing frictions, but it is impor- allow for the possibility of including lively discussions tant to carefully judge its impact on system cost. The game between sales and manufacturing; participants should increases the awareness for this trade-off by providing experience conﬂicts and their causes. The intended key information transparency. 123 92 Logist. Res. (2011) 3:89–100 2.2 Design Manufacturing decides which orders they would prefer to accept, these orders are kept (S5). Orders that manufac- For the game design, a relatively low level of complexity turing prefers to reject are passed back to the respective was chosen to allow for robust conveyance of the key sales person (S6). Each sales person makes the ﬁnal deci- learning. This included sacriﬁcing some levels of realism, sion on these orders (S7): they can overrule manufacturing for example, by keeping demand exogenous. Yet, the core by adding a wildcard (one per unit of demand) to the elements of the company situation including the roles of respective order, forcing production (S5); or else may agree sales and manufacturing were mapped. with manufacturing’s preference and reject the order (S8). The game board is depicted in Fig. 2: Part 1 shows the Scheduling The scheduling and production part is divided places of the two sales people, working individually as into 6 horizontal rows, representing 6 periods. These are ‘‘Sales A’’ and ‘‘Sales B’’; Part 2 includes the places of two used to keep track of time. Rows 1 (S9) to 3 (S15) repre- manufacturing employees working jointly together. Sales sent the order pool without a delay, whereas rows 4 (S18) and manufacturing can be separated by a screen. The top to 6 (S19) collect orders beyond the due date, which are corners of each part include a key and a short instruction subject to waiting costs. Production planning ﬁelds (S12– per player (for a larger scale, see Fig. 3). LEGO tokens S14) show the period capacity of 5 units and represent the are used to represent order cards and game tokens. plan to be executed in the current period (Production) and a Demand is differentiated by product type (red or white), look-ahead planning with a horizon of two periods (Plan- order type (standard [gray] or rush order [yellow]), and ning). The latter, however, can still be modiﬁed when new order size (small (one unit) or large (two units)). A standard information about accepted orders becomes available in the order has a desired lead time of 3 periods, a rush order of 1 next period. Manufacturing forwards the order cards period. A speciﬁc demand sequence was constructed to according to type: standard orders to the ﬁrst row (S9), meet the objectives of (1) fair treatment of sales people accelerated orders (following the yellow arrow) to the third who are individually compensated and (2) conveyance of row in the yellow ﬁeld (S10), and issues the corresponding the key learning that acceptance of rush orders can cause raw material (from S11) to the planning ﬁeld(s) (S12–S14). substantial delays. The latter was done by building large Production rules are as follows: between different colors, a lots and adding a rush order of the other color to the changeover (black token) is necessary; after a changeover demand sequence, which would cause changeovers and any color can be produced (‘‘clean machine’’). Each thus delays to orders already accepted. This demand changeover reduces the available capacity by one unit. The sequence was tested and validated in the experiment (see current setup can be seen on the board (S13). Any order in Sect. 3.1). the order pool (rows 1-6) can be produced; however, orders Before the start of the game, participants are instructed have to be produced without interruption. In the planning on game sequence and rules (as explained in the following) ﬁelds, manufacturing plans the production schedule for 3 as well as the framework: the game is played in periods, periods using the game tokens and decides on a production measurement of monetary components is in Thaler. sequence for the current period in the production ﬁeld in Demand consists of 2–3 orders (4–6 units) per period, the third row (S13). The entire production plan can be capacity per period is 5. The game starts in period 1. The revised until the production decision has been made. While initial state of the game includes open orders of previous sales decides on order acceptance including the implicitly periods. Each period is announced by the instructor and determined due date according to order type, manufactur- includes four key steps: ing decides on the schedule, thus determining the realized Incoming customer demand, order acceptance check, and lead time and delivery date. At the end of step 2, manu- decision Both sales persons receive order cards (see S1 in facturing ﬁxes the production schedule. Fig. 2), including information on product type, size, order Production and packaging Orders are (instantaneously) type (including requested lead times), and order size. The produced, the setup marker (S13) is changed to the last players build the corresponding LEGO -order cards (S2) color produced, and the ﬁnished products are attached to and forward the incoming orders to manufacturing (S4), if the order cards (in rows 1–6), which had triggered pro- applicable by traversing the screen (S3). Manufacturing duction; this movement is indicated by the black arrow tries to integrate orders into their planned schedule taking between production and order cards. Changeover tokens into account promised due dates if an order is accepted. For are collected in a bin (S16). a standard order, the planned lead time is 3 periods, i.e., the due date is set equal to the index of the current period plus Shipment Completely fulﬁlled orders in rows 3, as well as 2. For a rush order, the planned lead time is a single period, in rows 4–6, are delivered to the customer (S17). All other i.e., the due date is set to the end of the current period. order cards are forwarded one step as a means to keep track 123 Logist. Res. (2011) 3:89–100 93 Fig. 2 Game board: parts 1 and 2 123 94 Logist. Res. (2011) 3:89–100 Fig. 2 continued 123 order flow orderflow one unit one unit standard order standard order - maximum delay:3 - maximum delay:3 accelerated order accelerated order - maximum delay: 1 - maximumdelay: 1 -accel mportant! - acceler important! wildcards - treat orders wildcards preferentially without - usewhen preferences of riskinglost sales sales and production differ - one card per unit waiting cost - standard order: 1 stone perstep - accelerated order: 3 stones perstep Logist. Res. (2011) 3:89–100 95 Fig. 3 Details of game board: Sales: Key and Instructions Manufacturing: Key and Instructions Key Key key and instructions one unit one unit Order acceptance: a) generate (assemble) incoming orders b) pass incoming orders to manufacturing Production andPackaging: Order acceptance: c) wait for manufacturing’s preferences 1 3 a) receive incoming orders a) setcolormarkeronlast color in d) receive orders manufacturing prefers theproduction field; b) decide on preferences to reject and decide on order - accept: pass orders to manufacture goods acceptance; if order is accepted, b) transfer finished goods to order „incoming orders“ attach wildcard(s) and return order to - reject: pass order back cards; transfer changeover manufacturing; else, pass order to chips to blue bucket c) WAIT for sales’ decision and „rejected orders“ receive „wildcard“-orders Scheduling: Shipment: 2 4 a) forward incoming orders, a) transfer all cards to the next issue raw material field b) schedule / reschedule b)apply delay penalties where production necessary of time, e.g., from row 1 (S9) to row 2, from row 2 to row (costs accrue at the sales department) and costs due to 3. Orders in row 3, which could not be produced in time, overtime or expediting shipments to compensate for are transferred to the waiting rows (e.g., from row 3 (S15) delayed production. Lost sales are orders that were to row 4 (S18) for a standard order, from row 3 (S10) to accepted yet could not be produced within a certain time row 6 (S19) for an accelerated order) and tagged with limit of 6 weeks for standard orders and 2 weeks for rush penalty tokens (S20). After the maximum waiting time orders. Rejected orders are orders which, due to scarce (standard order: 3 periods, rush order: one period), orders capacities, could not be accepted, they do not have a are lost at a penalty cost (ﬁeld ‘‘lost sales’’). After for- negative impact in this short-term game setup. warding the order cards, the game period is ﬁnished. After each period, the instructor collects the ﬁnished 2.3 Benchmark model orders (S17) and sorts the tokens for evaluation in the accounting section (S21). These tokens are the basis for For comparison purposes, the problem for known demand bonus calculation and game evaluation. After all periods can be modeled as discrete time lot-sizing and scheduling have been played, the tokens are counted, the performance problem with setups (see e.g., [9]. The deterministic measures computed and the bonus payments calculated. problem with all demands being known represents the best Finally, the results are communicated to the participants. solution only the instructor is able to obtain. The game Game parameters used are detailed in the benchmark participants only have incomplete information, i.e., 2–3 model, see Sect. 2.3. orders but do not know the distribution between the two For performance measurement, EBIT, deﬁned as reve- products and the extent of rush orders. To determine the nue minus cost, was chosen as the objective. The measure centrally optimal solution for both, a demand sequence proﬁt (EBIT minus bonus payments) was additionally under certainty and uncertainty, we developed a (stochas- computed for the purpose of evaluating game results. tic) dynamic program. In the following, we only show the Revenue is the standard measure from which variable costs deterministic version and brieﬂy sketch the required are deducted to compute the gross margin of sales. extensions for the stochastic version. Wildcards have to be used by a sales person who wants The planning horizon is T = 17 where orders only to have an order produced, which manufacturing preferred arrive in the ﬁrst 15 periods, and the two remaining periods to reject. The rationale behind the cost is additional are used to manufacture waiting orders. In every period, handling cost as well as frictions in scheduling. Utilization there is a limited manufacturing capacity of C = 5. Two (produced units/available capacity) is a measure of products i = 1, 2 are considered and for each product, throughput. Waiting costs include rebates granted to the there exist two types of orders: rush (r) and standard (s) customer for a due date later than their desired due date orders. Rush orders have a unit margin of p ¼ 50 and a Scheduling: wait Production and packaging: wait Shipment: wait 96 Logist. Res. (2011) 3:89–100 due date of 1 week. They can be either small with a we need to determine production quantities x , number of it capacity requirement of one unit or large with a capacity setups u , and the new setup state z . The logic for the it t?1 requirement of two units. The respective demands are setup decision variable is r1 r2 denoted by d and d . Standard orders with demand size it it u it s s d have an unit margin p ¼ 50, the due date is 3 weeks, 1 z ¼ j^ x [ 0; z ¼ j^ x ¼ 0 ^ z ¼ i; z ¼ i^ x [ 0 ^ z ¼ i it i t it t it tþ1 t jt tþ1 and they require 2 units of capacity. Orders that were 0 otherwise accepted but cannot be ﬁnished on time (within 1 week for A setup for product i is required in the following cases: (i) rush orders and 3 weeks for standard orders after the due the machine is initially setup for the other product j and the date) are lost at a penalty cost v = 600. Delayed orders are s r production quantity for i is positive, (ii) the initial setup is subject to waiting cost w ¼ 100 and w ¼ 300 per unit i i for the other product j, i is not produced but the initial setup and unit of time after the due date has passed for standard state in the following period is for i, and (iii) the initial and and rush orders, respectively. ﬁnal setup status are for product i but the other product j is The optimization problem formulation exploits the fol- produced in between. Production quantities and setups are lowing properties of an optimal solution to reduce com- limited by the available capacity of 5 units. plexity: (i) there is at most a single setup operation for each product in a period, (ii) (accepted) rush orders are satisﬁed 2.3.3 State transition with priority before any standard order is satisﬁed, and (iii) orders of any type are satisﬁed ﬁrst-in-ﬁrst-out. Furthermore, The new state of the following period y is a function of the t?1 we assume a pure make to order, zero inventory regime, i.e., current state y and the decisions about order acceptance and no products that have not been ordered are manufactured. In production quantities. We do not show the system of equa- dynamic programming, a simultaneous optimization prob- tions but rather sketch the logic behind the state transition. lem is decoupled into a sequential problem by introducing For both products i = 1, 2, manufacturing quantities x are stages (here periods). At the beginning of every period, it used to satisfy waiting and accepted orders in the sequence previous decisions have resulted in an initial state (here rush orders ﬁrst, then oldest to newest standard orders. In orders of a certain age and the setup status of the machine). case, only a single capacity unit remains, i.e., an order with a Given this state, the optimal decision for the period (con- capacity requirement cannot be satisﬁed in full, manufac- sisting of order acceptance and a production schedule) is turing will be started and completed in the following period determined such that the sum of direct rewards (from (therefore reducing the capacity by one unit). accepting orders) minus costs (for manufacturing, waiting, and lost orders) plus all the costs that result from taking 2.3.4 Functional equations t = 1, 2,…,T optimal decisions in all future periods (given the current periods decision). Next, we describe the dynamic program For each given initial state, the objective is to maximize the by stating the state of the system at the beginning of every expected proﬁt for accepted orders minus costs for waiting period, the decisions to be taken, the state transition, and the and lost orders. The constraints ensure the bounds for functional equations. accepting orders and the manufacturing capacity constraint. 2.3.1 State r 1 2 s max V ðy ; z Þ¼ p ðr þ 2r Þþ 2p s t t t it i it it i i¼1 A state at the beginning of period t is represented by the s s s s r r1 r2 w ðy þ y þ y Þ w ðy þ y i i3 i4 i5 i i i number of waiting rush orders of size one (r1) or two (r2) r1 r2 r1 r2 s vðy þ y þ y x Þ units (y ; y ), standard orders of age j, y ,(j = 1,…,5), and i5 it i i i i ij the initial setup state of the machine z [ {1,2}. Each indi- þ V ðy ; z Þ tþ1 tþ1 tþ1 1 1 2 2 s vidual order pool state variable can take values between 0 s:t: r d ; r d ; s d ; i ¼ 1; 2 it it it it it it and 3. Let y denote the vector of all order state variables. ðx þ u Þ C it it t i¼1 2.3.2 Decision u 2f0; 1g; x ; r ; it it it Decisions to be taken are which of the incoming orders to r ; s 0 and integer, i ¼ 1; 2 it it accept, how many units of each product to produce, and V ðy ; z Þ¼ 0; ðxÞ ¼ max 0; xg Tþ1 Tþ1 Tþ1 1 r1 2 r2 setup changeovers between products. Let r d ðr d Þ it it it it denote the number of accepted rush orders of size 1(2) and The demand data and the optimal decisions are shown in s d the respective accepted standard orders. Further, Table 1. At the beginning of the game, there exist already 4 it it 123 Logist. Res. (2011) 3:89–100 97 accepted orders for product 1 that were accepted in periods 3 Game validation -1 and 0, respectively. Given this initial order pool and that the machine is initially setup for product 1, the optimal 3.1 Experimental design and implementation decisions for each period are determined by forward evaluation of the eventual plans obtained from the We designed a controlled experiment tested with university functional equations. students in order to validate the game’s effectiveness in The columns in Table 1 show the respective demands. meeting its key learning objectives. Figure 4 shows the In the optimal solution, all standard orders are accepted. four treatments, each representing a combination of a bonus payment and the availability of information/com- For rush orders, the numbers in brackets show the number of accepted orders. The optimal solution under full infor- munication. Treatment 1 [T1] is assumed to be the worst mation yields V = 4,000. For the decision problem under case and treatment 4 [T4] the desired solution. uncertainty, every period has several scenarios with In treatments 1 and 2 without information/communica- respective probability and demands. Decisions in every tion, manufacturing utilization was not visible to sales and period have to be detailed by scenario. We assume that only limited information on game parameters was avail- there will be exactly three orders in every period. The split able. After each period, sales were informed about the between the two products is uniformly distributed. Fur- number of delayed orders. With the availability of infor- thermore, with probability 2/3, one of the three orders is a mation and communication (treatments 3 and 4), visibility rush order. In case there is a rush order, the sizes of one or of the whole game board was enabled, and full information two units are equally likely. This in total results in 16 on the game parameters was given to the participants. demand scenarios for each future period. After the reali- Secondly, we manipulated the players’ performance zation of demand in every period, the optimal decision is measurement systems by means of incentive alignment. implemented for the realized demand. The expected value Bonus 1 (shown in Table 2) represents the initial situation using the above assumptions is 3862.57. However, given of misaligned incentives to visualize its resulting problems the realized demand sequence, this results in the same (treatments 1 and 3). optimal decisions. Manufacturing is primarily evaluated based on utiliza- tion. Waiting costs as the inﬂuential factor on company performance are only included with a small weight. Sales are compensated based on individual revenues and com- petes for scarce capacities. As the demand sequence includes rush orders causing changeovers, incentive con- Table 1 Customer orders, acceptance decisions, and manufacturing ﬂicts arise. Each sales person is interested in having each quantities incoming order produced, whereas manufacturing aims at s s r1 r2 r1 r2 td d x x d d d d 1t 2t 1t 2t 1t 1t 2t 2t minimizing the number of changeovers even if this causes delays. Additionally, not the gross margin, but only reve- -12 nues are included in the bonus of sales. This conceals the fact that one waiting step reduces the margin to zero. Also, 13 5 compensating sales primarily on the basis of revenue 2 2 1(0) 5 weighs customer service as a sales’ objective (measured by 3 2 1(0) 5 waiting costs) only insufﬁciently within the bonus. As an 43 5 additional bonus component, each sales person is penalized 5 2 1(1) 3 1 for using a wildcard as a measure of conﬂict. Bonuses of 6 2 1(0) 5 both sales and manufacturing include a penalty on waiting 7 2 1(0) 5 cost. Lastly, lost sales are penalized. 8 1 1(0) 5 Bonus 2 implements aligned incentives as the core pillar 92 1 5 of the solution approach to improve coordination (treat- 10 2 1(1) 2 2 ments 2 and 4). Bonus payments were constructed, con- 11 3 5 sidering both the realities of the case study and the 12 2 1(0) 5 implementation within the game. The bonus for both sales 13 2 1(1) 5 and manufacturing comprises an individual component (as 14 2 1(1) 3 1 for Bonus 1) with a weight of 80% and an overall proﬁt 15 3 5 sharing component with a weight of 20%. As proﬁt 16 5 depends on ﬁnal bonus payments, EBIT was used as a 17 3 measure for the bonus calculations. For the determination 123 98 Logist. Res. (2011) 3:89–100 Fig. 4 Overview of treatments Bonus 1 Bonus 2 incomplete information, Treatment 1 Treatment 2 no communication Impact of conflicting incentives on Impact of improved individual system performance. incentives and a profit-sharing component. complete information, Treatment 3 Treatment 4 communication Impact of complete information and Impact of improved bonus and communication. complete information and communication. Table 2 Details of bonus Bonus 1 Bonus 2 calculation Proﬁt sharing component: weight 20% 5% of EBIT Sales Individual component: 100% Individual component: weight 80% ?5 per sold unit of product ?5 per unit -1 per waiting period standard order -1 per waiting period standard order -3 per waiting period rush order -3 per waiting period rush order -1 per used wildcard -3 per used wildcard -100 per lost order -100 per lost order Manufacturing Individual component: 100% Individual component: weight 80% Initial bonus payment: 180 Initial bonus payment: 180 Target: 96%; ±10 per% utilization more/less -1 per waiting period standard order -10 per waiting period standard order -3 per waiting period rush order -30 per waiting period rush order -100 per lost order -100 per lost order of the individual part, utilization is excluded from manu- 3.2 Results facturing’s bonus and substituted by a 10 times higher penalty on waiting costs. This aligns manufacturing’s Table 3 gives an overview of average key performance objectives with company objectives. Sales’ incentives are measures and bonus payments for all four treatments for aligned through a penalty on wildcards 3 times higher, the ﬁrst 15 periods of the game. The gross margin is rel- which gives them an incentive to adhere to manufactur- atively stable. Note that the corresponding value of gross ing’s decisions. margin in the optimal solution is 3,600 (4,000 minus the revenue of 8 units produced after the horizon of 15 peri- The standard methodology of experimental economics was used (e.g., [4]. Eighty students were recruited as par- ods). Waiting costs show substantial differences, whereas wildcard costs are mostly negligible. EBIT differs between ticipants at the University of Mannheim, mostly graduate business students specializing in logistics. For each treat- the treatments, so does utilization, but to a smaller extent. ment, ﬁve individual sessions were conducted, and a Result I System performance is improved by incentive between-subject design with different participants was alignment as well as by information and communication. chosen. The students were assigned roles as sales (2 per The impact of aligned incentives on system performance is session) and manufacturing (2 per session). Participants larger than the impact of information and communication. were provided with instructions, both written and oral. Better company performance is reﬂected in higher bonus Instructions contained information on the bonus payment payments. including how their performance in the game (measured in First, the inﬂuence of bonus design is analyzed. Com- Thaler) translated into real monetary payments (in Euros) after the game. Compensation of the students consisted of a paring T1 and T2 yields signiﬁcant differences (p = 0.008) for all values but gross margin; T3 and T4 are signiﬁcantly ﬁxed show up fee of €7.50 and a performance-related bonus with an expected value of €3. different concerning all values except wildcard cost. With 123 Logist. Res. (2011) 3:89–100 99 Table 3 Overview of key performance measures and bonus setting, this result may differ depending e.g., on the total payments amount of the bonus compared to ﬁxed compensation. In order to sustainably implement an aligned incentive Bonus 1 Bonus 2 scheme, the results need to be Pareto-improving, i.e., not T1 T3 T2 T4 only the company but also the employees have to be better- Gross margin 3,380 3,460 3,450 3,560 off. An acceptance of a new policy by employees implies Total cost -2,008 -1,693 -252 -23 that higher bonus payments for all players should follow better company performance, which is conﬁrmed by the Thereof waiting cost -1,920 -1,660 -240 -20 Thereof wildcard cost -88 -33 -12 -3 results: EBIT (i.e., the total of proﬁt and bonus payments) and manufacturing bonus are positively correlated (Spear- EBIT 1,372 1,767 3,198 3,537 man’s q value of 0.931), EBIT and bonuses of the two sales Utilization (%) 90 93 93 95 persons are correlated with values of 0.849 and 0.724, Bonus sales (total) 301 323 332 354 respectively (all signiﬁcant at the 0.01 level, 1-tailed). Bonus manufacturing (total) 206 267 314 356 Additionally, the absolute bonus values of T2 and T4 Proﬁt 865 1,177 2,552 2,828 (Bonus 2) were signiﬁcantly higher than the values in T1. The second result focuses on the analysis of order acceptance/rejection decisions of sales and manufacturing EBIT, being the key measure, Bonus 2 leads to better (see Table 4). In the optimal solution, 34 standard orders results than Bonus 1 for both treatments with and without and 4 rush orders are accepted, whereas 6 rush orders are communication. This result conﬁrms the choice of mone- rejected. tary incentives as the central pillar to guide behavior. The Result II Conﬂict between sales and manufacturing cau- students were inﬂuenced in their behavior both by the ses inefﬁciencies through waiting costs. Acceptance of rush ﬁctional game setting and the intention to ‘‘win’’ and by the orders is a main cause for these inefﬁciencies. Rejection of real monetary pay. the ‘‘right’’ orders improves company performance. Secondly, the inﬂuence of information and communi- cation is reviewed: T2 and T4 are different with regard to In T4, the highest number of orders is accepted in the EBIT (p = 0.016) and gross margin (p = 0.056), however ﬁrst step, indicating the lowest necessity of direct interac- both with a small difference of absolute values. T1 and T3 tion at the interface and therefore little frictions. In T1, are not signiﬁcantly different. The result thus mainly holds manufacturing rejected the largest amount of orders with regard to T2 and T4, implying that information and (rejected orders and orders with wildcards). Comparing T1 communication improve system performance with Bonus and T4, the number of rejected standard orders is signiﬁ- 2. An explanation for T1 and T3 not being signiﬁcantly cantly different (p = 0.008), as well as the number of different is the observation of different risk attitudes: in T1 rejected rush orders (p = 0.032). In T4, no standard orders (without visibility of capacities), sales was often afraid of were rejected. lost sales and hence did reject more orders as compared to The usage of wildcards, which was subject to a cost of 5 T3, where capacities as well as waiting queues were visi- Thaler per unit, is interpreted as follows: orders rejected by ble. Bringing improved information and a better aligned manufacturing could be accepted by sales using a wildcard, incentive system together, T4 yields signiﬁcantly better thus overruling the recommendation of manufacturing. In results compared to T1 in all performance values T1, signiﬁcantly, more wildcards are used than in T4, as (p = 0.032 for gross margin, p = 0.008 for EBIT, waiting well as in T2 (p = 0.008 for standard orders, p = 0.056 for costs and wildcard cost). Bonus payments and information rush orders). With wildcards as a measure of conﬂict, T2 transparency lead to improved coordination of sales and and T4 (Bonus 2) are thus better than the original setting manufacturing. (T1) in this respect. Communication, which was observed When comparing the impact of bonus (T2) with the during the experiment, made wildcard usage unnecessary. impact of information (T3), we get the following results: Lastly, the correlation between order acceptance/rejec- Looking at EBIT and waiting cost, there is a signiﬁcant tion decisions and waiting costs as a measure for produc- difference between T2 and T3 (p = 0.095 and p = 0.032). tion delays, and thus low customer service is analyzed. The This indicates that an improved bonus (T2 to T1) has a number of accepted orders with wildcard as a measure of higher positive impact on total system performance (mea- cross-functional conﬂict are positively correlated with sured by EBIT) as a change in information availability (T3 waiting costs (q = 0.748, signiﬁcant at the 0.01 level to T1). This result shows a successful implementation of a (1-tailed)). The q value of rush orders with wildcards was parameterization that allows to conveying the key learning 0.528 and standard orders with wildcards 0.678, both sig- of the impact of incentive alignment. In a real world niﬁcant at the same level. This indicates that wildcard 123 100 Logist. Res. (2011) 3:89–100 Table 4 Order acceptance decisions incentive alignment by the design of bonus schemes and the role of information and communication. Bonus 1 Bonus 2 T1 T3 T2 T4 Acknowledgments The authors would like to thank the two anon- ymous reviewers for their helpful suggestions to improve the manu- # of accepted orders without wildcard: 22.4 29.4 31.0 34.0 script and Mirko Kremer, The Pennsylvania State University, for his standard input to this project. # of accepted orders without wildcard: rush 3.6 5.8 4.8 3.4 # of accepted orders with wildcard: standard 8.8 2.6 0.8 0.0 References # of accepted orders with wildcard: rush 2.4 1.4 0.4 0.4 # of rejected orders: standard 2.8 2.0 2.2 0.0 1. Bendoly E, Donohue K, Schultz K (2006) Behavior in operations # of rejected orders: rush 4.0 2.8 4.8 6.2 management: assessing recent ﬁndings and revisiting old Subtotal: # of accepted orders: rush 6.0 7.2 5.2 3.8 assumptions. J Oper Manag 24(6):747–752 Subtotal: # of accepted orders: standard 31.2 32.0 31.8 34.0 2. Cachon GP (2003) Supply chain coordination with contracts. In: de Kok AG, Graves SC (eds) Supply chain management: design, Subtotal: # of accepted orders without 26.0 35.2 35.8 37.4 coordination and operations, handbooks in OR & MS, vol 11. wildcard Elsevier, Amsterdam, pp 229–339 Subtotal: # of accepted orders with wildcard 11.2 4.0 1.2 0.4 3. Chayet S, Hopp WJ, Xu X (2004) The marketing-operations Total: # of accepted orders 37.2 39.2 37.0 37.8 interface. In: Simchi-Levi D, Wu SD, Shen Z-J (eds) Handbook of quantitative supply chain analysis: modeling in the e-business Total: # of rejected orders 6.8 4.8 7.0 6.2 era. Kluwer, Norwell, Massachusetts, pp 485–554 All values: average values per treatment 4. Croson R (2002) Why and how to experiment: methodologies from experimental economics. Univ Illinois Law Rev 2002(4): 921–945 5. Elgood C (1990) Using management games. Gower, Aldershot 6. Eliashberg J, Steinberg R (1993) Marketing-production joint usage leads to production delays. The number of accepted decision making. In: Eliashberg J, Lilien G (eds) Handbooks in rush orders with wildcard and waiting costs are correlated operations research and management science. Elsevier, Amster- (q = 0.528, p = 0.01). However, rush orders without dam, pp 625–880 wildcard and waiting costs show no signiﬁcant correlation. 7. Ellinger AE, Daugherty P, Keller SB (2000) The relationship between marketing/logistics interdepartmental integration and Yet, the subtotal of rush orders and waiting costs are cor- performance in US. Manufacturing ﬁrms: an empirical study. related with q = 0.547 (p = 0.01). Again, the parameter- J Bus Logist 21(1):1–22 ization allowed for a conveyance of the intended key 8. Ellinger AE, Keller SB, Hansen JD (2006) Bridging the divide learning of the negative impact of most rush orders in this between logistics and marketing: facilitating collaborative behavior. J Bus Logist 27(2):1–27 speciﬁc setting. Lastly, total order rejection and waiting 9. Fleischmann B (1990) The discrete lotsizing and scheduling costs are not signiﬁcantly correlated, yet rejected standard problem (DLSP). Eur J Oper Res 44(3):337–348 orders and waiting costs show a correlation (q = 0.507). 10. Gunasekaran A, Ngai EWT (2009) Modelling and analysis of Rejected rush orders are correlated with waiting costs built-to-order supply chains. Eur J Oper Res 195(2):319–334 11. Keskinocak P, Tayur S (2004) Due date management policies. In: (q = 0.547); this shows that not the total number of Simchi-Levi D, Wu SD, Shen Z-J (eds) Handbook of quantitative rejected orders is decisive but rejection of the ‘‘right’’ supply chain analysis: modeling in the E-business era. Kluwer, orders. Norwell, Massachusetts, pp 295–334 12. Knolmayer G, Montandon C, Schmidt R (2004) Interaktive Lernobjekte zur Logistik. Wirtschaftsinformatik 46(2):139–151 13. Kouvelis P, Lariviere MA (2000) Decentralizing cross-functional 4 Conclusion and managerial implications decisions: coordination through internal markets. Manage Sci 46(8):1049–1058 Based on an industrial case from the food industry, we 14. Malhotra MK, Sharma S (2002) Spanning the continuum between marketing and operations. J Operat Manag 20(3):209–219 developed a management game to involve decision makers 15. Mentzer J, Stank T, Esper T (2008) Supply chain management in the core conﬂicts at the manufacturing–marketing and its relationship to logistics, marketing, production, and interface, revenue versus utilization maximization. operations management. J Bus Logist 29(1):31–46 Thought as a training tool to support change management 16. Pinedo M (2002) Scheduling. Theory, algorithms, and systems, 2nd edn. Prentice Hall, Upper Saddle River and provide mutual understanding of different functional 17. Porteus EL, Whang S (1991) On manufacturing/marketing areas, we conducted a laboratory experiment with student incentives. Manage Sci 37(9):1166–1181 subjects to stress the importance of empirical game vali- 18. Shapiro BP (1977) Can marketing and manufacturing coexist? dation as real behavior might deviate from standard theo- Harvard Bus Rev 55(5):104–114 19. Talluri KT, van Ryzin GJ (2005) The theory and practice of retical predictions [1]. We found that the developed game revenue management. Springer, Berlin served the purpose of highlighting the importance of http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Logistics Research Springer Journals http://www.deepdyve.com/lp/springer-journals/incentive-alignment-at-the-manufacturing-marketing-interface-design-OumgFGWhnx

Loading next page...

References (26)

G. Knolmayer, Corinne Montandon, R. Schmidt (2004)
Interaktive Lernobjekte zur Logistik
Wirtschaftsinformatik, 46
D. Simchi-Levi, S. Wu, Z. Shen (2004)
Handbook of Quantitative Supply Chain Analysis: Modeling in the E-Business Era (International Series in Operations Research & Management Science)
B. Fleischmann (1990)
The discrete lot-sizing and scheduling problem
European Journal of Operational Research, 44
Rachel Croson (2002)
Why and how to experiment: Methodologies from experimental economics
University of Illinois Law Review, 2002
W. Lieberman (2005)
The Theory and Practice of Revenue Management
Journal of Revenue and Pricing Management, 3
(1977)
Can marketing and manufacturing coexist?
Subtotal: # of accepted orders with wildcard 11
M. Malhotra, Subhash Sharma (2002)
Spanning the continuum between marketing and operations
Journal of Operations Management, 20
Ellinger Enrique (1997)
The relationship between marketing/logistics interdepartmental integration and performance in U.S. manufacturing firms : an empirical study
GP Cachon (2003)
Supply chain management: design, coordination and operations, handbooks in OR & MS
Evan Parteus, S. Whang (1991)
On manufacturing/marketing incentives
Management Science, 37
Alexander Ellinger, S. Keller, J. Hansen (2006)
BRIDGING THE DIVIDE BETWEEN LOGISTICS AND MARKETING: FACILITATING COLLABORATIVE BEHAVIOR
Journal of Business Logistics, 27
P. Keskinocak, S. Tayur (2004)
Due Date Management Policies
A. Gunasekaran, E. Ngai (2009)
Modeling and analysis of build-to-order supply chains
Eur. J. Oper. Res., 195
Michael Pinedo (1994)
Scheduling: Theory, Algorithms, and Systems
Sergio Chayet, W. Hopp, Xiaowei Xu (2004)
The Marketing-Operations Interface
M. Salomon (1991)
The Discrete Lotsizing and Scheduling Problem (DLSP)
P Keskinocak, S Tayur (2004)
Handbook of quantitative supply chain analysis: modeling in the E-business era
E. Bendoly, K. Donohue, Kenneth Schultz (2006)
Behavior in operations management: Assessing recent findings and revisiting old assumptions
Journal of Operations Management, 24
Gérard Cachon (2003)
Supply Chain Coordination with Contracts
G. Nemhauser, A. Kan (1989)
Handbooks in operations research and management science
All values: average values per treatment
J. Mentzer, T. Stank, Terry Esper (2008)
SUPPLY CHAIN MANAGEMENT AND ITS RELATIONSHIP TO LOGISTICS, MARKETING, PRODUCTION, AND OPERATIONS MANAGEMENT
Journal of Business Logistics, 29
J. Eliashberg, R. Steinberg (1993)
Marketing-production joint decision-making
P. Kouvelis, M. Lariviere (2000)
Decentralizing Cross-Functional Decisions: Coordination Through Internal Markets
Management Science, 46
C. Elgood (1991)
Using Management Games

Publisher: Springer Journals
Copyright: Copyright © 2011 by Springer-Verlag
Subject: Engineering; Engineering Economics, Organization, Logistics, Marketing; Logistics; Industrial and Production Engineering; Simulation and Modeling; Operation Research/Decision Theory
ISSN: 1865-035X
eISSN: 1865-0368
DOI: 10.1007/s12159-011-0048-7
Publisher site: See Article on Publisher Site

Abstract

Logist. Res. (2011) 3:89–100 DOI 10.1007/s12159-011-0048-7 OR IGINAL PAPER Incentive alignment at the manufacturing–marketing interface: Design and validation of a management game • • Beate Zo ¨ beley Stefan Minner Christoph Kilger Received: 21 July 2010 / Accepted: 24 February 2011 / Published online: 12 March 2011 Springer-Verlag 2011 Abstract Supply chain coordination problems are fre- 1 Introduction quently found at the manufacturing–marketing interface. Inspired by a case study from the food industry, we ‘‘Can marketing and manufacturing coexist?’’ This ques- designed and validated a management game that focuses tion which Shapiro posed in 1977 has frequently been on potential conﬂicts between sales order acceptance and quoted. ‘‘Can marketing and manufacturing afford to not manufacturing utilization. We discuss how individual coexist?’’ was one answer [14]. Aligning the two functional behavior under distributed decision making can be areas has a signiﬁcant impact on company performance, improved to comply with overall company objectives if but looking at manufacturing and marketing from a system awareness is increased, incentive systems are resource-based and a market-based point of view visualizes carefully aligned, and cross-functional communication that these two essential functional areas are often clear protocols are improved. An empirical investigation in a opponents [8]. In many companies, interaction at the controlled laboratory experiment with university students interface of sales and manufacturing is coined by funda- shows the game’s effectiveness to achieve the key learning mental conﬂicts, lack of mutual understanding and com- objectives. The results show that both an aligned bonus munication, perturbing company efﬁciency [3, 18]. While scheme and information and communication increase sales is typically rewarded based on revenues, manufac- overall performance and decrease frictions between the two turing is rewarded for achieving high operational efﬁciency functional areas. As a further result from the experiment, and low production cost. With signiﬁcant changeover times we ﬁnd that an improved bonus scheme has a larger impact and a variety of customer-speciﬁc orders, this is a chal- than improved communication and information. lenging task. These diverging interests of sales and man- ufacturing naturally lead to conﬂicts [6]. The different tasks Keywords Marketing–operations interface Incentives and objectives of both areas are often reﬂected in the Management game Laboratory study compensation and mindset of the people involved, often resulting in suboptimal system performance [7]. Recent literature has proposed mechanisms to mitigate the adverse B. Zo ¨ beley effect of local incentives and private information, mostly Roche Diagnostics GmbH, GX, Sandhofer Straße 116, through various contractual arrangements providing 68305 Mannheim, Germany e-mail: beate.zoebeley@roche.com incentives for all players involved to make decisions that serve the entire system best [2]. However, the complexity S. Minner (&) of business reality does not allow perfect guidance of Logistics and Supply Chain Management, University of Vienna, decisions by incentives. Bru ¨ nner Straße 72, 1210 Vienna, Austria e-mail: stefan.minner@univie.ac.at By integrating compensation, transparency, and atti- tude into a single framework, we adopt a broad view on C. Kilger aligning individual behavior with total system’s objec- J&M Management Consulting AG, Willy-Brandt-Platz 5, tives. Based on this uniﬁed perspective, we develop a 68161 Mannheim, Germany business game focusing on coordination problems at the e-mail: christoph.kilger@jnm.com 123 90 Logist. Res. (2011) 3:89–100 manufacturing–marketing interface. A case study of a real At the case company, sales and manufacturing are company served as a starting point for our investigation and coordinated by an order management function. Ideally, all provided the basis for developing the game. Information incoming orders are processed centrally. However, desired was gathered by expert interviews and assessment of communication structures are often not adhered to, e.g., company data. The case company belongs to the food order placement and due date negotiations often take place industry and offers a wide and increasing variety of product between a customer and the responsible sales person. Order speciﬁcations. About 50% of the products are made to order acceptance and scheduling decisions can lead to waiting (MTO) being the focus of this study. Demand is sensitive to times including planned waiting times (e.g., the customer is the length and reliability of quoted lead times. Standard quoted a later due date than desired, a sales decision) and order lead times are on the order of 5 days. However, the unplanned waiting times (e.g., due to frictions in schedul- company receives a considerable amount of rush orders ing, on ﬁrst sight a manufacturing problem). In the man- which, upon order acceptance, have to be produced within agement game, these waiting times are combined into a 1–2 days. The manufacturing process is characterized by single queue. substantial changeover times, mainly due to cleaning pro- Optimizing a complex MTO production system with cesses. Therefore, producing rushed customer orders considerable changeover times and limited capacities fac- requires a careful trade-off between manufacturing and ing time-sensitive demand is a challenge on its own. Due revenue concerns. Despite a potentially negative impact, date management investigates how to optimize such a management has observed the acceptance of an increasing system, but most of the existing literature ignores both number of rush orders. pricing decisions and the impact of prices and lead times on Figure 1 shows a simpliﬁcation of the as-is processes. A demand [11]. Scheduling research (e.g., [16]) and revenue lack of coordination between decisions, as well as a lack of management [19] contribute solution approaches to MTO transparency and missing system awareness, leads to systems, but changeover times are scarcely included. Joint inefﬁciencies. Speciﬁcally, sales accept a variety of orders, consideration of order acceptance, due date determination, which often include speciﬁc features and are on short and scheduling is rare [10]. notice, without considering their negative effects on supply To achieve a coordinated pursuit of company objectives, chain costs. A reward system providing sales with incen- the behavior of decentralized decision makers has to be tives for achieving high sales volumes and manufacturing aligned. Monetary incentives constitute a central pillar. for achieving high operational efﬁciency, i.e., misaligned Implemented by means of compensation such as bonus incentives, is a major cause for conﬂicts and inefﬁciencies payments, they provide the core element of inﬂuencing and as local incentives reinforce local optimization. Compen- aligning behavior. This is the essence of agency theory as the theoretical basis to monetary incentive provision in dis- sation based on revenue naturally provides sales with incentives to accept as many orders as possible, irrespec- tributed decision making systems. Porteus and Whang [17] tive of their cost implications. Compensation based on use a principal-agent framework to investigate the coordi- individual revenues fosters competition between sales nation of the manufacturing–marketing interface. The people for scarce capacities and ampliﬁes the problem, company owner as the principal creates an internal market in especially in the presence of many customer enquiries for which manufacturing and marketing managers as agents rush orders. Lastly, non-monetary rewards such as nomi- operate in. Kouvelis and Lariviere [13] present a general- nating the ‘‘employee of the month’’ based on sales vol- ization of the internal market mechanism. They show that a umes reinforces the problem. system can be decentralized efﬁciently by distributing Fig. 1 System processes, SALES MANUFACTURING characteristics, and key Incoming “Pre Order Processing Queue” Fixed Production Lead- “Delay Queue” Finished Orders Time: Standard Goods decisions Capacity ... ... planned: customer is informed Lookahead not planned: customer is informed before placing an order after order has been placed Fixed Production Lead- Time: Accelerated Decisions: Communication? (1) Order acceptance (3) Due date assignment / (2) Due date determination scheduling Information on Workload? 123 Logist. Res. (2011) 3:89–100 91 decision control to a number of agents, while implementing learning objectives are (A) the effect of aligned incentives, suitable incentive mechanisms that align each agent’s indi- communication and information at the interface of sales, vidual goals with overall company objectives. Quantitative and manufacturing on company performance and (B) the performance measures can be complemented by qualitative potentially negative impact of rush orders. measures. However, the effectiveness of system coordina- The game aims at improving system awareness in gen- tion solely by means of monetary incentives is challenged by eral. Taking a process-oriented bird’s eye view on the several factors. The complexity of reality causes contracts to company, it allows recognizing the interdependencies be nearly unavoidably incomplete. Additionally, motivation between time-sensitive demand, order acceptance, and due theories suggest that decision makers are not solely pursuing date determination under capacity constraints and consid- monetary objectives; the decision maker’s attitude, e.g., erable changeover times. Due to the simpliﬁcations, cause- entrepreneurial thinking, can inﬂuence his actions. Lastly, and-effect relationships can be seen, and participants can the transparency of the system, including communication experience how their local actions affect the system as a and information, can have an impact on decisions. whole. Not only awareness for the processes is ameliorated, The manuscript is organized as follows. In Sect. 2,we but also comprehension of the behavior of all game par- discuss the development of the management game with its ticipants is enhanced [4, 15]. learning objectives. Section 3 reports on the design and The key learning is conveyed by contrasting a situation results of a game validation and further includes a cen- with misaligned incentives to one with aligned incentives tralized stochastic dynamic optimization model to bench- in two separate game rounds. With misaligned incentives, mark the game performance. Conclusions and discussion the participants should have the feeling that it is best for are given in Sect. 4. them if they decide independently. This, however, leads to lower the company performance. In the second round, they should feel that it is best to cooperate. This leads to 2 Management game development improved company performance and to a higher bonus of the participants, who thus learn that misaligned incentives We developed a management game for guiding behavior in can be a major cause for inefﬁciencies. Also, participants decentralized systems, speciﬁcally addressing the coordi- see the impact of inappropriate performance measures nation issues at the manufacturing–marketing interface. when, e.g., not utilization per se, but the number of orders delivered in time is the decisive criterion instead. Lastly, 2.1 Management games as interactive learning tools they experience that an effective bonus design can not only make the company better-off but also can increase the reward of each employee. Management games are an increasingly popular method of active learning. The term ‘‘Management Game’’ involves Rush orders often have a negative impact on supply the interaction between some models of a company, which chain costs and operating efﬁciency. However, this impact determines the impact of decision making and human of an incoming order can often not be easily assessed but behavioral elements [5] and the participants. Although less depends on the complex interplay between order charac- information is conveyed, active learning methods excel by teristics and the production system’s state. Regarding the the depth of learning which is increased by personal case study, one core problem is that sales people are not involvement [5]. One key beneﬁt of management games aware of the negative impact that accepting rush orders that are widely used as risk-free learning environments is to might have. In a production setting with restricted capac- understand the interaction between different functional ities and considerable changeover times, understanding areas. Knolmayer et al. [12] provide an overview on freely opportunity cost of changeovers is a vital component: not accessible, interactive learning objects in the area of the changeover induces cost, but the products that could logistics. have been produced and sold had the changeover not been We developed a partial model board game as an inter- carried out. Furthermore, the production of a rush order can active learning tool to be employed during company cause a delay of standard orders. The associated delay cost trainings following the idea of the ‘‘Beer Distribution can outweigh the margin the rush order would contribute to Game’’ as a role model, customized for tackling the case company proﬁts. In the game, optimizing system perfor- company’s problems of suboptimal system performance— mance involves the rejection of some, yet not all, rush misaligned incentives, lack of transparency, and lack of orders. The core learning is that a rush order can naturally system awareness. A board-based game was chosen to still be accepted even if causing frictions, but it is impor- allow for the possibility of including lively discussions tant to carefully judge its impact on system cost. The game between sales and manufacturing; participants should increases the awareness for this trade-off by providing experience conﬂicts and their causes. The intended key information transparency. 123 92 Logist. Res. (2011) 3:89–100 2.2 Design Manufacturing decides which orders they would prefer to accept, these orders are kept (S5). Orders that manufac- For the game design, a relatively low level of complexity turing prefers to reject are passed back to the respective was chosen to allow for robust conveyance of the key sales person (S6). Each sales person makes the ﬁnal deci- learning. This included sacriﬁcing some levels of realism, sion on these orders (S7): they can overrule manufacturing for example, by keeping demand exogenous. Yet, the core by adding a wildcard (one per unit of demand) to the elements of the company situation including the roles of respective order, forcing production (S5); or else may agree sales and manufacturing were mapped. with manufacturing’s preference and reject the order (S8). The game board is depicted in Fig. 2: Part 1 shows the Scheduling The scheduling and production part is divided places of the two sales people, working individually as into 6 horizontal rows, representing 6 periods. These are ‘‘Sales A’’ and ‘‘Sales B’’; Part 2 includes the places of two used to keep track of time. Rows 1 (S9) to 3 (S15) repre- manufacturing employees working jointly together. Sales sent the order pool without a delay, whereas rows 4 (S18) and manufacturing can be separated by a screen. The top to 6 (S19) collect orders beyond the due date, which are corners of each part include a key and a short instruction subject to waiting costs. Production planning ﬁelds (S12– per player (for a larger scale, see Fig. 3). LEGO tokens S14) show the period capacity of 5 units and represent the are used to represent order cards and game tokens. plan to be executed in the current period (Production) and a Demand is differentiated by product type (red or white), look-ahead planning with a horizon of two periods (Plan- order type (standard [gray] or rush order [yellow]), and ning). The latter, however, can still be modiﬁed when new order size (small (one unit) or large (two units)). A standard information about accepted orders becomes available in the order has a desired lead time of 3 periods, a rush order of 1 next period. Manufacturing forwards the order cards period. A speciﬁc demand sequence was constructed to according to type: standard orders to the ﬁrst row (S9), meet the objectives of (1) fair treatment of sales people accelerated orders (following the yellow arrow) to the third who are individually compensated and (2) conveyance of row in the yellow ﬁeld (S10), and issues the corresponding the key learning that acceptance of rush orders can cause raw material (from S11) to the planning ﬁeld(s) (S12–S14). substantial delays. The latter was done by building large Production rules are as follows: between different colors, a lots and adding a rush order of the other color to the changeover (black token) is necessary; after a changeover demand sequence, which would cause changeovers and any color can be produced (‘‘clean machine’’). Each thus delays to orders already accepted. This demand changeover reduces the available capacity by one unit. The sequence was tested and validated in the experiment (see current setup can be seen on the board (S13). Any order in Sect. 3.1). the order pool (rows 1-6) can be produced; however, orders Before the start of the game, participants are instructed have to be produced without interruption. In the planning on game sequence and rules (as explained in the following) ﬁelds, manufacturing plans the production schedule for 3 as well as the framework: the game is played in periods, periods using the game tokens and decides on a production measurement of monetary components is in Thaler. sequence for the current period in the production ﬁeld in Demand consists of 2–3 orders (4–6 units) per period, the third row (S13). The entire production plan can be capacity per period is 5. The game starts in period 1. The revised until the production decision has been made. While initial state of the game includes open orders of previous sales decides on order acceptance including the implicitly periods. Each period is announced by the instructor and determined due date according to order type, manufactur- includes four key steps: ing decides on the schedule, thus determining the realized Incoming customer demand, order acceptance check, and lead time and delivery date. At the end of step 2, manu- decision Both sales persons receive order cards (see S1 in facturing ﬁxes the production schedule. Fig. 2), including information on product type, size, order Production and packaging Orders are (instantaneously) type (including requested lead times), and order size. The produced, the setup marker (S13) is changed to the last players build the corresponding LEGO -order cards (S2) color produced, and the ﬁnished products are attached to and forward the incoming orders to manufacturing (S4), if the order cards (in rows 1–6), which had triggered pro- applicable by traversing the screen (S3). Manufacturing duction; this movement is indicated by the black arrow tries to integrate orders into their planned schedule taking between production and order cards. Changeover tokens into account promised due dates if an order is accepted. For are collected in a bin (S16). a standard order, the planned lead time is 3 periods, i.e., the due date is set equal to the index of the current period plus Shipment Completely fulﬁlled orders in rows 3, as well as 2. For a rush order, the planned lead time is a single period, in rows 4–6, are delivered to the customer (S17). All other i.e., the due date is set to the end of the current period. order cards are forwarded one step as a means to keep track 123 Logist. Res. (2011) 3:89–100 93 Fig. 2 Game board: parts 1 and 2 123 94 Logist. Res. (2011) 3:89–100 Fig. 2 continued 123 order flow orderflow one unit one unit standard order standard order - maximum delay:3 - maximum delay:3 accelerated order accelerated order - maximum delay: 1 - maximumdelay: 1 -accel mportant! - acceler important! wildcards - treat orders wildcards preferentially without - usewhen preferences of riskinglost sales sales and production differ - one card per unit waiting cost - standard order: 1 stone perstep - accelerated order: 3 stones perstep Logist. Res. (2011) 3:89–100 95 Fig. 3 Details of game board: Sales: Key and Instructions Manufacturing: Key and Instructions Key Key key and instructions one unit one unit Order acceptance: a) generate (assemble) incoming orders b) pass incoming orders to manufacturing Production andPackaging: Order acceptance: c) wait for manufacturing’s preferences 1 3 a) receive incoming orders a) setcolormarkeronlast color in d) receive orders manufacturing prefers theproduction field; b) decide on preferences to reject and decide on order - accept: pass orders to manufacture goods acceptance; if order is accepted, b) transfer finished goods to order „incoming orders“ attach wildcard(s) and return order to - reject: pass order back cards; transfer changeover manufacturing; else, pass order to chips to blue bucket c) WAIT for sales’ decision and „rejected orders“ receive „wildcard“-orders Scheduling: Shipment: 2 4 a) forward incoming orders, a) transfer all cards to the next issue raw material field b) schedule / reschedule b)apply delay penalties where production necessary of time, e.g., from row 1 (S9) to row 2, from row 2 to row (costs accrue at the sales department) and costs due to 3. Orders in row 3, which could not be produced in time, overtime or expediting shipments to compensate for are transferred to the waiting rows (e.g., from row 3 (S15) delayed production. Lost sales are orders that were to row 4 (S18) for a standard order, from row 3 (S10) to accepted yet could not be produced within a certain time row 6 (S19) for an accelerated order) and tagged with limit of 6 weeks for standard orders and 2 weeks for rush penalty tokens (S20). After the maximum waiting time orders. Rejected orders are orders which, due to scarce (standard order: 3 periods, rush order: one period), orders capacities, could not be accepted, they do not have a are lost at a penalty cost (ﬁeld ‘‘lost sales’’). After for- negative impact in this short-term game setup. warding the order cards, the game period is ﬁnished. After each period, the instructor collects the ﬁnished 2.3 Benchmark model orders (S17) and sorts the tokens for evaluation in the accounting section (S21). These tokens are the basis for For comparison purposes, the problem for known demand bonus calculation and game evaluation. After all periods can be modeled as discrete time lot-sizing and scheduling have been played, the tokens are counted, the performance problem with setups (see e.g., [9]. The deterministic measures computed and the bonus payments calculated. problem with all demands being known represents the best Finally, the results are communicated to the participants. solution only the instructor is able to obtain. The game Game parameters used are detailed in the benchmark participants only have incomplete information, i.e., 2–3 model, see Sect. 2.3. orders but do not know the distribution between the two For performance measurement, EBIT, deﬁned as reve- products and the extent of rush orders. To determine the nue minus cost, was chosen as the objective. The measure centrally optimal solution for both, a demand sequence proﬁt (EBIT minus bonus payments) was additionally under certainty and uncertainty, we developed a (stochas- computed for the purpose of evaluating game results. tic) dynamic program. In the following, we only show the Revenue is the standard measure from which variable costs deterministic version and brieﬂy sketch the required are deducted to compute the gross margin of sales. extensions for the stochastic version. Wildcards have to be used by a sales person who wants The planning horizon is T = 17 where orders only to have an order produced, which manufacturing preferred arrive in the ﬁrst 15 periods, and the two remaining periods to reject. The rationale behind the cost is additional are used to manufacture waiting orders. In every period, handling cost as well as frictions in scheduling. Utilization there is a limited manufacturing capacity of C = 5. Two (produced units/available capacity) is a measure of products i = 1, 2 are considered and for each product, throughput. Waiting costs include rebates granted to the there exist two types of orders: rush (r) and standard (s) customer for a due date later than their desired due date orders. Rush orders have a unit margin of p ¼ 50 and a Scheduling: wait Production and packaging: wait Shipment: wait 96 Logist. Res. (2011) 3:89–100 due date of 1 week. They can be either small with a we need to determine production quantities x , number of it capacity requirement of one unit or large with a capacity setups u , and the new setup state z . The logic for the it t?1 requirement of two units. The respective demands are setup decision variable is r1 r2 denoted by d and d . Standard orders with demand size it it u it s s d have an unit margin p ¼ 50, the due date is 3 weeks, 1 z ¼ j^ x [ 0; z ¼ j^ x ¼ 0 ^ z ¼ i; z ¼ i^ x [ 0 ^ z ¼ i it i t it t it tþ1 t jt tþ1 and they require 2 units of capacity. Orders that were 0 otherwise accepted but cannot be ﬁnished on time (within 1 week for A setup for product i is required in the following cases: (i) rush orders and 3 weeks for standard orders after the due the machine is initially setup for the other product j and the date) are lost at a penalty cost v = 600. Delayed orders are s r production quantity for i is positive, (ii) the initial setup is subject to waiting cost w ¼ 100 and w ¼ 300 per unit i i for the other product j, i is not produced but the initial setup and unit of time after the due date has passed for standard state in the following period is for i, and (iii) the initial and and rush orders, respectively. ﬁnal setup status are for product i but the other product j is The optimization problem formulation exploits the fol- produced in between. Production quantities and setups are lowing properties of an optimal solution to reduce com- limited by the available capacity of 5 units. plexity: (i) there is at most a single setup operation for each product in a period, (ii) (accepted) rush orders are satisﬁed 2.3.3 State transition with priority before any standard order is satisﬁed, and (iii) orders of any type are satisﬁed ﬁrst-in-ﬁrst-out. Furthermore, The new state of the following period y is a function of the t?1 we assume a pure make to order, zero inventory regime, i.e., current state y and the decisions about order acceptance and no products that have not been ordered are manufactured. In production quantities. We do not show the system of equa- dynamic programming, a simultaneous optimization prob- tions but rather sketch the logic behind the state transition. lem is decoupled into a sequential problem by introducing For both products i = 1, 2, manufacturing quantities x are stages (here periods). At the beginning of every period, it used to satisfy waiting and accepted orders in the sequence previous decisions have resulted in an initial state (here rush orders ﬁrst, then oldest to newest standard orders. In orders of a certain age and the setup status of the machine). case, only a single capacity unit remains, i.e., an order with a Given this state, the optimal decision for the period (con- capacity requirement cannot be satisﬁed in full, manufac- sisting of order acceptance and a production schedule) is turing will be started and completed in the following period determined such that the sum of direct rewards (from (therefore reducing the capacity by one unit). accepting orders) minus costs (for manufacturing, waiting, and lost orders) plus all the costs that result from taking 2.3.4 Functional equations t = 1, 2,…,T optimal decisions in all future periods (given the current periods decision). Next, we describe the dynamic program For each given initial state, the objective is to maximize the by stating the state of the system at the beginning of every expected proﬁt for accepted orders minus costs for waiting period, the decisions to be taken, the state transition, and the and lost orders. The constraints ensure the bounds for functional equations. accepting orders and the manufacturing capacity constraint. 2.3.1 State r 1 2 s max V ðy ; z Þ¼ p ðr þ 2r Þþ 2p s t t t it i it it i i¼1 A state at the beginning of period t is represented by the s s s s r r1 r2 w ðy þ y þ y Þ w ðy þ y i i3 i4 i5 i i i number of waiting rush orders of size one (r1) or two (r2) r1 r2 r1 r2 s vðy þ y þ y x Þ units (y ; y ), standard orders of age j, y ,(j = 1,…,5), and i5 it i i i i ij the initial setup state of the machine z [ {1,2}. Each indi- þ V ðy ; z Þ tþ1 tþ1 tþ1 1 1 2 2 s vidual order pool state variable can take values between 0 s:t: r d ; r d ; s d ; i ¼ 1; 2 it it it it it it and 3. Let y denote the vector of all order state variables. ðx þ u Þ C it it t i¼1 2.3.2 Decision u 2f0; 1g; x ; r ; it it it Decisions to be taken are which of the incoming orders to r ; s 0 and integer, i ¼ 1; 2 it it accept, how many units of each product to produce, and V ðy ; z Þ¼ 0; ðxÞ ¼ max 0; xg Tþ1 Tþ1 Tþ1 1 r1 2 r2 setup changeovers between products. Let r d ðr d Þ it it it it denote the number of accepted rush orders of size 1(2) and The demand data and the optimal decisions are shown in s d the respective accepted standard orders. Further, Table 1. At the beginning of the game, there exist already 4 it it 123 Logist. Res. (2011) 3:89–100 97 accepted orders for product 1 that were accepted in periods 3 Game validation -1 and 0, respectively. Given this initial order pool and that the machine is initially setup for product 1, the optimal 3.1 Experimental design and implementation decisions for each period are determined by forward evaluation of the eventual plans obtained from the We designed a controlled experiment tested with university functional equations. students in order to validate the game’s effectiveness in The columns in Table 1 show the respective demands. meeting its key learning objectives. Figure 4 shows the In the optimal solution, all standard orders are accepted. four treatments, each representing a combination of a bonus payment and the availability of information/com- For rush orders, the numbers in brackets show the number of accepted orders. The optimal solution under full infor- munication. Treatment 1 [T1] is assumed to be the worst mation yields V = 4,000. For the decision problem under case and treatment 4 [T4] the desired solution. uncertainty, every period has several scenarios with In treatments 1 and 2 without information/communica- respective probability and demands. Decisions in every tion, manufacturing utilization was not visible to sales and period have to be detailed by scenario. We assume that only limited information on game parameters was avail- there will be exactly three orders in every period. The split able. After each period, sales were informed about the between the two products is uniformly distributed. Fur- number of delayed orders. With the availability of infor- thermore, with probability 2/3, one of the three orders is a mation and communication (treatments 3 and 4), visibility rush order. In case there is a rush order, the sizes of one or of the whole game board was enabled, and full information two units are equally likely. This in total results in 16 on the game parameters was given to the participants. demand scenarios for each future period. After the reali- Secondly, we manipulated the players’ performance zation of demand in every period, the optimal decision is measurement systems by means of incentive alignment. implemented for the realized demand. The expected value Bonus 1 (shown in Table 2) represents the initial situation using the above assumptions is 3862.57. However, given of misaligned incentives to visualize its resulting problems the realized demand sequence, this results in the same (treatments 1 and 3). optimal decisions. Manufacturing is primarily evaluated based on utiliza- tion. Waiting costs as the inﬂuential factor on company performance are only included with a small weight. Sales are compensated based on individual revenues and com- petes for scarce capacities. As the demand sequence includes rush orders causing changeovers, incentive con- Table 1 Customer orders, acceptance decisions, and manufacturing ﬂicts arise. Each sales person is interested in having each quantities incoming order produced, whereas manufacturing aims at s s r1 r2 r1 r2 td d x x d d d d 1t 2t 1t 2t 1t 1t 2t 2t minimizing the number of changeovers even if this causes delays. Additionally, not the gross margin, but only reve- -12 nues are included in the bonus of sales. This conceals the fact that one waiting step reduces the margin to zero. Also, 13 5 compensating sales primarily on the basis of revenue 2 2 1(0) 5 weighs customer service as a sales’ objective (measured by 3 2 1(0) 5 waiting costs) only insufﬁciently within the bonus. As an 43 5 additional bonus component, each sales person is penalized 5 2 1(1) 3 1 for using a wildcard as a measure of conﬂict. Bonuses of 6 2 1(0) 5 both sales and manufacturing include a penalty on waiting 7 2 1(0) 5 cost. Lastly, lost sales are penalized. 8 1 1(0) 5 Bonus 2 implements aligned incentives as the core pillar 92 1 5 of the solution approach to improve coordination (treat- 10 2 1(1) 2 2 ments 2 and 4). Bonus payments were constructed, con- 11 3 5 sidering both the realities of the case study and the 12 2 1(0) 5 implementation within the game. The bonus for both sales 13 2 1(1) 5 and manufacturing comprises an individual component (as 14 2 1(1) 3 1 for Bonus 1) with a weight of 80% and an overall proﬁt 15 3 5 sharing component with a weight of 20%. As proﬁt 16 5 depends on ﬁnal bonus payments, EBIT was used as a 17 3 measure for the bonus calculations. For the determination 123 98 Logist. Res. (2011) 3:89–100 Fig. 4 Overview of treatments Bonus 1 Bonus 2 incomplete information, Treatment 1 Treatment 2 no communication Impact of conflicting incentives on Impact of improved individual system performance. incentives and a profit-sharing component. complete information, Treatment 3 Treatment 4 communication Impact of complete information and Impact of improved bonus and communication. complete information and communication. Table 2 Details of bonus Bonus 1 Bonus 2 calculation Proﬁt sharing component: weight 20% 5% of EBIT Sales Individual component: 100% Individual component: weight 80% ?5 per sold unit of product ?5 per unit -1 per waiting period standard order -1 per waiting period standard order -3 per waiting period rush order -3 per waiting period rush order -1 per used wildcard -3 per used wildcard -100 per lost order -100 per lost order Manufacturing Individual component: 100% Individual component: weight 80% Initial bonus payment: 180 Initial bonus payment: 180 Target: 96%; ±10 per% utilization more/less -1 per waiting period standard order -10 per waiting period standard order -3 per waiting period rush order -30 per waiting period rush order -100 per lost order -100 per lost order of the individual part, utilization is excluded from manu- 3.2 Results facturing’s bonus and substituted by a 10 times higher penalty on waiting costs. This aligns manufacturing’s Table 3 gives an overview of average key performance objectives with company objectives. Sales’ incentives are measures and bonus payments for all four treatments for aligned through a penalty on wildcards 3 times higher, the ﬁrst 15 periods of the game. The gross margin is rel- which gives them an incentive to adhere to manufactur- atively stable. Note that the corresponding value of gross ing’s decisions. margin in the optimal solution is 3,600 (4,000 minus the revenue of 8 units produced after the horizon of 15 peri- The standard methodology of experimental economics was used (e.g., [4]. Eighty students were recruited as par- ods). Waiting costs show substantial differences, whereas wildcard costs are mostly negligible. EBIT differs between ticipants at the University of Mannheim, mostly graduate business students specializing in logistics. For each treat- the treatments, so does utilization, but to a smaller extent. ment, ﬁve individual sessions were conducted, and a Result I System performance is improved by incentive between-subject design with different participants was alignment as well as by information and communication. chosen. The students were assigned roles as sales (2 per The impact of aligned incentives on system performance is session) and manufacturing (2 per session). Participants larger than the impact of information and communication. were provided with instructions, both written and oral. Better company performance is reﬂected in higher bonus Instructions contained information on the bonus payment payments. including how their performance in the game (measured in First, the inﬂuence of bonus design is analyzed. Com- Thaler) translated into real monetary payments (in Euros) after the game. Compensation of the students consisted of a paring T1 and T2 yields signiﬁcant differences (p = 0.008) for all values but gross margin; T3 and T4 are signiﬁcantly ﬁxed show up fee of €7.50 and a performance-related bonus with an expected value of €3. different concerning all values except wildcard cost. With 123 Logist. Res. (2011) 3:89–100 99 Table 3 Overview of key performance measures and bonus setting, this result may differ depending e.g., on the total payments amount of the bonus compared to ﬁxed compensation. In order to sustainably implement an aligned incentive Bonus 1 Bonus 2 scheme, the results need to be Pareto-improving, i.e., not T1 T3 T2 T4 only the company but also the employees have to be better- Gross margin 3,380 3,460 3,450 3,560 off. An acceptance of a new policy by employees implies Total cost -2,008 -1,693 -252 -23 that higher bonus payments for all players should follow better company performance, which is conﬁrmed by the Thereof waiting cost -1,920 -1,660 -240 -20 Thereof wildcard cost -88 -33 -12 -3 results: EBIT (i.e., the total of proﬁt and bonus payments) and manufacturing bonus are positively correlated (Spear- EBIT 1,372 1,767 3,198 3,537 man’s q value of 0.931), EBIT and bonuses of the two sales Utilization (%) 90 93 93 95 persons are correlated with values of 0.849 and 0.724, Bonus sales (total) 301 323 332 354 respectively (all signiﬁcant at the 0.01 level, 1-tailed). Bonus manufacturing (total) 206 267 314 356 Additionally, the absolute bonus values of T2 and T4 Proﬁt 865 1,177 2,552 2,828 (Bonus 2) were signiﬁcantly higher than the values in T1. The second result focuses on the analysis of order acceptance/rejection decisions of sales and manufacturing EBIT, being the key measure, Bonus 2 leads to better (see Table 4). In the optimal solution, 34 standard orders results than Bonus 1 for both treatments with and without and 4 rush orders are accepted, whereas 6 rush orders are communication. This result conﬁrms the choice of mone- rejected. tary incentives as the central pillar to guide behavior. The Result II Conﬂict between sales and manufacturing cau- students were inﬂuenced in their behavior both by the ses inefﬁciencies through waiting costs. Acceptance of rush ﬁctional game setting and the intention to ‘‘win’’ and by the orders is a main cause for these inefﬁciencies. Rejection of real monetary pay. the ‘‘right’’ orders improves company performance. Secondly, the inﬂuence of information and communi- cation is reviewed: T2 and T4 are different with regard to In T4, the highest number of orders is accepted in the EBIT (p = 0.016) and gross margin (p = 0.056), however ﬁrst step, indicating the lowest necessity of direct interac- both with a small difference of absolute values. T1 and T3 tion at the interface and therefore little frictions. In T1, are not signiﬁcantly different. The result thus mainly holds manufacturing rejected the largest amount of orders with regard to T2 and T4, implying that information and (rejected orders and orders with wildcards). Comparing T1 communication improve system performance with Bonus and T4, the number of rejected standard orders is signiﬁ- 2. An explanation for T1 and T3 not being signiﬁcantly cantly different (p = 0.008), as well as the number of different is the observation of different risk attitudes: in T1 rejected rush orders (p = 0.032). In T4, no standard orders (without visibility of capacities), sales was often afraid of were rejected. lost sales and hence did reject more orders as compared to The usage of wildcards, which was subject to a cost of 5 T3, where capacities as well as waiting queues were visi- Thaler per unit, is interpreted as follows: orders rejected by ble. Bringing improved information and a better aligned manufacturing could be accepted by sales using a wildcard, incentive system together, T4 yields signiﬁcantly better thus overruling the recommendation of manufacturing. In results compared to T1 in all performance values T1, signiﬁcantly, more wildcards are used than in T4, as (p = 0.032 for gross margin, p = 0.008 for EBIT, waiting well as in T2 (p = 0.008 for standard orders, p = 0.056 for costs and wildcard cost). Bonus payments and information rush orders). With wildcards as a measure of conﬂict, T2 transparency lead to improved coordination of sales and and T4 (Bonus 2) are thus better than the original setting manufacturing. (T1) in this respect. Communication, which was observed When comparing the impact of bonus (T2) with the during the experiment, made wildcard usage unnecessary. impact of information (T3), we get the following results: Lastly, the correlation between order acceptance/rejec- Looking at EBIT and waiting cost, there is a signiﬁcant tion decisions and waiting costs as a measure for produc- difference between T2 and T3 (p = 0.095 and p = 0.032). tion delays, and thus low customer service is analyzed. The This indicates that an improved bonus (T2 to T1) has a number of accepted orders with wildcard as a measure of higher positive impact on total system performance (mea- cross-functional conﬂict are positively correlated with sured by EBIT) as a change in information availability (T3 waiting costs (q = 0.748, signiﬁcant at the 0.01 level to T1). This result shows a successful implementation of a (1-tailed)). The q value of rush orders with wildcards was parameterization that allows to conveying the key learning 0.528 and standard orders with wildcards 0.678, both sig- of the impact of incentive alignment. In a real world niﬁcant at the same level. This indicates that wildcard 123 100 Logist. Res. (2011) 3:89–100 Table 4 Order acceptance decisions incentive alignment by the design of bonus schemes and the role of information and communication. Bonus 1 Bonus 2 T1 T3 T2 T4 Acknowledgments The authors would like to thank the two anon- ymous reviewers for their helpful suggestions to improve the manu- # of accepted orders without wildcard: 22.4 29.4 31.0 34.0 script and Mirko Kremer, The Pennsylvania State University, for his standard input to this project. # of accepted orders without wildcard: rush 3.6 5.8 4.8 3.4 # of accepted orders with wildcard: standard 8.8 2.6 0.8 0.0 References # of accepted orders with wildcard: rush 2.4 1.4 0.4 0.4 # of rejected orders: standard 2.8 2.0 2.2 0.0 1. Bendoly E, Donohue K, Schultz K (2006) Behavior in operations # of rejected orders: rush 4.0 2.8 4.8 6.2 management: assessing recent ﬁndings and revisiting old Subtotal: # of accepted orders: rush 6.0 7.2 5.2 3.8 assumptions. J Oper Manag 24(6):747–752 Subtotal: # of accepted orders: standard 31.2 32.0 31.8 34.0 2. Cachon GP (2003) Supply chain coordination with contracts. In: de Kok AG, Graves SC (eds) Supply chain management: design, Subtotal: # of accepted orders without 26.0 35.2 35.8 37.4 coordination and operations, handbooks in OR & MS, vol 11. wildcard Elsevier, Amsterdam, pp 229–339 Subtotal: # of accepted orders with wildcard 11.2 4.0 1.2 0.4 3. Chayet S, Hopp WJ, Xu X (2004) The marketing-operations Total: # of accepted orders 37.2 39.2 37.0 37.8 interface. In: Simchi-Levi D, Wu SD, Shen Z-J (eds) Handbook of quantitative supply chain analysis: modeling in the e-business Total: # of rejected orders 6.8 4.8 7.0 6.2 era. Kluwer, Norwell, Massachusetts, pp 485–554 All values: average values per treatment 4. Croson R (2002) Why and how to experiment: methodologies from experimental economics. Univ Illinois Law Rev 2002(4): 921–945 5. Elgood C (1990) Using management games. Gower, Aldershot 6. Eliashberg J, Steinberg R (1993) Marketing-production joint usage leads to production delays. The number of accepted decision making. In: Eliashberg J, Lilien G (eds) Handbooks in rush orders with wildcard and waiting costs are correlated operations research and management science. Elsevier, Amster- (q = 0.528, p = 0.01). However, rush orders without dam, pp 625–880 wildcard and waiting costs show no signiﬁcant correlation. 7. Ellinger AE, Daugherty P, Keller SB (2000) The relationship between marketing/logistics interdepartmental integration and Yet, the subtotal of rush orders and waiting costs are cor- performance in US. Manufacturing ﬁrms: an empirical study. related with q = 0.547 (p = 0.01). Again, the parameter- J Bus Logist 21(1):1–22 ization allowed for a conveyance of the intended key 8. Ellinger AE, Keller SB, Hansen JD (2006) Bridging the divide learning of the negative impact of most rush orders in this between logistics and marketing: facilitating collaborative behavior. J Bus Logist 27(2):1–27 speciﬁc setting. Lastly, total order rejection and waiting 9. Fleischmann B (1990) The discrete lotsizing and scheduling costs are not signiﬁcantly correlated, yet rejected standard problem (DLSP). Eur J Oper Res 44(3):337–348 orders and waiting costs show a correlation (q = 0.507). 10. Gunasekaran A, Ngai EWT (2009) Modelling and analysis of Rejected rush orders are correlated with waiting costs built-to-order supply chains. Eur J Oper Res 195(2):319–334 11. Keskinocak P, Tayur S (2004) Due date management policies. In: (q = 0.547); this shows that not the total number of Simchi-Levi D, Wu SD, Shen Z-J (eds) Handbook of quantitative rejected orders is decisive but rejection of the ‘‘right’’ supply chain analysis: modeling in the E-business era. Kluwer, orders. Norwell, Massachusetts, pp 295–334 12. Knolmayer G, Montandon C, Schmidt R (2004) Interaktive Lernobjekte zur Logistik. Wirtschaftsinformatik 46(2):139–151 13. Kouvelis P, Lariviere MA (2000) Decentralizing cross-functional 4 Conclusion and managerial implications decisions: coordination through internal markets. Manage Sci 46(8):1049–1058 Based on an industrial case from the food industry, we 14. Malhotra MK, Sharma S (2002) Spanning the continuum between marketing and operations. J Operat Manag 20(3):209–219 developed a management game to involve decision makers 15. Mentzer J, Stank T, Esper T (2008) Supply chain management in the core conﬂicts at the manufacturing–marketing and its relationship to logistics, marketing, production, and interface, revenue versus utilization maximization. operations management. J Bus Logist 29(1):31–46 Thought as a training tool to support change management 16. Pinedo M (2002) Scheduling. Theory, algorithms, and systems, 2nd edn. Prentice Hall, Upper Saddle River and provide mutual understanding of different functional 17. Porteus EL, Whang S (1991) On manufacturing/marketing areas, we conducted a laboratory experiment with student incentives. Manage Sci 37(9):1166–1181 subjects to stress the importance of empirical game vali- 18. Shapiro BP (1977) Can marketing and manufacturing coexist? dation as real behavior might deviate from standard theo- Harvard Bus Rev 55(5):104–114 19. Talluri KT, van Ryzin GJ (2005) The theory and practice of retical predictions [1]. We found that the developed game revenue management. Springer, Berlin served the purpose of highlighting the importance of

Journal

Logistics Research – Springer Journals

Published: Mar 12, 2011

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Incentive alignment at the manufacturing–marketing interface: Design and validation of a management game

Incentive alignment at the manufacturing–marketing interface: Design and validation of a management game

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Incentive alignment at the manufacturing–marketing interface: Design and validation of a management game

Incentive alignment at the manufacturing–marketing interface: Design and validation of a management game

References (26)

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies