PRODUCTION PLANNING AND ACTUAL DECISIONS: AN EMPIRICAL STUDY

Production Planning And Actual Decisions: An Empirical Study Jan C. Fransoo Vincent C.S. Wiers (* Eindhoven University of Technology, Department of Technology Management, Subdepartment of Operations Planning, Accounting and Control P.O. Box 513, NL-5600 MB Eindhoven, The Netherlands j.c.fransoo@tm.tue.nl, v.c.s.wiers@tm.tue.nl *) corresponding author Abstract Production planning decisions are usually made by human planners that are assisted by decision support systems. While it is widely argued in the literature that current decision support systems for production planning are generally inadequate, it is not clear to what extent human planners actually disregard the planning decisions proposed by the system. In this study, we investigate this question. In a setting in which the planning system’s model has an adequate representation of reality, we collect data on actual planning decisions and compare them to the planning decisions proposed by the system. We conclude that planners systematically and largely neglect the system's recommendations and that the extent of neglect is larger if the planning problem is more complex. 1 Keywords: production planning, human decision making, ERP systems. 1 Introduction Production planning systems have been deployed widely in the last three decades, with extensive presence of ERP systems. Although the presence and use of the system has been accepted in industry, the academic literature generally argues that the basic replenishment, lot sizing and production planning logic in ERP systems is incorrect and insufficiently represents the stochastic characteristics of the manufacturing system. The actual usage of the planning logic in ERP systems has not been investigated empirically, and in this study we intend to provide some initial and tentative insights into this phenomenon. We study a chemical company with a very simple material and resource structure that can be fully represented in a resource constrained single level MRP system. Data have been collected during a period of three months. During this period, both the planned orders proposed by the ERP system and the actual orders planned by the production planners are recorded and compared. Our results support the general notion that planners largely disregard the planned orders proposed by the ERP system, and that this effect is further enhanced if the complexity of the planning problem is increased. 2 Previous studies Although an enormous part of worldwide IT budgets is spent on implementing and maintaining ERP systems, there is not much research on the use of ERP systems by human planners. 2 Vollmann, Berry & Whybark (1992) discuss bottom-up replanning as a way to change the production plan that is generated by MRP. The objective of replanning the planned orders is to solve capacity problems, as these are not taken into account into MRP. However, it is assumed that the planner firstly sets the planning parameters in MRP correctly so that the replanning effort can be kept to a minimum. This latter part can be referred to as top-down planning. Empirical research also shows that in using ERP systems, there appear to be two strategies to change the output of the system: 1) changing the planned orders that are generated by the MRP algorithm, and 2) changing the system parameters so that the system generates a better output. Den Boer (1992) and Zoryk-Schalla et al. (2004) both discuss these two ways to change a production plan, to deal with capacity problems. Den Boer refers to ad-hoc changes and tactical changes: Ad-hoc changes are changes in already placed orders, borrowing materials from other stock points or renegotiating with suppliers in the hope of getting materials earlier. Tactical changes are made by setting the value of parameters in the ERP system, e.g. the ordering method, frequency of ordering, safety time and safety stock. Zoryk-Schalla describes the implementation and use of Advances Planning and Scheduling (APS) systems. She refers to Schneeweiss’s (1999) Aspiration Levels (AL) that can be set in the APS systems and which determine the output generated by these systems. However, given a specific Aspiration Level the planner may still find the output of the APS system unsatisfactory and decide to directly change the production plan. Hence, planners appear to resort to bottom-up replanning. The main finding of Den Boer also is that planners do not change the tactical parameters in the system; instead, they change the proposed orders of the system directly. In his research findings, there appeared to be no link between the main system parameters safety time and safety stock and the delivery performance of the plan. 3 Given the finding of previous empirical research that bottom-up replanning is preferred by planners instead of changing tactical planning parameters, the question can be asked what determines the use of ERP generated output by human planners. The more general question of why humans prefer to perform a task manually instead of trusting a decision support system, given the fact that cognition is bounded and that techniques can help humans to increase performance, is discussed by Kleinmuntz (1990). A common explanation is that people are unwilling to trust techniques they know are imperfect. Possibly erroneously, people also believe that increased mental effort improves performance. According to Kleinmuntz, this is particularly true for situations where they are confident about their expertise. The issue of trust in automation has also been studied by Muir (1994) and Muir & Moray (1996). The former paper presents a theoretical model of human trust in machines. In the latter paper, two experiments are reported to examine operators’ trust in and use of automation in a simulated supervisory process control task. Results showed that operators’ ratings of trust were mainly determined by their perception of its competence. Trust was reduced following any sign of incompetence in the automation, even one which had no effect on overall system performance. Another finding of Muir & Moray’s experiments is that operators’ trust changes very little with experience; whereas Kleinmuntz concludes that the use of decision aids decreases with the subject’s belief in his experience. Lee & See conclude that trust appears to be an important criterion in the use of systems in situations where uncertainty and complexity make an exhaustive evaluation of options impractical. Complexity and uncertainty are often associated to planning tasks. Trust is likely to influence reliance on complex, imperfect automation in dynamic environments that require the person to adapt to unanticipated circumstances (Lee & See, 2004). 4 Based on this reasoning, we define the following hypotheses: H1: Planners prefer to manually plan production in favour to using the output generated by ERP systems H2: The number of orders manually created by the planner increases when the complexity of the planning problem increases. 3 Planning environment In this study, we are considering one business unit of a large chemical company. The business unit operates seven plants across Europe. Each of the plants consists of a number of manufacturing units (reactors). The manufacturing units produce a variety of products. The manufacturing units that are part of our study produce a chemical product which is mostly delivered in bulk (trucks or containers) to industrial customers. Although theoretically the variety of products is extremely large (due to the possibility of blending in specific additions), in reality the number of products on each of the resources is fairly limited (typically around one dozen, with a few resources producing up to around three dozen products). The planning of the reactors has been centralized in a European Central Planning Office. All customer orders are received at the European Customer Service Centre that is located in the same building. The order handlers are located in the office adjacent to the office where the planners are located. There are four planners involved with planning the reactors, and one additional planner is responsible for planning the shipments from the plants to the customers. For each of the plants, one primary planner has been allocated. Also, for each of the plants, one of the other planners acts as a secondary planner, who can replace the primary planner in case of absence. Each of the planners has primary responsibility for one or two plants, and also carries secondary responsibility for one or two plants. Planners differ in their experience. 5 The production process and the role of the planner is depicted in Figure 1. We consider the planning process to the extent to which the planner influences it. The planners use an ERP system (deploying Material Requirement Planning logic, see, e.g. Silver et al., 1998, for an explanation on MRP logic) that generates planned orders. Production planning as carried out by MRP is the process of converting customer orders with a specific due date and quantity into production orders with a specific manufacturing time and quantity. A planned order can be viewed as an advise of the planning system which product to manufacture on which date and in which quantity. These planned orders are generated based on actual information regarding customer orders (actual orders placed by customers), demand forecasts (expectations of future orders made by someone else in the planning office), current inventory levels, required safety stock levels, and lot sizing rules. Customer orders and demand forecasts cannot be influenced by the planner and should be considered exogenous. Safety-stock levels and lot sizing rules are system parameters that can be set by the planner. Hence, the planner is able to change the parameters that drive the generation of planned orders. However, the planner can also choose to change every planned order individually. Note that the implementation of MRP logic in this ERP system is capacity constrained by forward loading, and that the bill-of-material consists of one level. Every night, a new set of planned orders are generated by the system, based on the current inventory levels, the parameter settings for safety stocks and lot sizes, new customer orders that have arrived on the preceding day, and the available production capacity. New demand forecasts can be supplied every week, but in practice are only updated into the system once a month. Figure 1 about here 6 Each morning, when the planner arrives, he sees a new list of planned orders, as well as all realized (produced) production orders on the previous day, and any manufacturing orders that have already been scheduled by him on previous days, but which have not yet been completed (either due to delay – which rarely happens – or due to the fact that the order is scheduled to be executed at a later date). The planner then needs to decide on the manufacturing orders. Using the planned orders as a suggestion, but using additional information such as inventory levels and available capacity, he needs to determine which orders actually to produce. In doing so, he can either: • convert a planned order directly into a manufacturing order • modify the planned order, and then convert it into a manufacturing order • create a new manufacturing order without a planned order being present • delete a planned order The planner will work through the planned orders until he has processed all of them with a specific time horizon (which he can determine himself). When he has allocated all manufacturing orders to a specific time slot, his work for this specific plant is finished for the day. He may also modify any manufacturing orders that had been created on previous days. This may include moving the order backward in time (postponing the execution), bringing the order forward in time, changing the order duration (and hence the lot size), and by changing production start times the planner may change the sequence of two or more orders. For a number of days, varying between roughly 2 and 12, manufacturing orders have been assigned by the planner to a particular reactor for a particular period. This is called the manufacturing schedule. Overnight, all manufacturing orders that are scheduled for 7 the following day only, are released to the specific plant. The people in the plant execute the schedule according to specification. The information flow for the planner that is supported by information systems is depicted in Figure 2. Figure 2 about here Both the input and results of the manual planning task are captured by the information system, making it possible to retrieve the data. Every day, the following information is available from the ERP system: • complete list of planned orders • complete manufacturing schedule (list of manufacturing orders assigned to a specific reactor in a specific time slot) • complete list of production orders that have actually been executed 4 Analysis and results Since ERP systems register all data that the planner uses as input and generates as output, they provide a potentially rich source of data for analysis of the behaviour of planners (Fransoo & Wiers, 2006). A disadvantage of these systems is that they do not keep track of the actual decision making process; only the results of the decision making process are captured. Also, ERP systems do not keep track of history, other than the financial history and possibly the actual production. Furthermore, systems do not maintain different versions of a production plan. Analyzing series of decisions by production planners thus requires a full download of the data at regular time intervals before they are overwritten or dumped. To investigate system’s use by human planners, we are interested in the difference between the orders as proposed by the ERP system and the orders that are generated by 8 the planners. Production orders can be either generated from ERP planned (proposed) orders, or they can be created manually. We expect that human planners will ignore most of the ERP output and create most production orders manually. Moreover, we expect that increasing complexity of the planning problem will increase the manual portion of the production plan generation. A list has been generated with all planned orders and production orders that were planned during the measurement period. This resulted in a list of 6224 planning entries, of which 3284 are unique orders. Every entry in the production plan is either of the following cases: I. A production order that has been created from a planned order II. A planned order that has not been used to create a production order III. A production order that has been created without a planned order Table 1 below shows the number of plan entries for every case. Table 1 about here Table 1 clearly shows that most production orders are created by the planner without a planned order. Moreover, most planned orders are not transferred to a production order. This means that most of the output of the MRP algorithm is ignored by the planner and the bulk of the production is specified by the planner. Therefore, we cannot reject H1. It has also been investigated if the extent to which the planner manually creates production orders depends on the complexity of the planning situation. In other words, if the number of article codes for a specific production line is higher, would there be more manual creation of production orders. The number of orders per category is visualized in Figure 3. Figure 3 about here 9 Figure 4 shows the number of products per work centre. The correlation coefficient of the category III orders and the number of unique article codes sorted by work centre is .856 (< .0005). Therefore, we also cannot reject H2. Figure 4 about here 5 Conclusion and discussion In this paper, we have made an analysis of the decisions of a number of production planners. We were specifically interested in the extent to which they accept the suggested production orders by the ERP system in place. Based on our analysis, we conclude that in this factory planners largely disregard suggestions from the ERP system. The planners could have partly avoided this by adjusting the parameter settings of the system, but they clearly choose to use a bottom-up planning strategy. The question can be asked why planners do not use the parameter settings in systems to change the output of the algorithms. The study of Den Boer (1994) showed the same results, and he attributes this to the fact that planners scarcely receive feedback on a tactical level. Therefore they do not make an attempt to improve the tactical system settings. This is an often experienced problem with ERP systems, which are standard software packages that need to be configured and parameterised (Stevenson et al, 2005). Moreover, it is not clear whether all the changes made by the planner really improve the performance of the plan. It may also be the case that by making many changes, the planners get the feeling that they are in control, disregarding the implications of their behaviour on actual performance. It is an interesting observation to see the planners changing individual orders frequently, because in other studies, it is stated that there are ‘natural’ deviations in any production system and that regardless of corrective actions, performance will return to the average level all the same (Stoop and Wiers, 1996). 10 Unfortunately, in this case study no data was available on the performance of the production units. Future research is needed on the contribution of the human planner on system performance. The same case study was used in Fransoo & Wiers (2006) to investigate the relationship between the number of actions and the action variety of planners. It was concluded here that an increased number of actions leads to an increased action variety, because the increase in number of actions was associated with an increase in exceptional situations in the planning environment, and this again made a more top-down decision making strategy necessary. However, there is an important difference between the use of topdown and bottom-up between these two cases: whereas in Fransoo & Wiers (2006) topdown decision making is associated by regarding the problem as a whole and designing more complex solutions (indicated by more action variety), an interesting finding of the study in this paper is how the solution is implemented, i.e. by not changing system parameters and trusting the system to come up with better solutions, but by changing most orders manually. In other words, in this case study, a top-down decision making strategy is implemented by using a bottom-up approach – changing the orders one-byone in the context of an overall purpose. 6 References Boer, A.A.A. den, 1992, Integration of Information in Logistic Operations. Proceedings of the IFIP WG 5.7 Working Conference: Integration in Production Management Systems, Eindhoven, August 24-27, pp.232-246. Boer, A.A.A. den, 1994, Decision Support for Material Planners. Production Planning & Control, 5, (3), pp. 253–257 11 Fransoo, J.C. & Wiers, V.C.S. 2006, Planners’ Action Variety: a field study using the daily planning data at a chemical company. Journal of Operations Management, 24, (6), 813-821. Hoffman, P.J., 1960, The paramorphic representation of clinical judgment. Psychological Bulletin, 57 (2), 116–131. McKay, K.N., 1992, Production planning and scheduling: A model for manufacturing decisions requiring judgement, (Ph.D. Thesis University of Waterloo). Schneeweiss, C. 1999, Hierarchies in Distributed Decision Making (Springer, Berlin). Silver, E.A., D.F. Pyke, Peterson, R., 1998, Inventory Management and Production Planning and Scheduling (John Wiley & Sons, New York) (3rd ed.). Stevenson, M., Hendry, L.C. & Kingsman, B.G., 2005, A review of production planning and control: the applicability of key concepts to the make-to-order industry. International Journal of Production Research, 43, (5), 869–898 Stoop, P.P.M., & Wiers, V.C.S., 1996, The Complexity of Scheduling in Practice. International Journal of Operations and Production Management, 16, (10), 37– 53. Wiers, V.C.S., 1996, A quantitative field study of the decision behaviour of four shop floor schedulers, Production Planning & Control, 7, (4), 383–392. 12 Planned Orders Forecast human planner MRP Production Orders Sales Inventory Status stock customers production line Figure 1: Planning & Control Structure Stock replenishment (MRP) Actual inventory Forecast Customer Orders Planned orders Human Planning Task Production orders Actual production Figure 2: Information flow MRP and human planner 120% 100% 80% III II I 60% 40% 20% 0% R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 Figure 3: Number of orders per category by work centre 13 Product Variety 14% 12% 10% 8% Product Variety 6% 4% 2% 0% R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 Figure 4: Product variety per work centre 14 Table 1: Plan Entries Planned order Yes No I III 359 1532 Yes Production order II No 1393 15