THE NEED FOR A HOLISTIC DEMAND PLANNING FRAMEWORK

Transcription

1 THE NEED FOR A HOLISTIC DEMAND PLANNING FRAMEWORK Abstract: Peter Nielsen Department of Production Aalborg University, Fibigerstraede Aalborg Oest Denmark, peter@production.aau.dk. Kenn Steger-Jensen Department of Production Aalborg University, Fibigerstraede Aalborg Oest Denmark, kenn@iprod.aau.dk. This paper addresses the subject of demand planning and problems in the current hierarchical demand planning approach. This paper contains three elements. First, the need for a new framework for hierarchical demand planning is discussed through a brief literature review. Second, findings supporting the need for a new framework from a case study of two manufacturing firms are presented. Third, a conceptual framework for a holistic demand planning approach addressing the identified issues is introduced. The presented framework offers a new perspective on how to design a generic company s demand planning processes and structures under an assumption of multiple demand planning objectives and performance criteria. Keywords: Demand planning, supply chain planning, best practices & benchmarking. 1 Introduction and paper outline Since Frederic Taylor s (Taylor, 1911) first presentation of scientific management, the field of Operations Management (OM) and Research (OR) has developed into a well defined research area where the general OM/OR-techniques (Johnson and Siskin, 1976; Silver et al., 1998) and techniques for hierarchical demand planning are in place (Bitran et al., 1982; Hax and Meal, 1975; Holt et al., 1960). Since then the field has undergone a large development with the emergence of such paradigms as e.g. consensus forecasting and the monolithic modeling approach. Despite these developments the techniques used to manage demand are to a large extend still based on the concept of mass production presented in Taylor (1911) and taken for granted in e.g. Holt et al. (1960). This paper addresses these issues and focuses on demand planning and the gap between expected performance and realized performance. The paper is structured as follows: Section 2 presents a literature review and discussion of the current state of demand planning theory. Section 3 presents an evaluation of current practices of hierarchical demand planning through a case study. Section 4 presents a discussion of the results garnered from the case study establishes a conceptual framework supporting holistic demand planning in a supply chain context. Section 5 concludes on the state of demand planning and presents avenues of further research. 2 Current state of demand planning in the supply chain context the need for a new framework In the work done by Hax and Meal (1975) and Holt et al. (1960) it is suggested that the best performance of a demand planning process is achieved when using a hierarchical approach. The

2 hierarchical demand planning approach (in e.g. Hax and Meal (1975)) is based on the presumption that the world is simple and stable and only assumes the perspective of Manufacturing Planning and Control (MPC). This limitation was acceptable when mass production for a single market was the norm but not in today s dynamic world of masscustomization and complex buyer-supplier relationships. Although the world has changed a great deal since these works were published, almost all Enterprise Resource Planning (ERP) systems are still based on the relationships presented in Hax and Meal (1975) and Orlicky (1975). The hierarchical demand planning approach has several planning levels and objectives depending on the utilization of the demand plan(s). A consequence of this is that several planning stages (Bitran et al., 1982; Hax and Meal, 1975) might exist. The demand planning approach presented in e.g. Hax and Meal (1975) is based on the underlying assumption that information and plans can be aggregated and separated within the boundaries of a given company or business unit without interference. This is of course a limited perspective on the complexity of the situation of a given company or business unit, something recognized half a century ago by Forrester (1958). Since the publication of these works many things have changed with regards to both decision support and the supply chain relationship situation of companies. Subsequently much attention has been directed to supply chain management (SCM) (Slack et al., 2004). As a result there is an abundance of literature that treats SCM (Beamon, 1999; Ganeshan et al., 1999), but none have presented a holistic framework for demand planning. The consequence is that we are stuck with a hierarchical demand planning approach which takes for granted that information and plans can be aggregated across several business levels or across a supply network without interference. In reality many companies face a situation with dynamic interaction across several levels both vertically within the company s boundaries and horizontally within the supply network. The basic problem is illustrated in figure 1. The left hand side of figure 1 indicates the dynamic and complex external context of the demand planning systems. This complex reality is mapped through a demand forecast and handled in a static and simple hierarchical demand planning structure delivering multiple plans for different control areas (right hand side of figure 1). Since the external context of a company is complex and dynamic while the hierarchical approach is simple and static, there is a discrepancy between the model (hierarchical demand planning approach) and the external context being mapped. So, when applying a hierarchical approach to a complex external context, poor performance of the business system must be expected. The hierarchical approach as such will be expected to perform well when the external context of the planning system is simple and stable, such as in a mass production, high volume and low variance environment. A second important issue regarding forecasting and demand planning, is that focus has been on optimizing a given subsystem to achieve either a higher precision of forecasts (Moon et al., 2003) or an optimal plan with regards to any number of criteria e.g. achieving a certain service-level (Silver et al., 1998). While these goals are important from a limited managerial viewpoint, there is a need for a broader perspective when evaluating the company s performance and designing the demand planning system. This is due to the multiple objectives serviced through demand planning. The aim should be to achieve a proper balance between the company s external context, e.g. supply network relationships, and the internal adaptation of a given business structure. To resolve this issue, a number of researchers and practitioners, including Lapide (2000), have called for a hierarchy of performance measures to give better guidance to

3 both executives and managers, while other researchers such as Folan and Browne (2005) claim metrics are poorly developed for SCM. Furthermore, while the techniques for handling the aggregation and separation of plans are in place, there is no proof of a direct correspondence between the overall performance of the business system and the performance of the subsystem of demand planning. This can be attributed to the fact that the methods presented in e.g. Holt et al. (1960) are based on equaling the minimization of production cost (from a MPC perspective) with the maximization of profit. The perspective and success criteria still dominants demand planning approaches today. The narrow focus on MPC means that other objectives such as logistics and sales planning are neglected. This leads to the provocative conclusion that the existing metrics within the field of OM are not linked to the performance of the business system as a whole. Issue Figure 1: The figure illustrates the stringent hierarchical planning approach opposite the dynamic and often chaotic reality of the system sought controlled. The conclusion must be that there is a need for a holistic demand planning framework treating companies in a more situational manner able to handle multiple objectives in a supply network. Furthermore a framework should include both an internal vertical hierarchical and an external horizontal integrated approach to forecasting and demand planning. This new holistic framework should include new metrics to evaluate the overall performance of the business system from multiple perspectives and not just MPC. 3 Case study To examine the use of demand planning in the Industry, a two company case study (Yin, 2003) has been conducted. The focus of the case study was twofold: To determine the current state of demand planning and to give the foundation for the development of a framework for demand planning. An action research approach (Coughlan and Coghlan, 2002) has been used when evaluating and changing practices and decision structures. The case study was conducted over a period of six months and included numerous meetings with production planners, sales representatives, logistic managers and senior management as well as datamining.

4 3.1 The participating companies structural context The two cases, now designated company A and B, were both manufacturing companies within different business areas. Both were traditionally family-owned and controlled through several generations and both companies produced virtually the same products as they had done for several decades. Both companies were involved in a global supply chain but here the resemblance ends. The cases were chosen because of the large difference between their supply chains and products and because they were thought to represent the standards of demand planning within the Industry Company A Company A has approximately 2500 employees worldwide, revenue of approximately 210 million and 10+ production and sales companies throughout the world. The business area examined within company A produces and sells customized packaging to industrial customers. The supply chain for company A is simple with only a few tiers through the whole chain. The products are engineered to order and they depend wholly on the end product. Despite of this, the products have a very simple Bill of Materials (BOM) consisting of dye and raw material. The product range is limited and consisted of about 150 active products. The products have a very dynamic lifespan and lifecycle as short as two years. The manufactured products have no generic lifecycle or demand pattern since sales depend wholly on the end product and a wide variety of business sectors are serviced. The production process is a continuous process treating raw material into the final form of the packaging. Company A has very few suppliers (>10) since they are in the business of transforming raw material to an end product. Forecasts are produced by a sales support department and are distributed through the company s ERP system. A MRP breakdown takes place in the ERP system and the final responsibility for production scheduling and capacity control is placed on a production planner aided by the ERP system. The control of outbound logistics is placed in the sales support department who simply order logistical support from a logistics department. The rather limited inbound logistics is handled by the production planner based on data from the ERP system. Four people were involved in forecasting and logistics control on a part time basis as well as one full time production planner Company B Company B has approximately 800 employees worldwide, revenue of 95 million, two production facilities and 10+ sales companies throughout the world. The business area examined within company B produces and sells spare parts to automobiles to wholesale customers, who then sell to garages and mechanics. The supply chain of company B is a complex network structure due to many interrelations between customers and suppliers. The products are standardized but the product range is wide with 6000 active products in the catalog. The products have a relatively simple BOM consisting of typically less than 30 components per product. The products to a large degree follow a typical product life cycle (Levitt, 1965) and have a lifespan of 20+ years without change to the product nature. Due to the large product range, the products are at varying stages of their life cycles and demand is changing with varying speed. The major part of the revenue is created by about 600 products (constitutes approximately 50% of the revenue) that have a steady demand, while the remaining 90% of the products have a more sporadic demand. The production process might best be described as flowshop. Forecasts are produced in either the sales or the logistics control department and are distributed through the ERP system in which a MRP breakdown is conducted. Responsibility for final production scheduling and

5 capacity control lies with a production planning department while the outbound and inbound logistics is handled by the logistics control department based on MRP-data from the ERP. At the time of the study company B had a logistics department of 15 people working more or less full time with demand forecasting and logistics control, as well as two production planners working full time with demand planning. 3.2 Forecasting and demand planning practices in the case companies Since the examined companies have different products, supply chains and production facilities they have different approaches and objectives for their demand planning process. Both companies used a hierarchical approach to demand planning, although the process varied between companies. Both companies were interested in participating in the case study mainly through an in-house desire to improve their forecasting processes since these were perceived as ineffective. At the start of the case study neither company used quantitative techniques to forecast demand or capacity Company A In company A forecasts on a stockkeeping unit (SKU) level were used as input to the demand planning process. These forecasts were then used in the ERP-system in a standard MRPbreakdown to determine a startdate for production. This MRP was distributed to a particular production facility where the plans were aggregated and used for a manual overall planning of production capacity. The final step was a separation of this plan to a production schedule in the ERP-system for the shop floor. Company A relied on manual updating plans and forecasts on several planning levels on a daily basis, resulting in as many as 30 demand plans per production facility per planning horizon of 45 days. The need for this constant rescheduling was caused by shifts in customer demand compared to forecasted demand and an internal objective of maximizing production capacity utilization Company B Company B used differentiated planning and control measures depending on the products demand patterns. In company B forecasts of total demand on SKU level was likewise used as input to the demand planning process. The demand forecasts were subsequently aggregated to product family level through a manual process. This plan was used as input to an overall scheduling of the capacity utilization at the production facility. This plan was automatically separated into individual product scheduling and based on this a final MRP breakdown was conducted in the ERP-system. Finally a distribution of inventory to different warehouses was conducted based on distribution keys. Company B relied heavily on large inventories to level capacity utilization and buffer against uncertainty. This meant that demand plans were changed about five times a month compared to a planning horizon of one month and forecasts with three months horizon. 4 Discussion and results To evaluate the current state of demand planning practices and the results expected through traditional frameworks several Key Performance Indicators (KPI s) are examined in the two companies. A number of typical KPI s for the evaluation of the performance of the planning process; forecast error (in the form of Mean Absolute Percentage Error (MAPE)), inventory

6 turnover (annual inventory turnover ratio of finished goods), total production lead time and rescheduling frequency, are evaluated and compared to the more traditional business KPI; profit margin. The number of products is included to give insight into the relative complexity of the examined area. The results are shown in table 1. Company A Company B MAPE >50% <25% Annual inventory turnover >12 5 Total production lead time 1 month 1 month Rescheduling frequency 30 times/planning horizon 5 times/planning horizon Number of products Profit margin 18% 11% Table 1: A short selection of KPI s for the two companies included in the case study. The KPI s presented in table 1 can only be seen as indicators of performance, since a number of factors affecting e.g. the inventory turnover ratio are not included. However the elements included are assessed to be the most significant factors affecting inventory turnover. Contrary to theory the demand planning systems performance on e.g. forecast precision does not directly affect the performance of inventory turnover. More importantly no link between forecast precision and the profit margin can be established. This is a critical point since both companies focused resources on improving the precision of their forecasts while neglecting other objectives of demand planning. While the precision of demand forecasts is important when planning demand, knowing what to forecast and with what horizon was found to be at least as important. The implication is that the resources used on achieving a higher forecast precision might not be warranted when compared with the benefits gained. On paper both companies had a stable hierarchical demand planning process in place, but at the same time both companies frequently (respectively 30 and 5 times per planning horizon) had to change plans, despite to some extend buffering against uncertainty through lead time and inventory. Buffering against all elements of variance and uncertainty would in both companies lead to an inflexible demand planning system and large inventories. Without equaling small inventories to profit, the further build up of stock was not seen as a way to achieve benefits in either company. The overall impression of the companies was that while the demand planning processes were well established and robust in their construction, they did not perform well due to external disturbances and conflicting objectives. This trend was apparent in both companies since their current demand planning approach was subjected to many external influences resulting in an unstable demand planning processes. Furthermore company B in particular had multiple objectives with their demand plan resulting in unsatisfactory performance on some of these. Neither company had significant problems with internal demand amplification since everything was linked through a MRP breakdown of demand. The problems experienced were however attributed to the hierarchical approach s inability to handle disturbances in the aggregation and separation of data as well as the current approach s lack of ability to handle multiple perspectives and objectives.

7 Drawing on experience from the case study, the problems attributed to the lack of success with demand planning can be categorized within two main themes: Structural and processual. The problems associated with demand planning in the two examined companies is caused by the structure of the demand planning system and not by the current processes seen in the companies. The structural problems arise because the current demand planning structure is subjected to external disturbances and is unable to meet the multiple demand planning objectives. As an example both companies experienced customers interfering directly in the demand planning process. While this is to be expected, the resources needed to buffer against these uncertainties, when not in a dominant position in the supply chain, results in a drastic drop in the overall performance of the business system. As such company A had many external factors disturbing the demand planning structure, but due to the limited number of products and planning objectives they were able to change plans manually and adapt. Company B had a more complex demand planning process due to the complexity of the production flow, the number of products to be managed and the number of demand planning objectives. As a consequence they were unable to correct plans to the same extend as company A. The conclusion must be that the system best able to correct for the shortfall of the hierarchical demand planning approach (company A), had a better performance on profit margin and inventory turnover, with the same production lead time and significantly worse forecast precision than company B. Furthermore the resources used to achieve this performance differed greatly in the two companies. Company A had four people involved part time and one involved full time in demand planning compared to almost twenty people involved full time in company B. The remedy for this problem seems to be a holistic demand planning framework. This framework should support the internal gearing of a company in relation to its external context while focusing on multiple objectives and the overall performance of the business system. The conceptualization of the framework can be seen in figure 2. The conceptual framework presented in figure 2 links the internal and external structures and processes. The framework takes a network approach to the supply chain, meaning different parts (business units) of a given company (business node) may interact with any number of different business units in other business nodes. A given business unit might have the role of both supplier and customer of any arbitrary business unit inside and outside its business node. As such the business system of a given business node consists of a number of business units. The behavior of these units can be described through two subsystems: The internal business system structure and the external business context. When designing a demand planning system for a given company, success occurs when there is a match between these systems (illustrated in figure 2 through the KPI/structure/process-wheel). Subsequently the process of demand planning can be coordinated through recognizing the need for multiple perspectives on demand planning due to external context and conflicting objectives of e.g. sales and manufacturing planning. The current hierarchical approach lacks the ability to model these complex relationships in a holistic manner and lacks the metrics to evaluate the performance of such a system. The new concept is based on matching internal and external structures with regards to objectives and exchanging information between planning areas. The framework includes hierarchical demand planning as a possibility, as well as e.g. consensus forecasting or the monolithic modeling approach. The conceptual framework sets guidelines for how external and internal processes should be supported by external and internal structures as well as how data and materials should be exchanged. Finally the framework matches KPI s of e.g. customer preferences and market characteristics to the internal adaptation of these external conditions. A demand planning system adhering to this

8 concept will not have the problems associated with the hierarchical approach since the focus explicitly is on tailoring a demand planning system to multiple perspectives and objectives. Business node Business unit Logistics Production Supply Network Sales dep. Quality control Planning Key Performance Indicators Key Performance Indicators Structural Processual Structural Processual Logistics infrastructure IT-infrastructure Planning hierarchy Organization and culture Physical production facilities and product nature Internal Dataexchange Demand planning and forecasting Purchasing and supplying Production and logistics KPI s KPI s Internal External Up- and downstream market dynamics IT-infrastructure Logistics Political and legislative structure External Dataexchange Coordination Figure 2: Holistic framework for design and evaluation of a demand planning system within a supply chain context. So, based on this it is possible to define a demand planning system structure that includes the elements given in figure 2 and given this, a number of deductions can be made. First, the process(es) and structure(s) can be evaluated independently. Second, a given process or structure will perform better than others along a number of KPI s under a given external context. Third, the KPI s can be used to evaluate an arbitrary demand planning system and can be used to benchmark business systems against each other. Fourth, the business system with a better match between external context and demand planning system structure will outperform business systems with inferior matches. This is an important point since it implies that best practice is situational and unique rather than generic and universal and still leaves room for the evaluation of business systems. Furthermore this brings the cost/benefit relationship to the forefront of the discussion when re-/designing demand planning systems. 5 Conclusion and further research The literature review reveals a lack of focus on the problems associated with using a hierarchical demand planning approach. The approach is well suited for planning and control in a mass production environment but does not match the dynamics governing most companies today. Furthermore the review reveals the metrics used, to evaluate the performance of the demand plan in a supply chain context, fail when applied to evaluate the performance of the business system. Planning Logistics

9 The conclusion is that there is a need for a holistic demand planning framework with new metrics. To test the conclusion reached through the literature review a case study of two companies demand planning practices was conducted. Based on the case study a number of conclusions can be reached. First, although the hierarchical demand planning approach used in both companies ought to be stable it was unstable and resulted in e.g. constant rescheduling of production plans. Second, the instability of the planning structures could be attributed to the dynamics of the external context and the inability of the demand planning approach to handling multiple perspectives and objectives. Third, when comparing the performance of the traditional KPI s for demand planning performance with the overall economic performance, no direct link was found. Fourth, the company able to circumvent the plans imposed by a hierarchical approach with fewest resources expended achieved the best overall performance with regards to both inventory turnover and profit margin. Based on this, the need for a framework supporting the gearing of the business system, with demand planning as an important integrated part, is apparent. This supports the need for a holistic demand planning framework and new metrics to evaluate the performance of demand planning structures with multiple objectives. Although not fully developed, the holistic framework presented in figure 2, seems better able to include the elements that in the end determine whether or not a given business system is successful. The further development of the framework is seen as an important step towards fully integrating demand planning in the business system and supply chain. The framework is based on success being situational and given by matching the internal structure with the external context of a company. This means that traditional measurements of success such as forecast precision or inventory turnover should not to be considered primary criteria of success for a demand planning system. As such the principal of best practice no longer apply, since best practice now per definition is a unique situational configuration of the business system. This paper is the first step in the development of a holistic normative demand planning framework and as such further research is needed to give a fully developed framework. The first step is to extend the framework to include a more complete overview of demand planning processes and structures and of the external context of companies. The second step is to fully develop the relationship between the external and internal structures, processes and KPI s. The third step is to quantify the variables included in the framework through case studies and empirically test the relationships. 6 References Beamon, B. M., Supply chain design and analysis: Models and methods. International Journal of Production Economics, Vol. 55, pp Bitran, G. R., Haas, E. A. and Hax, A. C., Hierarchical Production Planning: A Two-Stage System. Operations Research, Vol. 30, pp Coughlan, P. and Coghlan, D., Action research for operations management. International Journal of Operations & Production Management, Vol. 22, pp Folan, P. and Browne, J., A review of performance measurement: Towards performance management. Computer in Industry, Vol. 56, pp

10 Forrester, J. W., Industrial Dynamics a major breakthrough for decision makers. Harvard Business Review, Vol. 36, pp Ganeshan, R., Jack, E., Magazine, M. J., & Stephens, P., A taxonomic review of supply chain management research. In S. Tayur, R. Ganeshan & M. Magazine, Quantitative models for supply chain management (pp ). Boston, MA: Kluwer Academic Publishers. Hax, A. C. and Meal, H. C., Hierarchical Integration of Production Planning and Scheduling. In Geisler, M. A. (Ed.), Studies in Management Scinece, Vol 1, Logistics, Amsterdam pp Holt, C. C., F. Modigliani, J. Muth and H. Simon, Planning Production, Inventories and Work Force. Englewood Cliffs, N.J., Prentice-Hall. Johnson, R. D. and Siskin, B. R., Quantitative Techniques for Business Decisions, Prentice-Hall, Lapide, L., The Measures of Supply Chain Performance. Supply Chain Management Review, Vol. 4, pp Levitt, T., Exploit the Product Life Cycle, Harvard Business Review, 43, pp Moon, M. A., Mentzer, J. T. and Smith, C. D., Conducting a sales forecasting audit. International Journal of Forecasting, Vol. 19, pp Orlicky, J., Material Requirements Planning, McGraw-Hill, New York, Silver, E. A., Pyke, D. F. and Peterson, R., Inventory Management and Production Planning and Scheduling, 3 rd edition, John Wiley & Sons. Slack, N., Lewis, M. and Burns, H., The two worlds of operations management research and practice Can they meet, should they meet?. International Journal of Operations & Production Management, Vol. 24, pp Taylor, F., The Principals of Scientific Management. Harper & Brothers, New York and London. Yin, R., Case Study Research Design and Methods, 3 rd edition, Sage Publications. 7 Biography Peter Nielsen is Ph.D. student at Department of Production, Aalborg University where he also received his M.Sc. in Industrial Management. The Ph.D. project concerns the development of a demand planning framework. Focus research areas are demand planning and demand forecasting. Kenn Steger-Jensen is Associate Professor at Department of Production, Aalborg University (AAU), where he also received his M.Sc. and Ph.D. The Ph.D. project was a strategic collaboration between ORACLE, a leading software vendor of APS systems, and Department of production (AAU). Focus research areas are within Advanced Planning and Scheduling Systems (APS) and Supply Chain Integration.