Springer advanced planning in fresh food industries integrating shelf life into production planning

Matthias Lçtke EntrupAdvanced Planning in Fresh Food Industries Integrating Shelf Life into Production Planning With 63Figures and 31 Tables Physica-Verlag A Springer Company... It turns

in Fresh Food Industries

R Flavell (Ed.)

Modelling Reality and Personal

Modelling

1993 ISBN 3-7908-0682-X

M Hofmann/M List (Eds.)

Psychoanalysis and Management

1994 ISBN 3-7908-0795-8

R.L D'Ecclesia/S.A Zenios (Eds.)

Operations Research Models

Modelling Techniques for Financial

Markets and Bank Management

WHU Koblenz ± Otto Beisheim

Graduate School of Management

2001 ISBN 3-7908-1366-4

C Zopounidis (Ed.) New Trends in Banking Management

2002 ISBN 3-7908-1488-1

U Dorndorf Project Scheduling with Time Windows

2003 ISBN 3-7908-0033-3 H.M Arnold

Technology Shocks

2003 ISBN 3-7908-0051-1

T Ihde Dynamic Alliance Auctions

2004 ISBN 3-7908-0280-8 M.J Thannhuber The Intelligent Enterprise

2005 ISBN 3-7908-1555-1

C Clarke Automotive Production Systems and Standardisation

2005 ISBN 3-7908-1578-0

Matthias Lçtke Entrup

Advanced Planning

in Fresh Food Industries

Integrating Shelf Life

into Production Planning

With 63Figures and 31 Tables

Physica-Verlag

A Springer Company

Werner A Mçller

Martina Bihn

Diss., TU Berlin, D83

ISSN 1431-1941

ISBN-10 3-7908-1592-6 Physica-Verlag Heidelberg New York

ISBN-13978-3-7908-1592-4 Physica-Verlag Heidelberg New YorkCataloging-in-Publication Data applied for

Library of Congress Control Number: 2005927952

This work is subject to copyright All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illus- trations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Physica-Verlag Violations are liable for prosecution under the German Copyright Law.

Physica-Verlag is a part of Springer Science+Business Media

Softcover design: Erich Kirchner, Heidelberg

SPIN 11493334 88/3153-5 4 3 2 1 0 ± Printed on acid-free and non-aging paper

During the last three decades trade, industry and also academia became heavily involved in the development primarily intended towards more effective planning and control of logistical operations in supply chains Lately, these approaches be-gan to be directed towards fresh food supply chains Competitive fresh food sup-ply chains require that the aspects costs, quality, food safety and technology be taken into account simultaneously in a multidisciplinary way In recent years the issue of food safety got large emphasis in government, industry and society (con-sumers) The introduction of the General Food Law by the EU from January 2005

on even put more emphasis on the issue of food safety

It turns out that Advanced Planning and Scheduling Systems (APS) can play an important and integrative role in supporting decision making activities in fresh food supply chains by considering shelf life as an instrument to generate more added value and food safety Basically the work of Matthias Lütke Entrup is con-centrated on two research questions:

x Which requirements must APS systems cover in order to efficiently and effectively support production planning in fresh food industries?

x How can shelf life be integrated into production planning? How can production planning contribute to optimizing shelf life output?

In his study the author shows how these questions should be answered quately His results and conclusions are of paramount importance for integrating the issue of shelf life into production planning

ade-The study provides a wealth of insights and results which are significant both from a practical as well as from an academic point of view The research starts with an overview of current APS systems and highlights the need of a new genera-tion of planning software which aims at supporting decision making in supply chain management Although APS gain increasing acceptance in industry, a num-ber of issues remain, in particular at the detailed planning and scheduling level, which are not satisfactorily covered by the decision models to be found in the standard APS software packages This is truly the case for the fresh food indus-tries Undoubtedly, the most important planning issue regarding fresh food lies in the consideration of shelf-life So far, vendors of APS systems have taken many efforts to consider shelf-life issues in their planning systems, however, without covering all of the characteristics being important in Fresh Food Supply Chains (FFSCs) and fresh food production systems One of the main contributions of the study by Matthias Lütke Entrup is a comprehensive analysis of the planning re-quirements of fresh food industries on one hand and the decision support offered

by typical APS systems on the other Software packages from leading players in the market are assessed looking at the scope of shelf life integration and its capa-bilities to generate plans that optimize shelf life output

Based on the shortcomings of current APS systems, new quantitative planning models are developed and resolved These models consider shelf life planning problems in specific fresh food industries (yogurt production, sausage production and poultry processing) The models are based on the general block planning prin-ciple and are adapted to the needs of the specific fresh food planning applications Considerable care has been taken to obtain compact model formulations which can be solved very efficiently by use of standard optimization software Numerical experiments demonstrate the applicability of the planning models in realistic in-dustrial settings

As a result, the author makes clear that suppliers of APS software are currently unable to offer APS systems in which the integration of shelf life into production planning has been dealt with adequately Specifically, product freshness has been modeled by the author as part of the optimization and not as a constraint within the planning function This is indeed a new and creative contribution of Matthias Lütke Entrup to solving complex planning problems of considerable practical relevance The applications (case studies) have been selected carefully by the au-thor in such a way that many other application fields in fresh food industries could benefit from his results

Prof Dr Paul van Beek

Prof Dr Hans-Otto Günther

This research could not have been written without the support of many people Therefore, I would like to thank a number of them for their support and contribu-tions, knowing that the list is, of course, incomplete

First of all, I am indebted to my academic advisors Professor Dr Hans-Otto Günther of the Chair of Production Management at the Technical University of Berlin and Professor Dr Paul van Beek of the Operational Research and Logistics Group at the Wageningen University (NL) Professor Dr Hans-Otto Günther woke my interest in the field of Production Management and helped me to trans-form my ideas into a full research project Similarly, I am thankful to Professor

Dr Paul van Beek for his supervision of the work and his critical comments Working with both of them was a pleasure, they have always been accessible and created a stimulating research environment Additionally, I thank Professor Dr Kasperzak for assuming the chairmanship of the promotion committee

I would also like to thank the entire team of the Chair of Production ment consisting of Hanni Just, Dr Martin Grunow, Matthias Lehmann, Ulf Neu-haus, Martin Schleusener, and Onur Yilmaz for their helpfulness and the fruitful discussions Their comments proved to be very useful and resulted in several im-provements In addition, I am grateful to Thorben Seiler and Shuo Zhang for their support regarding the development and implementation of the models

Manage-Furthermore, I thank my employer A.T Kearney for the possibility to conduct this research and the continual support In particular, I highly appreciate the con-tributions of Dr Antje Völker, Jan van der Oord and Ferdinand Salehi as well as

of Dr Peter Pfeiffer and all other colleagues of the Consumer Industries and Retail Practice Dr Marianne Denk-Helmold and Judith Siefers deserve a special thanks for carefully reading and correcting the manuscript

The last words are dedicated to my family I thank my parents for their agement and their trust in me during all the years Finally, I thank Kathrin for her backing and her care She made me realize that there are other things in life than yogurt, sausages and poultry

Foreword V Acknowledgement VII Abbreviations XIII

1 Introduction 1

1.1 Introduction to the Field of Research 1

1.2 Research Objectives 2

1.3 Dissertation Outline 3

1.4 Conclusion 4

2 Advanced Planning and Scheduling Systems 5

2.1 Evolutionary Path of APS Systems 5

2.1.1 MRP I and MRP II 5

2.1.2 Assessment of the MRP Planning Concepts 8

2.1.3 Emergence of APS Systems 9

2.2 Structure of APS Systems 12

2.2.1 Overview 12

2.2.2 Strategic Network Design 14

2.2.3 Demand Planning 15

2.2.4 Supply Network Planning 17

2.2.5 Production Planning 18

2.2.6 Production Scheduling 19

2.2.7 Distribution Planning 20

2.2.8 Transport Planning 21

2.2.9 Available-to-Promise 21

2.3 APS Systems Market Overview 23

2.3.1 Available Market Studies 23

2.3.2 Market Size and Segments 24

2.3.3 Major Providers 25

2.3.4 Expectations for the Future 27

2.4 Implementation of APS Systems 27

2.4.1 Implementation Process Overview 27

2.4.2 Project Definition 28

2.4.3 Vendor Selection 30

2.4.4 Implementation 31

2.4.5 Implementation Risks 32

2.5 Assessment of APS Implementations 33

2.5.1 Benefits 33

2.5.2 Development Needs 34

2.6 Conclusion 35

3 Fresh Food Industries 37

3.1 Introduction 37

3.2 Definition and Segments 37

3.3 Characteristics of Fresh Food Supply Chains 38

3.3.1 Structures of Fresh Food Supply Chains 38

3.3.2 Economic Characteristics and Developments 41

3.3.3 Technological Characteristics and Developments 47

3.3.4 Social/Legal Characteristics and Developments 50

3.3.5 Environmental Characteristics and Developments 53

3.3.6 Summary 57

3.4 Characteristics of Fresh Food Production Systems 58

3.4.1 Overview 58

3.4.2 Formulation 59

3.4.3 Processing 60

3.4.4 Packaging 61

3.4.5 Storage and Delivery 62

3.4.6 Summary 63

3.5 Case Study 1: Yogurt Production 64

3.5.1 Market Segments and Case Study Overview 64

3.5.2 Raw Milk Collection 67

3.5.3 Raw Milk Preparation 69

3.5.4 Fermentation 70

3.5.5 Flavoring and Packaging 71

3.5.6 Storage and Delivery 72

3.6 Case Study 2: Sausage Production 72

3.6.1 Market Segments and Case Study Overview 72

3.6.2 Input of Ingredients 75

3.6.3 Grinding and Mixing 76

3.6.4 Chopping and Emulsifying 76

3.6.5 Stuffing and Tying 76

3.6.6 Scalding 77

3.6.7 Maturing and Intermediate Storage 78

3.6.8 Slicing and Packaging 78

3.6.9 Storage and Delivery 79

3.7 Case Study 3: Poultry Processing 80

3.7.1 Market Segments and Case Study Overview 80

3.7.2 Transport of Animals 82

3.7.3 Stunning and Bleeding 83

3.7.4 Scalding and Eviscerating 84

3.7.5 Chilling 84

3.7.6 Rough Cutting 85

3.7.7 Fine Cutting 86

3.7.8 Packaging 86

3.7.9 Storage and Delivery 87

3.8 Conclusion 87

4 The Fresh Food Industry’s Profile Regarding APS Systems 89

4.1 Methodological Remarks 89

4.2 General Requirements 90

4.3 Requirements for Strategic Network Design 93

4.4 Requirements for Demand Planning 95

4.5 Requirements for Supply Network Planning 100

4.6 Requirements for Purchasing & Materials Requirements Planning 101

4.7 Requirements for Production Planning and Production Scheduling 103

4.8 Requirements for Distribution Planning 109

4.9 Requirements for Transport Planning 111

4.10 Requirements for Demand Fulfilment and Available-to-Promise 114

4.11 Conclusion 116

5 Shelf Life in Fresh Food Industries 117

5.1 Shelf Life of Food Products 117

5.1.1 Definition and Limiting Factors 117

5.1.2 Determination of Shelf Life 119

5.1.3 Technological Shelf Life Extensions 120

5.2 Shelf Life Characteristics of Case Study Products 121

5.2.1 Case Study 1: Shelf Life of Yogurt 121

5.2.2 Case Study 2: Shelf Life of Sausages 122

5.2.3 Case Study 3: Shelf Life of Fresh Poultry 123

5.3 Shelf Life in Fresh Food Supply Chain Management 125

5.3.1 Literature Review 125

5.3.2 Role of Shelf Life in Fresh Food Supply Chains 127

5.4 Conclusion 128

6 Shelf Life Integration in APS-Systems 131

6.1 Introduction 131

6.2 SAP APO 131

6.2.1 System Overview 131

6.2.2 Shelf Life Integration 134

6.3 PeopleSoft EnterpriseOne 137

6.3.1 System Overview 137

6.3.2 Shelf Life Integration 139

6.4 CSB-System 140

6.4.1 System Overview 140

6.4.2 Shelf Life Integration 143

6.5 Summary and Conclusion 143

7 Shelf Life Integration in Yogurt Production 147

7.1 Problem Demarcation and Modeling Approach 147

7.2 Model Formulations 152

7.2.1 Model 1: Model with Day Bounds 152

7.2.2 Model 2: Model with Set-up Conservation 159

7.2.3 Model 3: Position Based Model 163

7.3 Computational Results 171

7.3.1 Simultaneous Optimization of All Lines 171

7.3.2 Line Decomposition Approach 173

7.3.3 Model Combination and “Pick-the-Best” Approach 174

7.4 Conclusion 177

8 Shelf Life Integration in Sausage Production 179

8.1 Problem Demarcation and Modeling Approach 179

8.2 Model Formulation 183

8.3 Computational Results 191

8.4 Conclusion 195

9 Shelf Life Integration in Poultry Processing 197

9.1 Problem Demarcation and Modeling Approach 197

9.2 Model Formulation 200

9.3 Computational Results 206

9.4 Conclusion 209

10 Conclusions and Recommendations 211

10.1 Summary of Results 211

10.2 Discussion 213

10.3 Recommendations for Further Research 215

References 217

3PL Third Party Logistics Provider

CPFR Collaborative Planning, Forecasting and Replenishment

CTP Capable-to-Promise

EDIFACT Electronic Data Interchange for Administration, Commerce and

Transport

HACCP Hazard Analysis Critical Control Point

ISO International Organization for Standardization

TP/VS Transport Planning / Vehicle Scheduling

1.1 Introduction to the Field of Research

With an approximate turnover of € 100 bn., the food processing industry is one of the major sectors of the German economy Ca 50% of this number is generated by fresh food industries such as the meat, dairy, fish, fruit, vegetables, or bakery in-dustry (Lebensmittel Zeitung 2001) Due to factors such as high variability of raw materials, intermediate and final products, fluctuating prices, or variable process-ing times and yields, production planning in fresh food industries is generally a challenging task

In this environment, Advanced Planning and Scheduling (APS) systems can constitute significant means of support for the planner Driven by developments in Supply Chain Management (SCM) and Information Technology (IT), APS sys-tems are a shift of paradigm in production planning since they address material re-strictions and capacity constraints simultaneously and not successively as Enter-prise Resource Planning (ERP) systems implemented today by most companies Hence, APS systems help to avoid high amounts of work-in-progress, to increase service levels, and to shorten planning times Moreover, APS systems allow opti-mizing the entire supply network by integrating several production sites, distribu-tion centers, suppliers and customers into one planning model However, imple-mentation numbers of APS systems in fresh food industries remain rather low, because many important requirements of these industries are not yet sufficiently covered

One of the most distinctive factors to consider in fresh food production ning is the limited shelf life of the products Shelf life restrictions directly influ-ence scrap rates, out-of-stock rates in the retail outlets and inventory levels Fur-thermore, consumers tend to buy the product that has the longest possible shelf life Being able to offer a longer shelf life than their competitors constitutes a piv-otal competitive advantage for food producers Hence, the provision of shelf life functions is crucial for APS systems in order to succeed in the fresh food industry Yet, only a few authors have considered the integration of shelf life into produc-tion planning (see Chapter 5.3.1)

be used as a guideline for other industries In literature, fresh food industries have not been subject to intense research regarding APS systems Most contributions dealing with APS systems are concerned with the automotive or the semiconduc-tor industry when looking at discrete parts manufacturing (see for example Schmelmer and Seiling 2002; Schneeweiss and Wetterauer 2002; Zeier 2002d) or with the chemical industry when looking at process industries (see for example Hurtmanns and Packowski 1999; Franke 2002; Kallrath 2002; Mekschrat 2002; Richter and Stockrahm 2002) Some research is also related to the food industry in general (e.g Wagner and Meyr 2002), however no author looks specifically at the requirements of fresh food industries

Pro-products are often neglected Furthermore, product freshness is only considered as

a constraint and is not part of the optimization The models developed for the three case study industries address these issues

1.3 Dissertation Outline

According to the two research questions, the dissertation is divided into two tions The first section (Chapters 1 to 4, see Fig 1) aims at answering the first re-search question and concludes with a comprehensive list of requirements The second section (Chapters 5 to 9) covers the integration of shelf life into production planning

sec-After having introduced the research subject, the dissertation starts with an overview of the current status of APS systems (Chapter 2) The most important functions of each of the software modules are described, and the level of support for the planner is evaluated The assessment relies on a literature review of APS systems and of production planning and scheduling, as well as on descriptions of selected APS systems This analysis provides an understanding of what these APS systems can offer

Fig 1.1 Structure of dissertation

&RQFOXVLRQVDQG5HFRPPHQGDWLRQV

6KHOI/LIH,QWHJUDWLRQLQ$36

6KHOI/LIH,QWHJUDWLRQLQ$366\VWHPV 6\VWHPV 6RIWZDUH(YDOXDWLRQ

Then the characteristics of fresh food industries are examined (Chapter 3) An overview is given on major segments, competition, developments and trends Lastly, the characteristics of Fresh Food Supply Chains (FFSC) and fresh food production systems are emphasized The results of this analysis provide the foun-dation for the subsequent development of the requirements of fresh food industries regarding each of the modules of an APS system (Chapter 4) The second section

of the thesis begins with an overview of the shelf life characteristics of perishable food products (Chapter 5) The reasons and influencing factors for shelf life limi-tations are examined and options to extend shelf life are evaluated Furthermore, the consideration of shelf life in the Operations Research (OR) literature is ana-lyzed Thereafter, the impact of shelf life restrictions on production planning is examined qualitatively Following this theoretical foundation of shelf life and its implications on production planning, several APS systems are evaluated with re-spect to how they cover shelf life (Chapter 6) For the analysis, three important players in the German SCM software market have been chosen (SAP, PeopleSoft, and CSB) Each software package is assessed based on the scope of shelf life inte-gration and its capabilities to generate plans that optimize shelf life output Based

on the deficits of current APS systems, new models are developed and resolved The models consider shelf life planning problems in specific fresh food industries (yogurt production, sausage production and poultry processing, Chapters 7 to 9) Special attention is paid to short term planning problems with a planning horizon

of one week In Chapter 10, the major findings are summarized and tions for further research are provided

devel-2.1 Evolutionary Path of APS Systems

2.1.1 MRP I and MRP II

The production planning and scheduling processes that have been implemented by most companies over the last 20 years rely on the Material Requirement Planning (MRP I) and Manufacturing Resources Planning (MRP II) logic (Davies et al 2002) Fig 2.1 provides an overview on the emergence of the different applica-tions and the related system architectures over time

Fig 2.1 Market penetration of planning systems (based on von Steinaecker and Kühner

2001)

The MRP I concept was developed and refined by J Orlicky at IBM and the consultant O Wight in the 1960s and 1970s (Walle 1999) It is a mathematical modeling tool to assist order planners in determining the needs of dependent com-ponents, such as raw materials, parts and sub-assemblies in a manufacturing or warehousing environment

MRP I is founded on the principle of successive planning and includes four main steps (see for example Tempelmeier 1999b; Walle 1999; Günther and Tem-pelmeier 2000; Steven and Krüger 2002):

)RFXVRIDSSOLFDWLRQ

x Step 1: The entire MRP I process is driven by the end-item schedule of

the Master Production Schedule (MPS), which indicates which end items have to be completed within a certain time period (Marbacher 2001), as well as confirmed customer orders The primary demand of each end-item is determined by additionally considering stock balances of end-items

x Step 2: Based on a product structure or Bill of Materials (BOM), the

pri-mary demand is translated into gross requirements of the related nents by “exploding” the end items through the BOM The BOM con-tains the complete product description, including the materials, parts, and components as well as the sequence in which the product is created (Chase et al 1998) Then the system makes a projection on the stock bal-ances by taking into account the previous stock balance and planned or scheduled receipts and calculates the net requirements for any given part (“netting”) Thereafter, simple lot sizing algorithms are applied (for ex-amples see Buffa and Sarin 1987) The results of this calculation are planned orders including a rough indication of timing

compo-x Step 3: Each activity required to produce a part is scheduled to determine

the capacity utilization of all necessary resources If a resource is utilized

by over 100% of its capacity, the planner tries either to manually shift non-critical orders or to schedule overtime hours

x Step 4: Finally, the planned orders are released to the production

depart-ment and assigned to specific resources (e.g machines or manpower) For each resource, the sequence of the orders is determined, e.g based on priority rules

MRP I is regarded as the underlying philosophy for all following production planning and scheduling concepts Its simple way of data calculation was particu-larly suited for the low performing information systems of the 1970s By applying MRP I, many companies realized significant benefits, especially in the field of in-ventory reduction in multi-echelon production environments However, the hierar-chical planning approach often led to infeasible plans, as MRP I assumes infinite capacity (Voß and Woodruff 2000; N.N 2002a) Feedback loops between input and output do not exist, and the orders for critical parts are often inflated to avoid stock-outs (Thaler 2001) Finally, as an isolated unit, MRP I only applies to a small part of the business function (Walle 1999)

The MRP II concept aims at eliminating the shortcomings of MRP I by grating additional planning modules It generally includes MRP I as one compo-nent Therefore, it did not fundamentally change but refine the planning logic (Bartsch and Bickenbach 2002) Fig 2.2 gives an example of the modules and the planning logic, which are usually incorporated in an MRP II system (Grünauer 2001)

inte-Fig 2.2 MRP II planning concept (based on Grünauer 2001)

MRP II improved upon MRP I in three distinct ways: First, the forecast of the demand of end-items is now embedded in the general business planning of the company Secondly, the introduction of feedback loops prevents infeasible plans from being generated by considering capacity constraints Finally, due to the in-creased computational capacities, more actual information concerning production planning can be managed which results in improved decision making (Kuhn and Hellingrath 2002)

The implementation of MRP II was pushed by a performance improvement of the underlying IT-hardware, leading to an integration of formerly separated infor-mation systems into modular Enterprise Resource Planning systems with one common database (Steven and Krüger 2002) The MRP logic is embedded in all major ERP systems, yet ERP systems go beyond the MRP II logic to manage a company’s entire business and overcome functional boundaries within a company The software is generally compiled in series of modules, each one covering par-ticular functional elements of the company such as sales, accounting, human re-sources, manufacturing, logistics and many others (Walle 1999) On a worldwide scale, the primary ERP vendors are Baan, Oracle, PeopleSoft, SAP and J D Ed-wards (O’Leary 2000) In Germany, SAP is by far the market leader, holding 58,3% of all ERP-related license and maintenance fees in 2000 (Kaftan and Kaf-tan 2002) Today, ERP systems constitute the basic architecture (“backbone”) of all business applications including APS systems The focus of ERP systems is to support cross-departmental and cross-functional transactions However, real plan-

%XVLQHVV)RUHFDVWLQJ 3URGXFWLRQ 6DOHV3ODQQLQJ

3URGXFWLRQ 3ODQQLQJ

2UJDQL]DWLRQDO2EMHFWLYHV

)LQDQFLDO 3ODQQLQJ

5HVRXUFH5HTXLUHPHQWV 3ODQQLQJ

5RXJK&XW&DSDFLW\ 3ODQQLQJ

&DSDFLW\5HTXLUHPHQWV 3ODQQLQJ

3XUFKDVH3ODQQLQJ DQG&RQWURO

2UJDQL]DWLRQDO2EMHFWLYHV

)LQDQFLDO 3ODQQLQJ

5HVRXUFH5HTXLUHPHQWV 3ODQQLQJ

5RXJK&XW&DSDFLW\ 3ODQQLQJ

&DSDFLW\5HTXLUHPHQWV 3ODQQLQJ

3XUFKDVH3ODQQLQJ DQG&RQWURO

ning support is only provided for isolated activities such as algorithms for lot ing (Steven and Krüger 2002)

siz-2.1.2 Assessment of the MRP Planning Concepts

The MRP concepts and their implementations in ERP systems have been subject

to intense research throughout the 1980s and 1990s Although many authors have criticized these concepts, the introduction of MRP I and II has led to significant benefits compared to past concepts Schonberger and Knod (1994) provide a se-lection of advantages that have been realized:

completions has a major effect on customer satisfaction

x Cut inventories typically in the range of 20% to 35% The cut in

inven-tory has been realized at the same time as the on-time completions creased

and 25% to 40% in assembly as well as overtime cuts between 50% and 90% This is mainly due to less time being required to halt one job and set-up a shortage-list job

x Increased productivity of support staff As the time for expediting is

duced and planning procedures are partly computerized, less support sources are necessary

re-placed by the MRP II package (e.g purchasing)

However, despite the progress related to the MRP concepts and its widespread distribution (particularly in larger manufacturing companies), many researchers as well as practitioners report numerous weaknesses of MRP (e.g Schonberger and Knod 1994; Meyr 1999; Günther and Tempelmeier 2000; Li et al 2000; Zijm 2000; N.N 2001; Tempelmeier 2001; Bartsch and Bickenbach 2002; Knolmayer

out-of-date and the scheduled quantities do not correspond to the sales orders

being booked

x The scheduling decisions rely on lead-times that have been specified in advance There is no link to the current situation on the shop floor For security reasons and due to the generally high risk-aversion of planners,

these lead-times are regularly greatly overestimated leading to

unneces-sary intermediate stocks Günther and Tempelmeier (2000) estimate that often the waiting times of an order caused by the organization are as high

as 85% of the total lead-time

x The size of production orders is determined without considering

inter-dependencies

x Furthermore, MRP overemphasizes the demand explosion while nearly

neglecting the lot sizing part.

x The focus of the systems is site-centric Other plants or even other ners in the Supply Chain (SC) such as customers, suppliers or shippers cannot be integrated in the planning cycle Consequently, MRP estab-

part-lishes a local optimum without optimizing the entire chain.

x The amount of time required to establish a plan or to re-plan is tial (the “MRP-run” is usually executed overnight or over the weekend)

substan-For this reason, plans are literally “stiff” as it is not easily possible to

change them

Although the benefits of the systems had been significant in the 1970s and 1980s, many companies were no longer satisfied with the planning results at the beginning of the 1990s The deficits were immanent to the systems and mainly due to the successiveness of the planning process Therefore, simply employing modern information technology could not eliminate them (Günter and Tempel-meier (2000) To further improve the performance of production planning, the planning philosophy had to change

2.1.3 Emergence of APS Systems

The development of the APS systems was motivated by the stated drawbacks of the MRP I and II planning logic Other drivers were the growing integration of business processes beyond site and corporate boundaries (Bartsch and Bickenbach 2002), improved optimization algorithms, and the significant performance in-crease and innovations in hardware technology in the 1990s (Kodweiss and Nadj-mabadi 2001) For example, an analysis performed by Bixby (2002) revealed that the solving power required to solve production planning problems has increased

by six orders of magnitude since 1987, which is related to an increase by three ders of magnitude of both the algorithmic and the machine speed

or-Advanced Planning and Scheduling systems aim in particular at supporting decision-making in SCM Some authors use the abbreviation “APS” for “Ad-vanced Planning Systems”; however, in this research APS refers to “Advanced Planning and Scheduling” These systems are not a substitute for an ERP system,

but can be regarded as a layer on top of an ERP system in order to support the planner in making decisions at all levels (see Fig 2.3; Davies et al 2002) The transactions are executed by means of Supply Chain Execution (SCE) systems (e.g order, inventory, transportation or warehouse management system) The APS system has access to the data of the ERP and SCE systems at any time; it can ma-nipulate the data and write the results of the calculation back into the ERP and SCE systems (Corsten and Gabriel 2002) However, the boundaries between APS, ERP and SCE are fluent Regarding specific APS systems it becomes difficult to distinguish between APS, ERP and SCE functions

Fig 2.3 Relation between SCE, ERP and APS systems (based on Knolmayer 2001b)

Most authors use the terms “APS system” and “Supply Chain Planning (SCP) system” interchangeably and define an SCM system as the sum of an APS (or SCP) and an SCE system Others discuss SCM systems and APS systems as equivalents, or regard APS systems as a subset of SCP systems (Knolmayer 2001a) Although different software providers have launched APS systems inde-pendently at different points in time, they all have a number of common basic characteristics (see for example Ferrar 2000; Grünauer 2001; N.N 2001; Hieber 2002; Knolmayer et al 2002; Meyr et al 2002b; Werner 2002a):

x All APS systems are decision support tools and not transaction systems

They prepare plans and provide the possibility to run what-if analyses of multiple production scenarios, but they do not provide facilities for insti-gating or recording material issues or movements

x APS systems can simultaneously compute plans and schedules for ple variables and constraints (e.g materials, resources, demands, etc.),

multi-by permitting them to generate plans that are optimized for multiple and user-defined criteria (cost, time etc.)

heuristics are integrated into the software packages to solve these ticated planning problems Powerful standard optimization software is

sophis-$366\VWHP

(536\VWHP

6&(6\VWHP

embedded in the APS systems so that even very complex problems can

be resolved

x Due to dramatic progress in hardware efficiency, APS systems provide a

very high processing speed as they use a dedicated server and in-memory

processing They do not rely on a database to store and locate the data used for the calculations, which avoids repetitive read and write transac-tions to and from the database

APS systems constitute a significant progress compared to ERP systems; many

of the disadvantages of ERP systems have been leveled off (see Table 2.1; based

on Ferrar 2000; Benninger and Grandjot 2001; Grünauer 2001; von Steinaecker and Kühner 2001; Davies et al 2002; Knolmayer et al 2002; Kuhn and Hellin-grath 2002; Steven and Krüger 2002) The primary differentiating factor between APS and ERP systems is the shift in the planning philosophy that had been un-changed since the 1960s (Tempelmeier 1999a) Constraints and bottlenecks, which have previously been neglected, are now taken into account

Table 2.1 ERP versus APS systems

Planning

philosophy

- Planning without considering the limited availability of key re- sources required for executing the plans

- Goal: First-cut requirements timate, feasible plans

es Push

- Sequential and top-down

- Planning provides feasible and reasonable plans based on the lim- ited availability of key resources

- Goal: Optimal plans

- Pull

- Integrated and simultaneous

Industry scope Primarily discrete manufacturing All industries including process

in-dustries Major business

areas supported

Transaction: Financials,

Controlling, Manufacturing, HR

Planning: Demand, Manufacturing,

Logistics, Supply Chain

Simulation

capabilities

Low High Ability to optimize

cost, price, profit

Manufacturing

lead-times

Fixed Flexible Incremental

calculations

The objective of production planning shifted from generating feasible plans to plans that are subject to company-specific optimization criteria Therefore, all planning parameters of a specific planning problem have to be considered simul-taneously Ideally, production lead-times that are fixed in the MRP logic can be reduced to the extent that an order-based pull production can be implemented In contrast to the MRP logic APS systems are also suited for process industries (Fleischmann 1998)

Due to complex processing conditions, the efficient reorganization of the ply chain is much more difficult in these industries than in discrete parts manufac-turing (Günther and van Beek 2003) Finally, another important difference is the decision support function, which is linked to the capability to quickly create new plans While it took a considerable amount of time – sometimes even a runtime of

sup-24 hours (Fritsche 1999) - to establish a plan with an ERP solution, APS systems deliver the results much faster due to the memory-resistant data storage This is especially true for smaller changes of parameters, since in this case ERP systems had to recalculate the entire plan

2.2 Structure of APS Systems

2.2.1 Overview

Today, APS systems cover most aspects of supply chain planning: from ment to sales and from strategic to operational decisions Different modules of APS systems support different tasks in the planning process (see Fig 2.4)

procure-Fig 2.4 Software modules of APS systems (based on Meyr et al 2002b)

'LVWULEXWLRQ

3ODQQLQJ

7UDQVSRUW3ODQQLQJ

'HPDQG3ODQQLQJ

$YDLODEOH

$YDLODEOHWRWRWR3URPLVH

'LVWULEXWLRQ

3ODQQLQJ

7UDQVSRUW3ODQQLQJ

'HPDQG3ODQQLQJ

$YDLODEOH

$YDLODEOHWRWRWR3URPLVH

The representation is based on the so-called Supply Chain Planning Matrix though the names of the modules differ depending on the APS provider, the sup-ported planning activities are generally the same (Meyr et al 2002b) However, most of the implementations do not cover all of the modules Usually only those desired by the customer are activated and installed Although not all APS provid-ers currently offer all modules, the clear trend is to supply an entire package cov-ering all planning tasks The following modules can be distinguished (at this place, only a short overview on the different modules is provided; for a detailed analysis

Al-of each module, it is referred to Chapter 2.2.2 to Chapter 2.2.9:

a long-term planning horizon up to ten years (Neumann et al 2002)

x Within a tactical or mid-term planning horizon, the Supply Network Planning (SNP) module aims at efficiently utilizing the company’s ca-

pacities Therefore, the purchasing, production and distribution functions are planned simultaneously The MPS is one important result of this mo-dule

x The Demand Planning (DP) module incorporates both strategic

long-term demand estimation and mid-long-term sales planning (Meyr et al 2002b)

x Most of the functions of ERP systems concerning production planning

and purchasing are now incorporated in the Purchasing & Materials quirements Planning (P&MRP) module However, as many companies

Re-have these functions already available in their legacy ERP system, this module is only seldom provided in APS packages (Meyr et al 2002b) and therefore not presented in detail in the following paragraphs

deter-mining lot sizes and detailed production schedules Depending on the production type, PP and PS can be performed on one or two planning levels (Stadler 2002)

build the most efficient transportation method (Davies et al 2002) While the DisP module is a more detailed representation of the Master Planning for the distribution part (Meyr et al 2002b), TP considers short-term fac-tors such as routing or vehicle availability

x Available-to-Promise (ATP) is meant to generate quick and reliable order

promises (Kilger and Schneeweiss 2002a)

While a representation of the APS modules along the dimensions business process and planning horizon as in Fig 2.4 is favored by many authors (see for example Corsten and Gössinger 2001b; Kilger and Müller 2002; Rohde 2002; Neumann et al 2002), some elements are missing or misleading As planning ac-tivities for different industries can vary noticeably, industry-specific planning so-lutions should be added, in particular at the mid- and short-term planning levels Moreover, the aspect of collaboration with suppliers and customers should also be integrated due to its importance in implementing efficient and effective SCM (Meyr et al 2002b) Finally, Tempelmeier (2001) criticizes that the multi-location

structure of a network is not visible and that DP and ATP have nothing to do with planning, but instead with the generation of input data Likewise, he emphasizes that the MRP calculations should be shifted from purchasing to production since APS systems do not replace the MRP calculations, but rely on their results However, despite the deficits stated above, the representation depicted in Fig 2.4 is still suited to describe the current status of APS systems, because the struc-ture of the APS systems has developed according to the two dimensions of the SC planning matrix (Tempelmeier 2001) As shown later, the weaknesses of the rep-resentation coincide with the actual weaknesses of APS systems in general In the following chapters, the planning objectives of each of the modules are analyzed in detail Major input and output factors are described Special attention is paid to the solution algorithms that are incorporated into the modules

2.2.2 Strategic Network Design

The SND module provides support for key strategic decisions concerning the figuration of the entire SC (Hartweg and Bruckner 2001) The module is essential for industries with frequently changing SCs and material flows (Hauptmann and Zeier 2001) The planning horizon typically ranges from three to ten years (Goet-schalckx 2002) Primary users of the module are business development depart-ments or consultancies (Fraunhofer Gesellschaft 1999) The issues addressed can

con-be structured around four major pillars (Davies et al 2002):

x Product Strategy: number and main characteristics of products as well as

markets to be served;

strat-egy, investment decisions, supplier selection;

x Logistics Strategy: number, locations and echelons of distribution

cen-ters, sourcing strategy, investment decisions; and

new technology introduction

SND models must tie together all relevant decision variables and constraints lated to countries, periods, products, facilities, transportation channels, product flows and inventories just to mention a few examples Therefore, they are very large (Geoffrion and Powers 1995; Goetschalckx 2002) To limit the model size, products, suppliers and customers are usually aggregated to zones (Vidal and Goetschalckx 1997) Some APS providers also incorporate model size limits to keep the models solvable (Steven and Krüger 2002) The modeling support pro-vided by the SND module generally comprises numerous functions Typical fea-tures are multi-echelon and multi-period modeling, finite capacities for sourcing, production, distribution and transportation facilities or piecewise cost functions representing economies of scale (Goetschalckx 2002) In addition, the model should also account for governmental issues such as tariffs, duties and transfer prices (Cohen and Lee 1989) Stochastic features have not yet been integrated into most systems, although future estimates incorporate a high degree of uncertainty

re-To cope with this, scenarios can be developed that describe the best, the worst and the most realistic case Either cost minimization or profit maximization can be chosen as objective function The models are defined as LP or MILP models and can hence be solved with standard solvers Nonetheless, in order to achieve rea-sonable solution times, a significant level of technical expertise is required to limit the model size (Goetschalckx et al 2002) For an example from industry, it is re-ferred to Cohen and Lee (1989)

Due to the necessity to reduce complexity and in order to reach a high level of abstraction, the models can only constitute a decision support tool for the SC de-sign team The results should be evaluated with a “healthy skepticism” (Goet-schalckx 2002), as the planning results have a high influence on all following planning steps Although the results of the SND module have the highest impact

on the SC, it is interesting to notice that APS providers generate only marginal cense fees with this application Many companies hesitate to implement the mod-ule for two reasons: First, the strategic decisions for which the module is used are company specific to a high degree, so that an individual support cannot be pro-vided by standard applications Secondly, these decisions are not made very often, and therefore most companies prefer to use simple Excel-based solutions

li-2.2.3 Demand Planning

The objectives of the DP module are to forecast and plan future demand (Davies et

al 2002) DP is relatively easy to install thanks to the limited interactions with other modules In addition, the results of the DP module are required as input fig-ures for the other modules Therefore it seems reasonable to start an APS imple-mentation with this module Two levels can be distinguished within DP: The long-term demand is generally forecasted for several years on a product group or prod-uct family level It serves to support the SNP module The mid-term demand is elaborated on a Stock Keeping Unit (SKU) level with a planning horizon of months or weeks and can generally be partitioned by customers, regions, seg-ments, or distribution channels (Kuhn and Hellingrath 2002) The mid-term de-mand figures constitute the input data for several other modules such as SNP, DisP or ATP Finally, the short-term demand is derived from the orders in the ERP system

Forecasting in APS systems relies on three components (Wagner 2002) First, statistical forecasting methods assist the planner in making estimations derived from historical data Within statistical forecasting, time series methods assuming demand to follow a certain pattern can be distinguished from causal methods, which focus on the relationships between two series (dependent and independent variable) Examples for time series methods are (Davies et al 2002; Meyr 2002):

x Moving Average: smoothing time series to reduce period-to-period

varia-tion;

seasonal and error components;

x Exponential Smoothing / Holt-Winters: smoothing time series by

assign-ing greater weights to most recent observations and includassign-ing trend and seasonality through decomposition;

mod-eling a series using trend, seasonal and smoothing coefficients that are based on moving average, auto regression and differential equations;

x Croston: forecasts the length of periods and the size of demand for

spo-radic demand

Examples for causal methods are (Davies et al 2002; Meyr 2002):

and many independent variables using the least squares method;

endogenous and exogenous variables using the least squares method to model mutual causality;

weighted connections

A major shortcoming of both types of methods is that only inventory exit umes are registered in ERP systems Unfulfilled demand results in an inventory exit of zero, although there is a demand for the specific product (Tempelmeier 1999a)

vol-Secondly, judgment factors are incorporated to correct and improve the cal forecast Data on promotions and marketing campaigns (own and competitors), customer feedback on special products, or cannibalization with regard to product launches can be integrated into the forecast by judgment factors (Seidl 2000) Changes in pricing policies are another primary driver for demand volatility that can only be captured by judgment factors (Bolton 1998) Davies et al (2002) name several methods for integrating qualitative factors:

statisti-x Panel consensus: consensus of experts to yield a better forecast than a

single expert’s opinion;

x Sales force composite: average forecast from independent inputs of

sev-eral salespeople;

that are tabulated and modified in reaching conclusions

Thirdly, the collaboration component assures that input for the demand ning process can be collected from all involved departments such as marketing, sales, procurement, or logistics (Rojek 2000) Collaboration is not only necessary

plan-to gather all relevant information, but also plan-to get an organizational agreement for the planning results (Smith et al 1998) In case of an external collaboration, de-mand forecasts from customers are integrated as well

In addition to the statistical forecasting functions, features that support the ity control of the forecast, the selection of methods and parameters, and forecast-ing based on product life cycles are integrated Most common measures to control forecast accuracy are the mean squared error, the mean absolute deviation, the

qual-weighted mean absolute percent error, and the mean absolute percentage error (Wagner 2002; Smith et al 1998) After having calculated the accuracy of a method, most APS systems can automatically generate a proposition as to which method or which combination of methods with which parameter adjustment achieves the highest forecast accuracy; frequent forecasting methods changes, however, lead to plan nervousness (Corsten and Gössinger 2001b) Forecasts based on life cycles should only be generated if the demand curve of the product is relatively similar to the compared product, which is not often the case in real life (Corsten and Gössinger 2001a) Lastly, most DP modules can support the calcula-tion of single-stage safety stocks

With respect to the DP module, Cap Gemini Ernst & Young (2002a) expect creased capabilities in external collaboration to be the most important develop-ments for the forthcoming years This trend can be observed for other modules as well but it is particularly important for DP The collaboration function also allows the customers to enter demand data directly into the system In addition, the APS providers aim to better integrate the DP module with other APS modules In many cases, APS systems consist of modules that have been acquired from different APS providers, and are therefore not yet as integrated as they should be

in-2.2.4 Supply Network Planning

The SNP module aims at synchronizing the flow of materials along the SC It ports mid-term decisions concerning efficient utilization of production, transport and supply capacities, seasonal stocks as well as the balancing of supply and de-mand (Rohde and Wagner 2002) SNP can be carried out on an intra- or inter-company level In the latter case, the “strongest” partner in the SC is in charge of the planning as he has the highest degree of added value, and is often within clos-est vicinity to the final customer (Kuhn and Hellingrath 2002) To integrate all demand peaks, the planning horizon has to cover at least one seasonal cycle (Neu-mann et al 2002) Frequently, the planning horizon covers 12 months and is di-vided into periods of a week or a month (“time buckets”) Inputs to the SNP mod-ule are the determinations of the SND and DP module and data on capacity, costs and stock levels per plant or Distribution Center (DC) Due to the complexity of the problem, only bottleneck resources can be modeled in detail Moreover, the BOMs of all products or product groups are required to gain information on input-output coefficients (Rohde and Wagner 2002)

sup-In order to solve the optimization problem, LP or MILP techniques are applied; most vendors use a mix of internally developed and third party solvers like ILOG/Ceplex or DASH/Express (Shepherd and Lapide 1999) The most important decision variables include sourcing, production and transportation quantities as well as inventory levels for every product, period and plant To increase the solv-ability of the model, most vendors distinguish between hard and soft constraints While hard constraints have to be fulfilled, the violation of soft constraints is only penalized by the model (Davies et al 2002) For instance, if the demand cannot be fulfilled completely, penalties are imposed for the lacking volume Another tech-

nique to reduce model complexity is aggregation Rohde and Wagner (2002) tinguish the aggregation of time, decision variables and data When aggregating the time, several smaller periods are consolidated into a larger period The aggre-gation of decision variables generally refers to the consolidation of production and transportation quantities by aggregating for example products to product families This is generally more difficult for the supply side than for the delivery to the final customers due to the influence of differential supply costs among suppliers and their greater uniqueness in terms of which supplier provides what (Geoffrion 1977) Finally, aggregating data comprises the grouping of, for instance, produc-tion, transport or inventory capacities, or the grouping of product demand data into product family demand (Rohde and Wagner 2002)

dis-The outputs of the model are aggregated production and distribution plans that are transmitted as requirements for the detailed plans to the PP, PS, DisP, TP, and

to the P&MRP module The main purpose of the SNP module is thus to coordinate the detailed plans, yet the achieved planning results cannot be used without further processing (Steven and Krüger 2002) Therefore, for Rohde and Wagner (2002) the most important planning results are the planned capacity usage and the amount

of seasonal stock at the end of each time bucket, both of which cannot be mined in the short-term planning modules as they need a full planning cycle Only the planning results of the first time bucket are obligatory for the short-term plan-ning modules (frozen horizon), while the requirements for the following periods are determined in later planning runs with actualized data (Corsten and Gössinger 2001a; Thonemann et al 2003)

deter-2.2.5 Production Planning

Among all modules, the importance of an IT support is considered highest for PP and PS (Nienhaus et al 2003) In the PP process, the aggregated production plan resulting from the SNP module is subsequently disaggregated to provide an opti-mized production plan for each site of the SC Hence, the responsibility is gener-ally located at the site level The planning horizon differs between weeks and months with time buckets of days or weeks The typical granularity is on machine group level; further detailing is then performed within the PS module (Kortmann and Lessing 2000) The production and distribution quantities of the SNP module for each time period constitute the most important input for PP as it sets the frame within which the decentralized PP decisions can be performed Other directives usually include the amount of overtime to be used, the availability of upstream items in the SC or the amount of seasonal stocks to be built up (Stadler 2002) According to Kuhn and Hellingrath (2002), the determination of the production plan is done in two steps First, exploding the BOM or the recipes disaggregate the production requirements of the SNP The BOM explosion is also frequently per-formed in the P&MRP module or even in the underlying ERP system The second step comprises lot sizing and a rough scheduling of the resulting production orders for all parts (on machine group level) with regard to capacity restrictions, shift plans or alternative resources

The output of the PP module is a site level production plan including the sponding capacity and material requirements that have been leveled according to critical capacity constraints The further detailing of the plan is then conducted within the following PS module The PP (and also the PS) capabilities of APS sys-tems vary considerably in sophistication concerning type and level of constraints, strength of the solvers or exception alerting (Davies et al 2002) For instance, Tempelmeier (1999a) criticizes the lot sizing support as insufficient and Davies et

corre-al (2002) state that the support of lot sizing can differ between a simple unit lot sizing and the calculation of economic lot sizes based upon manufacturing con-straints

2.2.6 Production Scheduling

The purpose of PS is to schedule in detail the resulting production orders from the

PP module In most cases, the planning responsibility is decentralized at the duction department level to respect local particularities (Kuhn and Hellingrath 2002) The typical planning horizon comprises a couple of hours or days (Kort-mann and Lessing 2000) The PS module considers a variety of constraints or manufacturing rules such as changeover times, routing requirements, resource preferences, or demand priorities (Davies et al 2002) Stadtler (2002) emphasizes that the objectives of PS are mainly time oriented (e.g reducing makespan, the sum of or the maximum lateness, or the sum of flow or setup times) Cost oriented objectives (e.g reducing variable production costs, setup costs or penalty costs) can also be integrated To solve the planning problem, APS systems use rather simple heuristics (Tempelmeier 1999a) Three types are usually available:

optimal) solutions to combinatorial decision problems by solving straint satisfaction problems consisting of variables, domains and con-straints In contrast to LP/MILP techniques, the user can influence the search strategy (Klein 2002b)

crossover) that are recognized in the evolution of the natural world to find solutions for planning problems (Knolmayer et al 2002) It finds near-optimal solutions within a reasonable time (Klein 2002a), and has only low hardware performance requirements (Stache 1997)

se-quence In this case, time gaps are searched for that result in only minor adjustments to the old plan in order to avoid plan nervousness (Stadtler 2002)

The generation of schedules can either be performed on a two level planning hierarchy as described above using the two modules PP and PS, or in a single planning step when PP and PS are integrated The decomposition decision de-pends on the production type (Stadtler 2002):

x In process organization with many different machines of similar tions, multi-stage production processes and many lot sizes within the

func-planning interval, a decomposition of the overall decision problem into

two planning levels is required to reduce the computational burden

x However, for an automated flow line with scarce resources, sequence pendent set-up times, and a smaller number of products, a two stage planning hierarchy is not adequate because sequence dependent set-up times cannot be adequately represented by time buckets Furthermore, the lot sizing and sequencing decision cannot be separated in that case as lot sizing depends on scheduling and vice versa Therefore, with regard to

de-the smaller number of products, one single planning step is preferable

2.2.7 Distribution Planning

DisP is part of the mid-term planning with a planning horizon in the range of days

to months The objective is the planning of inventory levels of final products and

of the distribution of the final products to the customer, with the objective to timize the trade off between inventory holding cost and transport cost Several de-cisions are to be made within DisP (Fleischmann 2002):

op-x The determination of aggregate transport quantities for every transport

link in the SC is the most important activity to be performed within DisP

x The frequencies of regular transports set target values for short-term

de-cisions on shipment quantities and determine the size of the transport lot

x A framework for the selection of distribution paths with regard to limits

of order size is set in DisP (e.g direct delivery if order volume exceeds

to areas have been determined in SND SNP delivers the quantities to be shipped and variations in seasonal stocks Finally, demand forecasts and safety stocks are added from DP (Fleischmann 2002) Generally, DisP overlaps with the SNP mod-ule to a large extent and hence, can only increase the planning performance in the case of a transport network with many far-off lying nodes having an identical range of products and materials (Steven and Krüger 2002)

The results of the distribution planning process are the primary input for the short-term TP module Hence, a tight integration between both is a must How-ever, as the TP module has often been acquired by the APS providers and not de-veloped in-house, the integration is frequently incomplete Consequently, most vendors currently focus on an increased integration of DisP with TP because a consideration of the constraints of the transportation capabilities will result in a more efficient distribution plan (Cap Gemini Ernst & Young 2002a)

2.2.8 Transport Planning

Based on the distribution plan, TP as a short-term planning module seeks to build the most efficient transportation method considering constraints such as costs, routing information, availability and speed of vehicles, loading constraints and mix, and timing (Davies et al 2002) The planning horizon is the same as for PS, that is, it ranges from hours to days (Kortmann and Lessing 2000) Knolmayer et

al (2002) name three major functions of TP:

x The Load Consolidation and Vehicle Scheduling function helps to

con-solidate the load to destination locations and aims at achieving high cle fill rates

vehi-x Route Determination supports the planner to find the best route through a

network with regard to time and cost

x Carrier Selection allows one to choose from several carriers and usually

includes an Internet-based tendering process

Most APS systems apply a combination of heuristics and LP/MILP procedures

to solve the planning problem The planner can intervene by means of a user face to integrate specific load optimization strategies (Davies et al 2002) Several additional features are available within specific APS solutions Some examples in-clude tracking and tracing functions (Bergmann and Rawlings 1998; Lang 2002a)

inter-or even a cargo revenue-maximizing assistant to sell overcapacities (Davies et al 2002) Nonetheless, in spite of all the offered functions, the use of the TP module

is generally only reasonable if the company manages a significant own fleet The

TP module’s offered functions are particularly useful for logistics providers bers and Plewnia 2001; Steven and Krüger 2002) Regarding the future develop-ment, it is expected that a wide range of functional enlargements will occur in TP Examples include industry-specific solutions for LSP, fleet management functions

(Nos-or the integration with on-board computers (Cap Gemini Ernst & Young 2002a)

2.2.9 Available-to-Promise

The major objective of the ATP process is to generate fast and reliable order promises to the customer and to shield production and purchasing against infeasi-bility (Kilger and Schneeweiss 2002a) For Kuhn and Hellingrath (2002), order promising is one of the key tasks in SCM because it links planning tasks that are independent of customer orders and the planning tasks related to customer orders APS systems are particularly suited to support ATP due to their high processing speed (Werner 2002b) In the traditional approach of order promising, orders were quoted against production lead-time if there is no inventory available This has led

to infeasible quotes as supply or capacity constraints have not been considered (Kilger and Schneeweiss 2002a) According to Fischer (2001), the ATP approach

of contemporary APS systems can be structured around four activities:

First, the checking of Product Availability is the core of the ATP module The basis for the product search constitutes available inventories and the quantities

calculated in the SNP (the SNP quantities available for order promising are also called “Available-to-Promise”) The ordered volume of the product is reserved and the available quantity in the corresponding time bucket of the SNP is reduced accordingly If the ordered volume is not available, an alternative delivery date can be proposed or pre-defined rules can be applied (Schneider and Grünewald 2001) The search can be extended to alternative products (see Fig 2.5, Number 1), alternative locations (Number 2), or both (Number 3) The system can even review production commitments and re-plan the production This procedure is called Ca-pable-to-Promise (CTP, Number 4)

Secondly, the Initial Order Promising function aims at confirming the delivery date and quantity to the customer A company that is able to consistently make re-liable promises over a long period of time creates an important competitive advan-tage (Kilger and Schneeweiss 2002a)

Thirdly, ATP supports measures and decisions regarding temporary delivery inability This is especially important when only a part of all customer orders can

be satisfied with the available volume In that case, a variety of shortage allocation rules can be applied Some examples are (Fischer 2001):

x Allocation proportional to the volume of customer orders;

x Allocation proportional to the customers turnover;

x Allocation proportional to the customer’s demand forecast or

x Allocation according to predetermined priority rules

Fourthly, Due Date Control and Re-Promise are also essential for customer isfaction as re-planning of already confirmed orders can never be totally avoided due to unexpected events The objective of ATP in that case is to identify potential delivery bottlenecks as early as possible

sat-Fig 2.5 ATP rule representation (based on Knolmayer et al 2002)

&73'LPHQVLRQ

3URGXFW'LPHQVLRQ

&73'LPHQVLRQ

3URGXFW'LPHQVLRQ

The granularity of the ATP with respect to the product dimension depends on the SC decoupling point (see Table 2.2) In a Make-to-Stock (MTS) environment

as it can be found in most Consumer Packaged Goods (CPG) industries, the ATP

is generally on a finished products level due to short customer order lead-times

As a high number of configurations is possible in a Make-to-Order (MTO) ronment (e.g computer industry), the forecast and the ATP are based on product group and component level Depending on its configuration, an order can consume multiple resources Characteristics of a Configure-to-Order environment are long production lead-times and difficult forecasts with BOMs being only partially available In that case, capacities must also be considered so that ATP resembles the CTP procedure Regarding the time, the ATP module is usually represented in the same granularity as the supply given by the master plan Therefore, quoting an order means consuming quantities from a particular time bucket (Kilger and Schneeweiss 2002a)

envi-Table 2.2 ATP granularity (based on Kilger and Schneeweiss 2002a)

With regard to the future development of the ATP module, Cap Gemini Ernst

& Young (2002a) foresee a further development of so called “Advanced Order Promising” A very prominent example is “Profitable-to-Promise” in which the profitability of an order can be determined and taken into account to facilitate the fulfilment decision Another expected development is the further expansion of ATP/CTP functions in the areas of multi-echelon ATP and true CTP, meaning that

in fact all modeled capability constraints are considered Currently, this is not sible in most cases due to the high complexity of the planning problem

pos-2.3 APS Systems Market Overview

2.3.1 Available Market Studies

Although the APS software market has been subject to intense market research in the recent years, most companies in CPG industries assess the market for APS sys-tems as not very transparent (Lebensmittel Zeitung and PwC Consulting 2002) One of the first comprehensive market overviews was developed by the Fraun-hofer Gesellschaft (1999), which compares 20 APS systems based on a detailed questionnaire An overview of the most important market studies is given by Kortmann and Lessing (2000) The market studies covered include for example

Manufacturing

Environment

Order Lead-Time ATP Granularity

the Supply Chain Magic Quadrant and the SCP vendor “footprint” of the Gartner Group, the Manufacturing Systems Software Database of Cahners Business In-formation and the Manufacturing Enterprise Applications Comparison guide They analyzed the results in detail and mention these major shortcomings:

x The information for the market studies are given by the APS providers and has not been subject to intensive checks;

x Lacking comparability of the systems;

x Very technology-oriented analysis;

x No integration of reference implementations in the studies; and

x Few indications on pricing

Based on these drawbacks, the authors have also integrated an enquiry of users

in addition to the interrogation of software providers and give indications on the pricing model of the providers More recent market studies from 2002 and 2003 are more focused on specific industries or functions Trier’s (2002) comparison of

18 APS software packages stresses the importance of implementation issues sulting companies in the Netherlands have published two overviews on APS sys-tems in process industries The study prepared by Davies et al (2002) includes a detailed analysis of ten APS software packages focusing on consumer products and process industries The analysis of Cap Gemini Ernst & Young Netherlands (2002a) concentrates on the Dutch APS market and investigates specifically on semi-process industry functions Only very few market studies are not limited to the analysis of single APS systems and its capabilities, but provide data on the to-tal market size and development The most comprehensive analysis is provided AMR Research (2001 and 2003)

Con-2.3.2 Market Size and Segments

The worldwide market for SCM software has a total annual volume of about $ 5

bn This number includes APS systems as well as SCE software After a period of dramatic growth, sales reached a peak in 2001 In 2002, the SCM market has seen

a decline for the first time and shrunk by ca 6% Reasons for this development are the general economic slow down and the end of the Internet hype However, de-spite this first decrease in SCM software spending, the market is expected to fur-ther increase but at a slower growth rate than in the late 1990s (AMR Research 2003)

With regard to the types of income of APS providers, software license fees and implementation fees generate the largest part of income for the software providers Although the SCM business is basically a license business, the implementations often include process redesign and other consulting services, which result in a high portion of implementation fees However, the implementation fees are typi-cally a one-time income so that the software providers have to sell a significant number of new implementations every year This is increasingly difficult in a sluggish economy as the IT budgets for new projects are the first to be cut in case

of cost reduction programs

APS providers obtain most of their income from larger companies This is mainly due to the fact that the benefits of implementing such software are much bigger in larger companies, which have to handle a far higher complexity regard-ing sites, departments, or products These customers can already improve their op-erations significantly on an inter-enterprise basis, whereas smaller customers have

to integrate their customers and suppliers right from the beginning to fully benefit from employing the software

Finally, although the implementation of APS systems generally provides a quick Return on Investment (ROI), the investment represents a high risk for smaller companies if the implementation fails or if not all targeted benefits can be realized

2.3.3 Major Providers

Despite several takeovers in recent years, the market is still relatively fragmented The major players in the market are i2, SAP, PeopleSoft/J.D Edwards and Manu-gistics Due to the fact that most providers offer both APS and SCE functions, a clear distinction between APS and SCE providers is rather difficult Therefore, the SCM software market (which includes APS as well as SCE) is first categorized to understand the different players SCM software providers can be divided into five categories (Hellingrath et al 2002):

x Integrated SCM- and e-business suite providers: Solutions that cover

al-most every aspect of SCM (APS and SCE) Generally, these providers start with an APS system and extend the system to create an integrated

“SCM-Toolsuite” Prominent examples are i2, Manugistics, and SAP

x Specialized SCM suite providers: Providers in this category are similar to

providers in the first category, but are more concentrated on special tasks

or industries (e.g AspenTech for the process industry) Their competitive advantage often consists of sophisticated algorithms to solve specific planning problems Several providers in this category are currently enlarging their range of products and will soon be part of the first cate-gory

enlargement and complement to their ERP systems Examples are Baan, PeopleSoft, Oracle, SCT or J.D Edwards The SCM part has frequently been added through a company takeover (e.g Baan acquired Berclain, CAPS Logistics, CODA and Aurum; PeopleSoft acquired Red Pepper and J.D Edwards which bought Numetrix before) Pillep and von Wrede (1999) emphasize the tight integration of the SCM and the ERP system of the solutions in this category Therefore, these providers leverage their ERP customer base to sell the APS system

x APS niche players: Providers that develop individual solutions for special

tasks in the SC or for target groups (e.g flexis or ICON) By focusing on special activities, they can frequently offer attractive solutions with re-

gard to price and quality and are especially suited for small/mid-size terprises

systems such as Inventory Management or Warehouse Management Due

to the high variety of activities along the SC, this category is very geneous

hetero-Fig 2.6 SCP magic quadrant (based on Peterson et al 2002)

When looking specifically into the APS segment, many authors (e.g Buscher and Jelken 2000; Kortmann and Lessing 2000; Grünauer 2001) use the representa-tion of the Gartner Group (“SCP Magic Quadrant”; Peterson et al 2002) to illus-trate the market position of the APS systems (see Fig 2.6) This instrument aims

at supporting companies when selecting an APS system and integrates a variety of factors along the dimensions “Ability-to-Execute” and “Completeness of “Vi-sion” Only providers with an own multi-module package and a credible vision are accepted The “Ability-to-Execute” is a score for the sustainability of the company and the product and integrates factors such as breath and depth of the product, the technical expertise, service levels, the stability of the product and of the manage-ment, the financial performance as well as the marketing of the company Hence,

the “Ability-to-Execute” score indicates how a provider corresponds to today’s

re-quirements The “Completeness of Vision” assesses the vision of the solution with regard to cost, functions, technology, service and sustainability Another important factor for the judgment is the integration of the solution components Therefore,

the second dimension of the matrix shows how a provider will correspond to ture requirements

L

$VSHQ7HFK -'(GZDUGV

3HRSOH6RIW