The Essentials in Managing Demand

A redesign of Wavin’s demand

planning process

Master’s Thesis

MSc. Business Administration

Specialisation Organisational & Management Control

‘The Essentials in

Managing Demand’

A redesign of Wavin’s demand planning process

by Frank Voortman

University of Groningen Faculty Economics & Business

F.V.H. Voortman (1736531)

E-mail: f.v.h.voortman@student.rug.nl Address: Lossersestraat 50

Postal code: 7527 PG Enschede

Supervisor: drs. M.M. Bergervoet Co-assessor: dr. J.S. Gusc

Preface

By writing this thesis I finalise a joyful, challenging, and satisfying period as a student in Groningen; a memorable period in my life. Although my master has almost come to an end, attending the first courses of the bachelor still feels a short time ago. The specialisation Organisational & Management Control (O&MC) focused on control of companies, addressing different contexts and various perspectives. The inspiring courses with challenging themes make me look back on a successful closing of my study Business.

This thesis is written in combination with a six-month internship at Wavin’s head office in Zwolle. To my opinion, it proved to be an ultimate showcase to put theory – and particularly control theories - into practice. The subject of my research, forecasting demand, addresses a process which is really critical for many organisations. Well-designed forecasting processes provide organisations a very valuable source of management information. As such, I’m proud to present a thesis which is not solely a mandatory closure of my study, but also is a relevant case for other organisations.

The road which a student covers during a graduation internship is said to be intense and bumpy. I simple agree. The completion of this thesis was not possible without the critical - though constructive - feedback of my supervisors. I want to thank Wavin, and in particular Eelco Spaans, for the opportunity to graduate under his supervision. With his challenging mind-set, he showed me what practical thinking really is about. My university-supervisor Marcel Bergervoet provided pragmatic views on how to link the observed practices to usable theories, an activity which I – honestly – struggled with at the start of the research. During this intense period of hard work I’ve been well supported by my direct environment. Thanks go to the many colleagues at Wavin who expressed real interest in the developments of my research. Special thanks go to Johan, Ruurd, Nisrine, Serge, and Eelco - my colleagues from ‘the second floor’ - with whom I had many interesting conversations and pleasant coffee breaks. In addition, I want to thank friends and family for their continuous interest and support. As I (temporarily) moved back to my parents’ place during the research period, I want to express gratitude for taking care of me and several of the ‘distracting’ activities such as the groceries, cooking, and especially ironing…

With their care and understanding, the above mentioned people motivated me to complete this ‘masterpiece’ which I’m proud of.

Please enjoy the reading!

[Accountability and forecasting …]

Management Summary

This research is conducted in behalf of Wavin, which is faced by problems in its method to produce accurate sales forecasts. Sales forecasts are projections into the future of expected demand, given a stated set of environmental conditions. These projections are a critical information source used in decision making, and essential to effectively manage a business.

Within Wavin, the demand planning process is designed to produce these sales forecasts. Forecasting practices currently are organised at the level of Operating Companies (local per country). The method Wavin is using to forecast demand is certainly not 'best practice'. Wavin’s current performance in forecasting demand is not up to standard. Several organisational reasons can serve as an explanation; lack of ownership, scattered processes, different practices not sharing best practice, lack of in-depth knowledge on forecasting, and so on. These problems lead to the following problem statement:

Research objective

The purpose of this research is to contribute to a redesign of Wavin' demand planning process, with the aim to improve sales forecast performance.

Wavin expresses forecast performance by the measure of forecast accuracy. This measure indicates how well forecasts are able to predict future sales. Considering the size of Wavin’s business, too high forecasts result in excessive amounts of inventories (high costs), while too low forecasts result in stock-outs (decline in customer satisfaction and possibly customers themselves).

Accurate sales forecasts are required to effectively balance demand and supply. The demand planning process involves several sources of information, and participation from various functions. The general approach to produce a forecast commences with a quantitative ‘baseline forecast’, calculated by the system. After calculation, market intelligence is added; several functions (Sales/ Marketing/ Planning) have knowledge of events – e.g. unexpected orders, promotions, or machine shutdowns - that impact the initial forecast. The procedure to combine all these information sources to generate an accurate sales forecast is ineffective, leading to the following research question:

The first need is to diagnose performance of the current demand planning practices. By linking forecast performance to four maturity stages, the forecasting audit (Moon et al., 2003) provides a structured approach to asses current forecast performance. Seventeen key elements covering four forecast dimensions are diagnosed as being in phase one (immature) to phase four (best in class). Complementary, a literature research was conducted to identify criteria for effective forecasting. As two of these criteria are addressed in the forecasting audit, the emphasis was placed on remaining criteria; forecasting knowledge, clear definition of rules and procedures, clearly assigned responsibilities and accountability, and structured performance management.

Central research question

The diagnosis by these concepts identified three strategic themes to focus on in the process redesign. Limited functional integration, limited knowledge of (technical) forecasting, and limited performance management, together lead to an ineffective demand planning process.

Strategic Themes: What goes wrong?

I. Limited Functional _Integration

 Sales concentrated management approach fails assigning accountability  Consensus meetings are missing; no consensus on forecasts is reached

 Collaboration between forecasting (Sales-located) and other function is minimal  Financial forecast spread sheet in Germany is an (undesired) island of analysis

II. Limited (technical) forecasting knowledge

 Top down approach is set as an exception, rather than a standard

 Incomplete segmentations are used to determine ‘priority’ products to forecast  Impact of external factors on accuracy not considered in forecast focus

 Incomplete use of quantitative and qualitative forecasting techniques  A coherent training structure (both general & technique usage) is lacking

III. Performance Limited Management

 Multidimensional performance metrics are missing

 No structured method to provide performance feedback to forecast developers  No incentives (rewards) linked to forecast performance

Table 11: Diagnosis of key elements and forecasting benchmark dimensions Based on these strategic themes, improvements are proposed in the form of a process redesign. The logistical framework (Visser & van Goor, 2007) provides a comprehensible structure for redesign; providing recommendations for the elements of process, organisation, systems, and performance measurement. As (technical) knowledge in forecasting is one of the causes of low forecast performance, additional recommendations are provided in regard of trainings.

Based on the research performed in this study, four key recommendations are proposed to overcome the diagnosed problems and address the three strategic themes. The recommendations for Wavin – in particular the assignment of a Demand Manager – are catalysts in improving performance of the essential demand planning process.

Key recommendations

1. Assign a Demand Manager, who takes responsibility over demand planning within Wavin. 2. Make Demand Meetings cross-functional, attended by roles from different functions.

Table of Contents

Chapter 1: Company Introduction ... 6

1.1 Company History ... 6

1.2 Wavin anno 2012 ... 6

1.3 Strategy 2015 ... 6

1.4 Organisational structure ... 7

Chapter 2: Research Design ... 9

2.1 Initial motive ... 9

2.2 Problem analysis ... 10

2.3 Problem statement ... 11

2.4 Research methodology ... 12

2.5 Research outline ... 14

Chapter 3: Theoretical Framework ... 15

3.1 Demand Planning & Sales forecasting management ... 15

3.2 Dimensions of sales forecasting management ... 15

3.3 Stages of effectiveness ... 19

3.4 Design criteria in forecasting ... 19

3.5 Strategic themes ... 23

3.5 Conceptual model ... 23

3.6 Sub-questions ... 23

Chapter 4: Description current practices ... 25

4.1 Functional integration ... 25

4.2 Forecast approach ... 26

4.5 Design Criteria ... 29 4.6 Chapter summary ... 31 Chapter 5: Analysis ... 33 5.1 Forecast Dimensions ... 33 5.2 Strategic Themes ... 34 5.3 Chapter summary ... 37 Chapter 6: Redesign ... 38 6.1 Process design ... 38 6.2 Monthly procedure ... 39 6.3 Overview improvements ... 41 Chapter 7: Conclusion ... 44

7.1 Conclusions from the research ... 44

7.2 Recommendations ... 45

7.3 Conclusions about the research ... 46

7.4 Limitations & suggestions for future research ... 48

References ... 49

I. Articles ... 49

II. Books ... 51

III. Wavin documentation... 51

IV. Websites ... 51

Glossary

Terminology Description

Sales Forecast Projection into the future of expected demand, given a stated set of environmental _conditions. Sales Forecast

Management Sales forecasting management is the broader concept of using tools to manage the sales forecasting process (Mentzer, 2005) Baseline Forecast The initial (quantitative) forecast produced by forecasting software. Often incorrectly called the statistical forecast, as this forecasts also includes corrections and other

methods.

Market Intelligence Primarily external data collected and analysed by a business about both potential and _{current market in which the business is operating.} Forecast

Developers Participants in the demand planning process who provide input in the form of forecast information. KPI Key Performance Indicator, a leading indicator to manage performance. These measures _{are commonly used by management to control process at lower organisation levels.} MTO- production Make-to-order, production strategy where production is triggered once the customer _{orders. Commonly characterised by no stocks of end-products.} MTS- production Make-to-stock, production strategy where production is driven by demand forecasts. In _{this production strategy stocks of end-products are common.} SAP Business application software, named after its German developer. Widespread package of _{business software, which enables to integrate the total business in one program.}

SAP BW ‘Business Warehouse’ or BW is a packaged, comprehensive business intelligence product. It is concentrated on a data warehouse, optimised to support decision making by management.

APO DP According to SAP literature, Advanced Planner and Optimizer is designed to help a company improve production planning, pricing, scheduling, and product shipping. The APO Demand Planning module used to plan demand (techtarget.com)

SKU(L) Stock-keeping unit, a unique product or item for sale in a store or other business. _{Forecasting on SKU per location (country) is called SKUL, and is performed by Wavin.}

S&OP

Chapter 1: Company Introduction

In this chapter the introduction to Wavin is discussed. Some attention is spent addressing Wavin’s history, its views on the future, and also the situation of the company as today. Basic understanding of Wavin and its business helps to understand the problem addressed in this research.

1.1 Company History

Wavin's story begins in 1950s Zwolle, the Netherlands, where company founder Johan Keller was director of WMO, the local water utility. Supplying a largely rural district and a few urban centres, the company's water supply reached its customers through a large network of mostly iron pipes. Before long, WMO found itself battling against serious pipe corrosion and a significant loss of water. Keller decided that something needed to be done urgently. Keller went searching for alternative materials to produce pipes, and found relief in materials which protected electricity cables from water. This application used PVC to protect cables from water; why could it not protect water from soil?

Although plastic pipes were already available in the 1950s, Keller wanted them tailor-made for the supply of drinking water. Finding none on the market, he set to work in a small workshop in Zwolle, and, in 1953, succeeded in producing the first plastic pressure pipes for potable water with diameters of 100 mm and larger.

Founded in 1955

WMO's solution to its pipe corrosion problem attracted the interest of utility companies both nationally and internationally. Unable to cope with the increasing demand for its pipes, however, and aware that as a government organisation its first duty was to serve the local population, WMO created an independent company to focus solely on pipe production. In August 1955, Wavin (a contraction of the words WAter and VINyl) was founded. It had 52 employees.

Recent developments

In the years that followed, Wavin grew rapidly and concentrated on developing sound management practices and a solid internal structure. In 2006 Wavin entered the NYSE Euronext stock exchange in Amsterdam and was included in the Amsterdam Midcap index. Six years later, in 2012, Mexichem, leader in Latin America in plastic pipe systems and in the chemical and petrochemical industry, acquired Wavin, leading to a delisting.

1.2 Wavin anno 2012

Today Wavin is the leading supplier of plastic pipe systems and solutions in Europe. The company is headquartered in Zwolle (the Netherlands) and has a presence in 25 European countries. The company employs approximately 6,000 people and reported revenue of approximately EUR 1.3 billion for 2011. Outside Europe, it has a global network of agents, licensees, and distributors.

1.3 Strategy 2015

1.4 Organisational structure

Wavin has its organisation structure according a matrix design. It has specialised functions grouped at Central office, and ‘operational’ activities assigned to the geographic regions. To slightly anticipate to the scope of this thesis; this research was conducted in behalf of the SCOPEX department (supply chain), located at the Central office in Zwolle, and concerned forecasting practices in four operating companies, split over two regions.

Geographical segments

Four geographical regions are assigned (figure 1); revenue in 2011 was approximately € 1.3 Billion. Whereas revenues in the regions NWE and SWE are relatively stable and slightly declining, the regions SEU and CE have become Wavin’s growth markets.

Figure 1: Wavin’s organisational structure and relative share in total revenue Business segments

Wavin manufactures pipes and fittings for transportation of different sorts of substances (e.g. potable water, sewage water, gas). Operational activities are strategically split in two Strategic Business Units (SBU’s). In 2011, Building & Installation (i.e. above ground) was roughly responsible for 37% of the revenue generated. Civils & Infrastructure (i.e. below ground) generated 63% of the total revenue.

Figure 2: Wavin’s business segments Wavin’s supply chain

Wavin fabricates a large variety of products; from ‘ordinary’ sales of pipes & fittings to complete projects including different types of customised products. Timely production and supply of this wide assortment of products is a continuous challenge. Wavin’s supply chain (figure 3) is spread over many countries and services several types of customers. Wavin’s operations involves 40 manufacturing sites, while Sales Organisations are spread over 25 countries. To maximise efficiencies of scale & scope, an intercompany trade (ICT) structure is applied.

Chapter 2: Research Design

This chapter describes the methodology applied in this research. First the initial motive is addressed. This introduction to the problem forms the basis for the problem analysis and results in the problem statement of this research. This chapter concludes with the research method and outline.

2.1 Initial motive

Sales forecasts are essential for effective planning and decision making. Within Wavin, the demand planning process is designed to produce these sales forecasts. Forecasting practices currently are organised at the level of Operating Companies (OC’s), i.e. regionally. The method Wavin is using to forecast demand is certainly not 'best practice'. Wavin’s current performance in forecasting demand is not up to standard. Several organisational reasons could serve as an explanation; lack of ownership, scattered processes, different practices not sharing best practice, lack of in-depth knowledge on forecasting, and so on. These characteristics result in a low quality output; as forecasts are used as important input for several

planning processes, this has a high impact on the organisation. .

2.1.1 Demand & the supply chain

A sales forecast is defined as a ‘Projection into the future of expected demand, given a stated set of environmental conditions. (Mentzer, 2006), and provides information required to manage demand and supply. Management of demand and supply is important for effective operations, as holding abundant quantities of inventories is a costly practice for a production company selling 90.000+ different products in 25+ countries. The tension of managing demand

and supply can best be expressed by three central themes in Supply Chain Management; within Wavin this relation frequently was referred to as ‘the Devils triangle’.

To stick to the same terminology, the ‘devils dilemma’ is the objective to optimise the three SCM-themes (figure 4). There is a continuous need to minimise the level of inventories (too much is very costly), manage capacity (short production cycles are inefficient; long cycles are inflexible) and maximise service levels by having sufficient inventories (stock-outs are destructive). For effective supply chain management - and tackling the devils dilemma - high quality planning information is required: the sales forecast.

2.2 Problem analysis

In essence, a problem is the discrepancy between an actual and desired situation. In other words: the gap between situation ‘as is’ and ‘to be’. This research is classified as a problem solving research, which is quite typical for a Business research. It aims to ‘provide an answer to the complete knowledge needs for solving the problem’ (De Leeuw, 2001), or in simpler words: solving the organisation’s problem.

2.2.1 Forecast performance

Wavin considers forecast accuracy as the major performance indicator of the demand planning process. Forecast accuracy is a quantitative indicator influenced by many (qualitative) variables. This research focuses on those variables to identify the root cause of Wavin’s low forecast accuracy.

2.2.2 Problem owners & Impact analysis

A problem owner is a person who to some extend is worried with the status quo; often problem owner is the person expected to solve the problem (De Leeuw, 2001). The consequences of low quality sales forecast are substantial for several functional areas within Wavin. Since sales forecasts are used as input for many successive (planning) processes, the impact is specified in more detail, as displayed in table 1.

Function Inaccurate forecasts cause problems in… Breakdown of forecast error…

Supply Chain

SCOPEX Director SCO Manager Supply Chain Director NWE

…capacity planning …inventory planning …optimising asset utilisation

… service levels (on-time delivery in full)

… when over-forecast

Excessive inventory Inventory holding costs Transhipment costs Obsolescence Reduced margin

… when-under forecast

Order expediting cost Higher product cost Cost of lost sales

Lost companion product Reduced customer satisfaction Production OPEX Director OPEX Manager …production planning …distribution planning

…planning capital req. & future plant & equipment …raw material purchasing schedules

Finance

Financial Director Business Controller

…financial planning & budgeting

…projecting costs and determining profit levels

Sales

Country Directors

Sales Managers Sales Representatives

…sales planning

…determining and specifying sales targets …monitoring sales performance (year-to-date) …assuring information of sufficient stock availability

Marketing

Business Unit Directors Marketing Managers

Product Managers

…marketing plans

…determining price strategies …guidance of promotional activities

General Management &

Human Resource …business planning …manpower planning

2.2.3 Categories of problems

A preliminary intake with the initiator of this research – the SCO Manager at Central office - was held at the start of this research. Based on this conversation, elements were identified that contributed to the lack of insight and consequently, low forecast performance. Five categories of problems were diagnosed, displayed in table 2.

2.3 Problem statement

The problem statement is a sharp formulation of a targeted research product, objective and scope. It comprises a research objective, the scope of the research and the questions to be answered (De Leeuw, 2001. A special variant of the problem statement is advised when the aim of a research is a redesign. Important when a redesign is the aim of a research is three questions to be addressed. These are: what do you want to produce for whom, which conditions must be met, which limitations are set? (De Leeuw, 2001) and are addressed in the research objective and research scope.

Based on the problem analysis and impact analysis conducted in the previous paragraph, the problem statement is the following:

Problem statement

Wavin’s process to produce sales forecasts is ineffective; resulting in low sales forecast performance.

Control Problems

 Control over the demand planning process is lacking. This prevents Wavin to effective manage the

forecasting process; as a result output quality is low. Control problems are caused by four other sets of problem, as described below.

Lacking insight in Rules & Procedures

 Rules & Procedures concerning the process are unclear, lacking and/or not followed. Problem owners

indicate that next to blueprints and manuals, a clear working method is lacking.

Insufficient forecasting knowledge

 General knowledge in the concept of forecasting, and technical knowledge of forecasting methodologies is

lacking. Current working methods of demand planners vary and are unclear.

Lack of responsibilities & ownership

 Responsibilities/ accountabilities, and ownership relating to the process are unclear. Some problem owners

indicate concepts as authorisations and responsibilities being relevant, but undefined. Other terminologies mentioned are employee support, commitment, ownership, and trust in the quality of the forecast.

Communication is lacking

 Communication between, and within – different functions is unclear. Uniformity in the procedures is also

indicated as having impact on accuracy; knowledge sharing & best practices plays a role. It is unclear how performance feedback regarding demand planning takes place.

2.3.1 Research objective

The research objective is a logical product from the problem statement. A research objective states the goal a researcher wants to achieve with his study, formulated in one sentence.

Research objective

The purpose of this research is to contribute to a redesign of Wavin' demand planning process, with the aim to improve sales forecast performance.

2.3.2 Central research question

To achieve the research objective as mentioned above, the research question needs to be answered. Central research question

Which adjustments to the current demand planning process are required to increase sales forecast performance?

2.4 Research methodology

This paragraph provides an overview of the research methods applied in this research, the procedures of data collection & analysis, report delivery and research scope.

2.4.1 Data collection

The selection of a case-study was chosen as the method of research; which is defined as an intense examination of a case (or cases), with the purpose to generalise conclusions for broader contexts (De Leeuw, 2001). Several types of sources are used in this research, of which the examination of documents and reality (e.g. interviews, observation of working method) are mostly used (De Leeuw, 2001). This research commenced with the preliminary conversation to discuss the categories of problems (table 3) faced by Wavin. The next step was to collect company forecast documentation. Process documentation (e.g. improvement initiatives, templates, guidelines, reports) and general company information (e.g. organisation charts, common methods of working, training materials) were examined.

2.4.2 Data analysis & report delivery

As to objective is to propose improvements, it is a clear problem solving research. This thesis makes use of two types of research to achieve its research objective. First step in this research is to determine which factors have impact on the demand planning process. The sales forecasting audit (Moon et al., 2003) proposes a clear methodology to diagnose the current state of Wavin’s demand planning practices.

As the purpose of this report is contributing process improvements, not solely conducting an audit, emphasis is placed on how a demand planning process should look like. Based on the initial problems stated in the problem analysis (table 2), a literature study has been conducted to identify design criteria. This study compares twelve articles on forecasting and demand planning, and based on this research six design criteria are identified (Appendix A).

The audit, supplemented with criteria, provide an integral tool to diagnose and improve Wavin’s demand planning process. Two deliverables are involved in this research. The first is a presentation to the SCO department; proposing improvements for process redesign. The other deliverable is this thesis, documenting the research methodology and findings from the research.

2.4.3 Research scope

A clear scope is needed to attain focus and work effectively towards the research objective. The scope of this research is dual of nature; it relates to the research result and the process (De Leeuw, 2001).

The objective of this research implies a group solution to Wavin’s demand planning process; the redesign should be generic of nature, with minor local adjustments. To come to a representative sample, processes in four countries were examined. Certain assumptions were made at the start of this research. The client, e.g. Wavin, provided a research problem with proposed areas of research. This input formed a basis for the problem statement.

In forecasting literature, both terminologies of Sales Forecasting and Demand Planning are used to describe the process to produce an accurate sales forecast. The feasible reason is that demand and supply continually have impact on each other. The scope of this research is limited to Sales Forecasting Management; managing the activities which are required to produce a high quality sales forecast. Companies can manage demand (e.g. by promotions, pricing strategies) and demand is constrained by the organisation (e.g. capacity

Figure 6: Context of Forecasting Demand

2.5 Research outline

After a short company introduction to Wavin, the research methodology is discussed; determining the structure of this thesis (figure 7). The theoretical framework discusses the theories used in this thesis; identifying two elements. First, a literature study was conducted to identify design criteria in relation to design of a forecasting process. Once determined, these design criteria and the forecasting audit (Moon et al., 2003) were applied to describe and analyse the current demand planning process. The output of this analysis is the identification of three ‘strategic themes’ - subjects to focus on for forecasting improvements. The redesign is concentrated on these strategic themes, and provides improvements to increase forecast performance. The last chapter discusses the conclusion; making distinction between conclusions from the research, about the research, recommendations to Wavin, and suggestions for future research.

Chapter 3: Theoretical Framework

This chapter discusses the theories which are used to answer the research question. It consists of an introduction of the concept demand planning, followed by the theory used to describe, analyse and redesign the current demand planning process. Once the theories are presented a link is made to the conceptual model and resulting research questions.

3.1 Demand Planning & Sales forecasting management

The subject of this thesis is conceptualised by sales forecasting management, which ‘is the management of the sales forecasting function within an organisation’ (Mentzer et al., 2005). Within Wavin, the demand planning process is designed to generate sales forecast. The process is generally concerned with the question which volumes of which products can be sold at which locations. As such, sales forecasting serves as a critical linking function between internal decision making and uncontrollable, external factors that have the potential to affect the demand for a firm's products (Mentzer & Moon, 2005).

Sales forecasting management involves the synthesis of internal and external (market & customer) information to estimate sales volume over a specified period of time. Both concepts of demand planning and sales forecasting are intertwined and mutual dependent, as expressed in the S&OP junction box (figure 5) and context of forecasting demand (figure 6).

3.2 Dimensions of sales forecasting management

Sales forecasting performance is determined by four dimensions affecting the forecasting process. These dimensions provide a structured framework to diagnose Wavin’s current forecasting practices.

Dimension Key element

I. Functional integration

 Organisational location of forecasting group  Existence & form of consensus forecast meetings

 Forecasting C3 _{between forecasting & other functional areas}

 Recognition of forecasting needs of various functional areas  Accountability/ performance rewards for personnel involved in

developing the forecast

II. Forecasting approach

 Relationship between forecasts & plans

 Orientation of the forecasting approach (top-down vs. bottom-up)  Understanding of forecast contribution to the supply chain  Forecast segmentation of products by importance

 Use of quantitative and qualitative forecasting techniques  Training in technique usage

III. Forecasting systems   Intra-company and supply chain electronic links Information availability (reports and performance metrics)  Degree of systems knowledge in the organisation

IV. Performance measurement

 Measurement and use of accuracy

 Recognition of the impact of external factors on accuracy  Measurement and use of other performance measures

(costs & customer service)

In total 17 key elements are grouped to four sales forecasting dimensions (table 3). Once characteristics of

key elements and dimensions are determined, these can be classified to a ‘stage’ of forecasting effectiveness; which are discussed in paragraph 3.3.

3.2.1 Functional integration

Functional integration is concerned with the role of collaboration, communication, and coordination of forecasting management with the other business functional areas of marketing, sales, finance, production and logistics (Moon et al., 1999). Functional integration consists of five key elements, discussed below. Organisational location of the forecasting group

The location of forecasting activities varies among organisations, and is an important indicator of the maturity of forecasting practices in organisation. Responsibility of forecasting is often found in Operations, Sales, Marketing, Logistics, and Finance. Until recently, Supply Chain was the department mostly assigned to forecast responsibility, while separate forecasting functions rapidly increase in popularity (Chaman, 2008). Four management approaches are distinguished in forecasting literature (Mentzer & Kahn, 1997), which collide with the location of forecasting responsibility.

Approach Responsibility for forecasting Drawback

Independent Each function has own forecasting efforts Ineffective and inefficient forecasting

Concentrated One function responsibility for forecasting Forecasting beneficial for one function

Negotiated Representatives per function combine input Functional bias can result in disagreements

Consensus Forecasting Champion/ Function Superior, but resource intensive

Table 4: Management Approaches to sales forecasting Existence and form of consensus forecast meetings

Consensus meetings are required to exchange cross-functional views on forecasts, and reach consensus and ownership over the sales forecast. Consensus is expressed in a single number forecast, which leads to superior forecasting practices and makes it easier to align demand with supply (Chaman, 2008).

Forecasting C3, between forecasting group and other functional areas

This concept relates to the degree of collaboration between forecasting and other functional areas (table 5). It divides functional integration in three components of Forecasting C3_{. The higher the stage of Forecasting} C3_{, the easier and more effective forecast information can be collected and processed in sales forecasts.}

Forecasting C3_{, Components of functional integration}

Communication The written, verbal and electronic information shared between organisational functions. Coordination The formal structure and required meetings between two or more organisational functions.

Collaboration The degree of which internal developers (functional areas) and external developers (key _{customers) aim to common goal setting i.e. common performance goals.}

Table 5: Characteristics of Forecasting C3

Recognition of forecasting needs of various functional areas

Function Forecast needs Function Forecast needs

Production Production cycle unit SKU based Logistics Order cycle unit SKUL based

Sales Quarterly Euro sales-territory-based Finance Annual Euro based

Marketing Annual Euro product-based

Table 6: Forecasting needs per Function Accountability/ performance rewards for personnel involved in developing the forecast

Apart from forecasting responsibility, accountability must be clearly defined. This includes close monitoring and involvement of forecast developers. To embed this accountability, several measures can be applied (e.g. clear rules, consensus). In addition, performance rewards are a strong measure to increase accountability. They direct behaviour, an aspect which must not be underestimated. Caution must be place in correctly determining a reward structure: rewards must be aligned with each other to elude politics and game-playing.

3.2.2 Forecasting approach

The forecast approach addresses the question which products are forecast and how they are forecast. It addresses the usage of techniques and the relationship between forecasting and planning (Moon et al., 2003).

Relationship between forecasts & plans

Forecast must be used in to facilitate business planning (Mentzer, 1999). Companies must also understand that forecasts and the business plan are intertwined: forecasts do not only affect plans, but also the other way around. As discussed in paragraph 3.1, demand can be managed.

Orientation of the forecasting approach (top-down vs. bottom-up)

There are two approaches to forecast orientation, covering three levels (Chaman, 2008). The levels identified are: aggregate, category and SKUL (Stock Keeping Unit per Location, often simply ‘product’, see glossary). A top-down approach calculates forecasts based on aggregate levels; e.g. combined product demand for a complete region. This demand consequently is cascaded down (often by regression techniques) to demand per product and location (i.e. SKUL). A bottom-up approach uses time-series techniques to calculate demand on SKUL-level, and aggregates this to combined - e.g. regional, domestic, worldwide - product demand. Category level aggregates demand to a ‘halfway-level’, where demand per category is calculated based on combined demand of individual SKU’s.

Understanding of forecast contribution to the supply chain

Understanding forecast contribution to the supply chain reflects how organisations see the role of forecasting and its impact on the business. Complete understanding is required to create cross-functional commitment of forecasting.

Forecast segmentation of products by importance

Use of quantitative and qualitative forecasting techniques

In forecasting literature, several groupings of forecasting models are distinguished. In general, a distinction is made between qualitative and quantitative forecasting techniques. Qualitative forecasting techniques rely mainly on judgmental data, while quantitative methods estimate relationships from quantitative data (Armstrong, 2008).

Training in technique usage

Sales forecasters must have a good understanding of sales forecasting techniques. This is only maintained when trainings in forecasting - e.g. for forecasters and other developers – are frequently organised. It involves issues as ‘how to use the forecasting system’ and ‘how to determine appropriate forecast methods’.

3.2.3 Forecasting systems

The system dimension encompasses the computer and electronic communications hardware and software used in sales forecasting (Mentzer et al., 1999). Forecasting systems enable large amounts of sales forecast information to be created, analysed and shared between users.

Intra-company and supply chain electronic links

Sales forecasting is a process that relies on intensive usage of information systems. This is required because large amounts of data need to be collected, processed, analysed, and transferred. For effective and efficient use, integration of forecasting systems is a must.

Information availability (reports and performance metrics)

Information availability is determined by the accessibility of reports and performance metrics. All forecast developers and uses must have easy access of forecasting reports. Sound enough; these must contain the performance metrics required by forecast users.

Degree of systems knowledge in the organisation

System knowledge covers the familiarity of forecast developers and users with the system. To make effectively use of forecasts, employees must have sufficient knowledge how to access and apply (e.g. customised forecast reports) forecast information.

3.2.4 Forecasting performance measurement

Performance measurement is of crucial importance for effective monitoring and control of sales forecasting. It considers the metrics used to measure sales forecasting effectiveness and its impact on business operations (Moon et al., 1999).

Measurement and use of accuracy

Recognition of the impact of external factors on accuracy

Both internal and external factors have impact on forecast accuracy. External factors, such as economic conditions and competitive actions affect demand; making it more difficult to forecast accurately. In addition, internal factors can also lead lower forecast accuracy; unfulfilled demand is partly a function of forecasting error and partially a function of operational error.

Measurement and use of other performance measures (costs & customer service)

As forecast accuracy has impact on overall business performance (Mentzer et al., 1998), these overall performance measures – often called multidimensional performance metrics - must be included in measuring forecasting performance (Moon et al., 2003). In addition, to increase forecasting ownership, forecasting KPI’s must be owned across functions; e.g. departments involved in the forecasting process.

3.3 Stages of effectiveness

In the previous paragraph, the four dimensions of sales forecasting have been explained. Performance on these dimensions can be linked to the sales forecast stages of effectiveness - or benchmark stages, from now on simply called ‘stages’ - displayed in Appendix B. Four stages are identified, each having characteristics corresponding to the management approaches discussed in paragraph 3.2.1. This relation makes sense, as an organisation’s management approach determines the level of functional integration, and all four dimensions are inextricably intertwined (Mentzer, 1999). Diagnosis of both the stages of effectiveness and design criteria determine strategic themes for improvement.

In practice, very few organisations have all forecast dimensions performing in the highest stages (stage 4). Stages of effectiveness provide two aims for management. First, they provide a broad set of targets which managers can aim for. As the consensus approach is superior, each organisation should strive to reach stage fourth characteristics. This really is worth achieving, as user satisfaction significantly increases when companies move from an independent approach (28% satisfaction) towards the consensus approach (70% satisfaction). Second, these concepts provide a tactical path to gradually improve forecasting practices, expressed by strategic themes. An overview of the sales forecasting dimensions, key elements, and their characteristics in each stage of effectiveness, is displayed in Appendix B.

3.4 Design criteria in forecasting

In the previous paragraphs the sales forecasting dimensions – and their stages of performance - are introduced. These combined theories enable to describe and assess the sales forecasting process; but to provide an answer to the research question (how to make improvement to the process), additional information is required. Design criteria consist of several elements which – when applied - can improve forecast performance. A table-wise overview of the literature study is displayed in Appendix A.

Six design criteria for an effective forecast design. The process must address…

…knowledge of (technical) forecasting approach: focus, simplicity & segmentation …clear rules and procedures

…clearly defined responsibility & accountability …well-structured cross-functional collaboration …clear performance management

3.4.1 Knowledge of (technical) forecasting approach: focus, simplicity & segmentation

A broad supported view in forecasting is to keeping it focused and simple. There is no proof that complex methods are more accurate than simple methods (Armstrong, 2005); often these only increase the risk of information overload (Mentzer, 2006). Another focus criterion is concerned with determining where, and when, forecasting efforts add value. Segmentation of products is performed to filter large groups of products and focus on the important products (Moon & Mentzer, 1999). The focus should be on the products having a significant impact on business performance (Wildhagen, 2011).

To achieve focus, organisations should define which products (SKU’s) and/ or categories are most (and least) important to forecast. This can be achieved by using an ABC-classification (Mentzer et al., 1998); sometimes referred to as the 80/20 or Pareto rule. This classification enables organisation to focus on high revenue (or volume) products (or customers).

Once these important products are distinguished, a second objective is to analyse the dataset, search for demand patterns and apply the right forecasting approach (Moon et al., 2003) and forecasting techniques. Demand patters indicate how product demand changed over time, and need to be modelled correctly to produce accurate forecasts. Several segmentations can help in attaining focus, such as price elasticity (Rego, 2011) and inventory impact (Milliken, 2011); which are displayed in Appendix D.

3.4.2 Clear rules and procedures

Typical for many cross-functional processes, clearly defined rules and procedures determine the success or failure of the forecasting process. Rules refer to both the technical steps (performed by the forecasters, mostly in the forecasting system) and cross-functional revision (performed by, and between, forecasters and other forecast developers).

Technical steps in the forecasting system

Eight steps of technical activities are distinguished to generate a sales forecast (Urs, 2008); these are viewed from the perspective of the forecasting system, and are displayed in table 8.

Sequence Technical process steps

1 2 3 4 5 6 7 8

Load historical data and create master data for planning Clean historical data

Generate statistical forecast

Prepare a forecast of new product introduction Override statistical forecasts with a judgmental input Adjust baseline forecast for promotion

Manage VMI and CPFR processes Generate one number forecast

can be quantitatively forecast. For products with scores above one, no conventional statistical techniques

must be used, and manual revision is advised.

Another important rule to address is the use of so called ‘expert systems’. These are model selecting tools which automatic select a forecasting method. The temptation to use this ‘easy ride’ comes with a drawback: real understanding of the driver of demand is not challenged when using his option. Forecasters are advised to only use this option for the least important products (Chaman, 2008). Two rules are essential in selecting the right forecast models: 1) regularly monitor performance of forecasting models and 2) use ex-posts forecasts to determine how good a model is (Chaman, 2008). Next to models, other examples of parameters to be monitored are segmentations, alerts, outliers (Rego, 2011) and exception rules (Milliken, 2011). Rules about the size and direction of qualitative adjustments are required, as research has proved that many ‘small’ adjustments decrease, rather increase accuracy (Fildes, 2008). Knowledge of fundamentals of models and modelling is important to attain focus, and necessary to link right model to the right demand pattern (Chaman, 2008). As product demand changes over time, demand patterns change with them; regular motoring of model performance is required (Chaman, 2008). Factors to consider in understanding demand patters are consistency of data, outliers and structural changes, seasonality, trends, forecast horizons and periods of history selected to base forecasts upon (Chaman, 2008).

Cross-functional revision

Information exchange also encompasses forecasters (forecasting system) and other forecast developers. A consensus meeting is already identified as important elements in forecasting process. It involves all functions in the forecasting process; enabling a company to tap into a large cross-functional pool of forecast information. Second, these meetings provide additional opportunities to include up-to-date forecast information in the forecast (reducing the need for forecast updates). Third, the cross-functional character brings functions on the same page; this which increases credibility and ownership (Chaman, 2008). To continually improve forecast performance a feedback loop must exist, evaluating all forecast developers (Wildhagen, 2011). This argument is supported by the requirement to record logic and measure value of all judgmental adjustments made to forecasts (Fildes, 2008).

Judgmental adjustments are used in qualitative forecasting techniques, and are useful when: 1) recent events are not reflected in the data, 2) historical data are limited and/ or 3) experts possess good domain knowledge about future changes that have not been included in the model (Armstrong, 2006). In addition, judgemental methods are used in situations where the 80/20% (Pareto) rule applies, where markets are highly volatile and where long forecast horizons must be covered (Chaman, 2008). Judgmental forecasts are ‘most likely to be accurate when the situation is well understood and simple, there is little uncertainty, and the experts receive accurate, timely, and well-summarized feedback about their forecasts (Armstrong, 2003). These methods are typically used for the introduction of a new product, changes in design, pricing or advertising and to predict competitor behaviour.

3.4.3 Clearly defined responsibility & accountability

All articles used in determining sales forecasting criteria recognised that tasks, responsibilities and accountability have to be clearly defined. Forecast participants need to know which tasks are required of them, who has responsibility for each step in the process, and who is end accountable. This is an important criterion for reaching commitment in forecasting, which is one of the critical success factors of forecasting performance (Moon et al, 2003). In addition, clear definition of tasks and responsibilities reduces the element of game-playing (Moon & Mentzer., 1999).

Accountability and responsibility are delicate elements in sales forecasting. A central subject of responsibility is the element of (de)centralisation. Centralisation is a solution which results in better control over processes, although caution must be taken. A common credo in centralisation is to organise decentralised what must be done decentralised, and to organise centralised what can be done centralised. A common example of centralisation of forecasting responsibilities is the assignment of forecast responsibility to an unbiased, credible role. Traditionally this was a Supply Chain role, but this function is rapidly losing terrain to a forecasting champion (e.g. manager) or special forecasting function (Park, 2008). Forecast consensus is a special concept which substantially increases accountability in forecasting. Consensus forecasts imply linking the ‘front-end’ (marketing, sales) with the ‘back-end’ (supply chain, operations) of the organisation, reconciling unconstrained and constrained forecasts (Rego, 2011). Consensus forecasts significantly increase accountability, as single number forecasts are nothing more than achieving consensus and holding everyone accountable (Park, 2008).

As forecasting is one out of three core responsibilities of Sales (next to generating revenue and maintaining customer relations), full commitment of Sales is essential in forecasting (Chaman, 2008). Four design criteria must be act upon to achieve commitment from Sales: 1) Make forecasting part of their job, 2) minimise game-playing, 3) keep it simple, and 4) Keep it focused. Disentangle sales goals from forecast goals (e.g. volumes vs. Euro’s) can help to minimise game playing.

3.4.4 Well-structured, cross-functional collaboration

This important factor is discussed in paragraph 3.2. As forecasting is a task which must be performed together, the needs (forecast requirements per function) of all forecast developers must be act upon (Moon et al., 2003). In addition, integrated forecasting systems must provide opportunities for easy communication and transfer of data (Mentzer & Moon, 1997).

3.4.5 Structured performance management

3.4.6 Commitment to forecasting

As management encompasses all activities to plan, do, control and act in the forecasting process, senior commitment is a crucial criterion in forecast improvement. Not surprisingly, ‘top management support is the top ranked predictor of implementation success, and lack of top management support is the top ranked predictor of implementation failure’ (Moon et al., 2003). Commitment encompasses sufficient availability of organisational resources to training (Park, 2008), forecasting software (Chaman, 2008), and reward systems.

Commitment from senior management significantly improves when the role of a forecasting champion is assigned, as attaining top management support is one of his key contributions (Mentzer et al, 1989). Among other factors, real senior management commitment requires concrete investments: trainings must be organised, a reward structure implemented, and forecasting KPI’s must be included in senior management compensation plans.

Many organisations committed to forecasting create a central Forecasting function. The head of this function is often referred to as the forecast champion (although the forecast champion can also reside in other functions). This role is responsible for coordinating the forecasting process; and in addition takes responsibility over forecast performance. Examples of forecast champion’s core activities are advocating forecasts as critical, company-wide management function, development of sales forecasting skills, and creating consensus in forecasting (Mentzer & Moon, 1997). This role is a great contributor to forecast improvement, as continuous improvement of any business function will not happen if it is not part of someone’s job (Mentzer et al., 1997).

3.5 Strategic themes

Strategic themes cut across multiple dimensions of forecasting management, and are so pervasive or cause such wide-ranging problems that they demand special attention and discussion (Moon et al., 2003). Based on the diagnosis these themes determine the focus for demand planning improvements.

3.5 Conceptual model

Concluding from the literature discussed in this chapter, the conceptual model (figure 8) of this research is presented. In the problem statement is expressed that Wavin’s process to produce a sales forecast is ineffective. Four forecasting dimensions, consisting of 17 key elements (black squares), identify elements affecting forecasting effectiveness. Performance stages are linked to each forecasting dimensions, resulting in a comprehensible framework to diagnose the current stage of sales forecasting effectiveness. The result of merging these theories leads to a model which is able to 1) identify four, comprehensible dimensions affecting sales forecast performance, 2) assess performance of these dimensions by a full range of potential ‘causes’ (key elements); and 3) provide a set of design criteria (arrows) for process improvements.

3.6 Sub-questions

Figure 8: Conceptual model CENTRAL RESEARCH QUESTION

Which adjustments to the current demand planning process are required to increase sales forecast performance?

SUB-QUESTIONS

1. How does the current design of Wavin’s demand planning process look like? What is the current stage of…

 …functional integration?  …the forecast approach?  …forecasting systems?

 …performance measurement?

2. To which extent does the demand planning process adhere to the identified design criteria? 3. Which strategic themes must be focused on in the redesign?

4. How should the demand planning process be redesigned? What are the implications for…

Chapter 4: Description current practices

To determine Wavin’s demand planning performance, the diagnosis is split into two steps. This chapter describes Wavin’s current sales forecasting practices, where the analysis in chapter five focuses on those aspects that have most priority to be improved. Together, these chapters provide a structured diagnosis which forms the basis for the redesign in chapter six.

This chapter describes current forecasting practices by two elements. First the forecasting audit (Moon et al., 2003), is used to assess the four forecasting dimensions and their characteristics (key elements). As some of the identified design criteria are addressed in the forecasting audit, the description focuses on the remaining design criteria.

The result of this chapter is dual of nature. First an overview of the forecasting dimensions and key aspects is provided. The ‘score’ of Wavin’s current forecasting practices is represented by number one to four. Each score resembles a performance stage; one is the lowest score and four the highest. The last part of this chapter covers current performance on the four design criteria; 1) technical forecast approach, 2) clearly defined responsibility & accountability, and 3) rules & procedures and 4) performance management.

4.1 Functional integration

The location of the sales forecasting function (and practices) in an organisation is the first dimension affecting overall success of the sales forecasting process.

Organisational location of the forecasting group

The practice of forecasting is the responsibility of the Sales department. Demand Planners are located in sales and coordinate the process. The Italian demand planner reports to the Director Marketing & Sales, the German Demand Planner reports to the Sales Director. Demand planning is organised at country-level. This implies that in principle, Demand Planners forecast for one market (i.e. country). In 2004, Wavin chose to combine forecasting activities; the German Demand Planner became responsible for forecasting the Belgium, Dutch and German market. Although the German Demand Planner officially reports to the Country Director, the German Demand planner states that the Supply Chain Manager for the region NWE is his ’officious boss’.

Existence and form of consensus forecast meetings

Once a month, a moment is planned in which forecasts are discussed. This takes shape by a phone call between Demand Planners and Sales Manager (the Netherlands, Belgium), or takes place by a face-to-face meeting (Germany, Italy). The attending roles in these demand meetings differ per country. In Italy, the Demand Planner meets with the Marketing & Sales Director. In Germany the Demand Planner is accompanied by a Sales Manager and some Sales reps (depending on their availability). In Italy and Belgium, no feedback between the Salespeople and Demand Planner takes place. The Sales function and a Supply chain role are not involved in the meeting, implying the complete line-up required for consensus is lacking.

SUB-QUESTION 1

Forecasting C3, between forecasting group and other functional areas

Three functional departments are involved in sales forecasting. Sales is responsible for forecasting, and requires information from the Marketing department; both are located ate the Operating companies (OC’s). The Marketing & Technology department is located at Central office and houses European Product Managers; these are responsible for providing forecast information for European product ranges (assortment sold in several countries). Product Managers are responsible for product ins and phase-outs, and inform Demand Planners about the impact on product forecasts.

An example of failing communication is provided by the Business Unit Director Hot & Cold, who gave an example where a product to be phased-out. This implies a stop in production and purchase of components. For some reason, a forecast for this product was still generated. Consequently, the BU-director stated that ‘one of our purchasers ordered fittings for a couple months’ sales volumes. This was a real waste, as these fittings could not be used in normal products’. This example illustrates that forecasting C3 between functions is not optimal.

Recognition of forecasting needs of various functional areas

Forecasting needs of forecast developers and users are assessed by checking five forecasting requirements (Mentzer, 2005). Forecasting needs are all recognised; an overview of the forecast requirements recognised by Wavin is displayed in Appendix D.

Accountability/ performance rewards for personnel involved in developing the forecast

Forecasting is currently concentrated at the Sales department, and determined as Sales’ responsibility. This setup was chosen to enlarge Sales’ commitment to forecasting. Procedures for forecasting new products (phase-ins) and discontinuance of old products (phase-outs) are undocumented. Documentation clearly defining accountability of forecast developers is non-existent. Forecast accuracy is included in the measurement of Demand Planners and Salespeople. Individual targets in regard to forecasting are not set; rewards or other incentives that stimulate forecast performance are non-existent.

4.2 Forecast approach

Wavin’s forecast approach encompasses what is forecast and how it is forecast. Relationship between forecasts & plans

Orientation of the forecasting approach (top-down vs. bottom-up)

Wavin forecasts on category-level; this implies a top-down orientation. Planning groups are used as a standard working method. These planning groups are created by Demand Planners, and contain bundles of assortments (which in turn contain individual products). When this approach leads to products (SKUL-level) forecast incorrectly, manual adjustments are performed by demand planners. The planning groups are an attribute of lean forecasting (Kahn & Mello, 2005); it saves Demand Planners and Salespeople a considerable amount of time. Rather managing (forecasting) tens of thousands of individual products, only about 30 to 90 planning groups have to be managed (Appendix E).

Understanding of forecast contribution to the supply chain

Within Wavin it is known that forecast accuracy has large impact on the supply chain (figure 1). Nonetheless, the importance and impact of forecasting is not understood by Sales (the function responsible for forecasting). An important factor to understand is that forecast accuracy is affected by capacity constraints. In the current setup, judgmental adjustments to baseline forecasts are only provided by Sales and Marketing, lacking ‘constrained’ forecast information. Incorrect or insufficient use of forecasts is also expressed by the Project Manager Business tools, stating that ‘the business unit Cable Ducting in France doesn’t look at the demand plan. In that case a Sales department can provide input for the demand planning process, but for whom are you doing it for? This is a good example that a coherent and shared understanding of forecasting currently is not present.

In Germany, an additional forecasting procedure is in use. Apart from the ‘ordinary’ sales forecasts, a financial forecast is calculated for the German market. This forecast is driven by the same information as the sales forecast, but links an average selling price to forecast volumes; resulting in a Euro-based forecast. This forecast is used for financial planning and, in addition, provides a format for the forecast forms which are submitted by sales. The Sales function lacks clear understanding in forecasting and consequently does not take responsibility over the process. The result is a situation where rules are unclear, not adhered by, and do not lead to effective forecasting.

Forecast segmentation of products by importance

Several segmentations are offered by the system (Appendix D), although correct use (focus) of this segmentation is lacking. Some products are indicated as important; a priority index of some kind (including definitions, rules, and criteria) is clearly lacking. As a result, demand planners lack focus; many manual corrections are performed at various products, rather than structurally focusing on ‘critical’ products. Differentiation of important products is not reflected in accuracy targets: one target is applied: forecast accuracy of 85% on A-products (see Glossary). This focus on A-products is an important segmentation, yet more segmentation by importance is required.

Use of quantitative and qualitative forecasting techniques