Comparing the Outcomes of Two Decision Support Models: The Analytical Hierarchy Process and Pugh Matrix Analysis

(1)

Comparing the Outcomes of

Two Decision Support Models:

The Analytical Hierarchy

Process and Pugh Matrix

Analysis

Using an actual multi-criteria decision-making situation

Jämförande av två beslutsstödjande modellers utfall: den Analytiska hierarkiska processen och Pughs matris analys

Med hjälp av en verklig multikriteriebeslutsfattande situation

Madeleine Burgren & Lina Thorén

Faculty: Health, Science & Technology Subject: Industrial economics

Points: 30 ECTS

Supervisor: Jonas Berghel & Dan Nordin Examiner: Mikael Johnson

(2)

(3)

Preface

This master thesis represents the final requirement of the five-year program Industrial Engineering and Management at Karlstad University. The research subject was defined during the fall 2014 based on the authors’ education and interest in the area of technology and business. The motivation of the thesis comes from the challenge companies are facing when making strategic decisions. Furthermore, combining the technical, environmental and business knowledge gives better insight in how to interweave those areas. The study also inspired cooperation with both Åmotfors Energi and Sweco in Karlstad and Gothenburg.

We would like to thank the people at Sweco for their support, and especially our supervisor, Monika Bubholz, who has been showing extensive interest through the whole process. Additionally, we would like to thank Jan Hallgren and Pär Höglund at Åmotfors Energi who always have been showing their encouragement in our work and provided us with necessary information to make the thesis complete. Furthermore, we would like to give a great thanks to our supervisors at Karlstad University, Jonas Berghel and Dan Nordin. Their collaboration has been valuable for the development of the thesis. Last but not least, we would like to thank our family and friends who have been a part of our journey and supported us all the way through.

__________________ __________________

Lina Thorén Madeleine Burgren

Karlstad University, June 2015

(4)

Abstract

Since businesses are constantly changing, making right decisions is a critical factor in order to achieve good results. In the thesis, two different decision support models are tested and the outcome is compared. This is done in cooperation with a company, Åmotfors Energi, who is facing a decision on how they can make use of their 30 GWh heat which they today do not have a paying customer for. Nine alternatives are used in the models and evaluated with seventeen different criteria. The purpose of this study is to compare and interpret the outcomes of two decision support models: the Analytical Hierarchy Process and Pugh Matrix Analysis. The purpose is also to investigate the main factors that influence the outcome of the models.

The main research strategy was to use experimental design where three experts with various technical skills have scored the alternatives in both models. The alternatives have been carefully developed through an idea generation and idea selection phase.

The results show that the models give different result when it comes to ranking the alternatives, both between the models and between the different experts. The empirical findings establish that the outcome from the models should be interpreted that the lowest scored alternatives can be eliminated for further research. The alternatives with the highest score should be further investigated before a decision could be made. Furthermore, what mainly affects the result is based on human factors.

Keywords: Multi-Criteria Decision Making, Decision Support Models, Analytical Hierarchy Process & Pugh Matrix Analysis

(5)

Sammanfattning

Företag står inför ständiga förändringar och att fatta rätt beslut ses som en kritisk faktor för att uppnå goda resultat. I denna uppsats testas två beslutsstödjande modeller där utfallet av dem jämförs. Detta görs med hjälp av företaget Åmotfors Energi som står inför ett beslut om hur de kan använda deras 30 GWh värme som de i dagsläget inte har någon betalande kund för. Nio alternativ används i modellerna och utvärderas med sjutton uppsatta kriterier. Syftet med denna studie är att testa, jämföra och tolka resultatet från två beslutsstödjande modeller, den Analytiska Hierarkiska Processen och Pughs Matris Analys. Syftet är också att utreda vilka huvudfaktorer som påverkar utfallet av modellerna.

Den huvudsakliga forskningsstrategin var ett experiment tre experter med olika tekniska färdigheter har poängsatt de olika alternativen i modellerna. Alternativen är omsorgsfullt framtagna genom en idégenereringsfas och en idéurvalsfas.

Resultaten visar att modellerna ger olika resultat när det kommer till att rangordna alternativen, både vid jämförelsen mellan modellerna men även mellan experterna. Den empiriska studien visar att resultatet från modellerna bör tolkas som att de lägst rankade alternativen kan uteslutas, och de högst rankade alternativen bör utvärderas vidare innan ett besluta kan tas. Det som huvudsakligen påverkar resultatet baseras på mänskliga faktorer.

Nyckelord: Flerkriteriumbeslutsfattande, Beslutsstödjande modeller, Analytiska hierarkiska processen & Pughs matris analys

(6)

List of tables

Table 1-‐ An example of the matrix when the weighting is filled in ... 9

Table 2 -‐ An example when a reference alternative and the scoring are filled in ... 9

Table 3 -‐ Total weighted score ... 10

Table 4 -‐ An example of a traditional scoring scale ... 12

Table 5 -‐ Random index (Saaty 2013) ... 14

Table 6 -‐ Alternatives for evaluation ... 28

Table 7 -‐ Criteria and weighting factors ... 29

Table 8 – Traditional scoring scale ... 31

Table 9 -‐ Random index (Saaty 2013) ... 34

Table 10 -‐ Result from evaluation with PMA and AHP ... 34

Table 11 -‐ Result from expert no.1 AHP and PMA ... 35

Table 14 -‐ Investment cost greenhouse ... 43

Table 15 -‐ Costs, Revenues and Margin-‐ greenhouse ... 43

Table 16 – Payback period and present value – greenhouse ... 45

Table 17 -‐ Costs, Revenues and Margin -‐ drying timber ... 49

Table 18 -‐ Investment cost – drying timber ... 49

Table 19 -‐ Payback and present value -‐ drying timber ... 50

List of figures

Figure 1 -‐ A typical decision matrix ... 7

Figure 2 – A simple example of an AHP model ... 11

Figure 3 -‐ Waste hierarchy ... 17

Figure 4 -‐ Waste to energy (Sysav 2003 referred to in Renhållningsverksföreningen 2005) ... 18

Figure 5 -‐ An overview of the research’s process steps ... 20

Figure 6 -‐ Process from idea generation to idea development ... 22

Figure 7 – Construction of the Pugh matrix used in the experiment ... 30

Figure 8 – Construction of AHP matrix in the expiriment ... 31

Figure 9 – Matrix [A] ... 32

Figure 10 – Matrix [A] together with the normalized numbers (b) ... 32

Figure 11– All numbers divided with the sum of each row ... 32

Figure 12 -‐ Energy balance in a greenhouse (After Möller Mielsen 2007) ... 39

Figure 13 – Temperature in the greenhouse during one year ... 44

Figure 14 -‐ Heat demand in greenhouse and heat available at Åmotfors Energi ... 45

Figure 15 -‐ Batch kiln (After Träcentrum 2015) ... 47

Figure 16 -‐ Schematic picture of the drying process (After Dincer & Sahin 2004) .... 48

(9)

List of abbreviations

AHP Analytical Hierarchy Process

CHP Combined heat and power

CI Consistency index

CR Consistency ratio

CSR Corporate social responsibilities

MCDM Multi-criteria decision making

PMA Pugh Matrix Analysis

PV Present value

RI Random index

(10)

(11)

1. Introduction

This chapter aims to give the reader a good insight into what this study is about. It presents the background and purpose, followed by the research questions that lay the foundation for the study.

Background 1.1.

Today’s businesses and organizations are changing rapidly which has made decision making an important part of organizational operations to achieve successful results. Managers have to choose the right strategy in order for the organization to increase their profit (Akdere 2011; You et al. 2015). One of the fastest growing problem areas during the last decades is multi-criteria decision making (MCDM). Decision making has gone from a single-criteria decision environment, where profit was the only important criteria, to a multi-criteria situation where it has been significant to consider several aspects that are important for the stakeholders (Triantaphyllou et al. 2000). Thus, MCDM is when an organization is faced with several alternatives which need to be evaluated while taking numerous criteria into account (Thokala & Duenas 2012). However, it is being used in several business areas and it is often observed in engineering, scientific, and technology fields (Mardani et al. 2015; Frey et al. 2007; Triantaphyllou et al. 2000). MCDM is correspondingly used in finance (Steuer & Na 2003).

There are several MCDM models1_{to choose from, and it is difficult to know}

which one to use for a specific situation (Triantaphyllou et al. 2000). The two models investigated in this thesis are the Analytic Hierarchy Process (AHP) and Pugh Matrix Analysis (PMA). They are both well-known models applied in MCDM (Cervone 2009; Renzi et al. 2013; Rezaei 2015). AHP is a popular model that has been applied in different areas such as planning, selecting the best alternative, resource allocation and resolving conflicts (Subramanian & Ramanathan 2012). PMA is often used when a decision has to be made where a number of alternatives are presented and the inventor, Stuart Pugh, designed the model in order to select between design alternatives.

Further, the goals of the process are to converge on a strong alternative and help the team understand the reasons for their choice (Frey et al. 2007). AHP

(12)

is similar to PMA; it is a model that considers different criteria (Saaty 2008). However, AHP is using pairwise comparison of the criteria, which makes the model more complex than PMA (Rezaei 2015).

Renzi et al. (2013) state that several researches have been done using different MCDM models, however, a few investigations actually compare different models. Even though the models are well known, only a small fraction of engineers are using them. One reason for the limited use may be due to the fact that the experts in the industries rely on their expertise and know-how. Renzi et al. (2013) and Triantaphyllou et al. (2000) state that comparing different MCDM models is challenging but also urgent because of the subjectivity and complexity involved in many MCDM models.

Åmotfors Energi is a Swedish company producing heat and electricity through waste incineration, and they are facing a MCDM situation. They need to find an advantageous alternative to take care of 30 GWh excess heat they get from their process per year. The company use waste as their fuel and partly because of their location, they have no customers in the closest area. Therefore, to broaden their horizon, it is important to alternate different ideas before starting the MCDM process. The choices they have are: selling the resource to a customer farther away, building a business and running it by themselves, or sell the heat to someone that can run a business in Åmotfors. For example, the heat can be used for heating different premises, drying processes and water-heating processes. Since the company has several possibilities, this thesis problematization was partly to tackle it through a multi-criteria decision-making perspective by choosing reliable and suitable decision support models to evaluate alternatives.

Problem 1.1.

(13)

them and how the result should be interpreted. To sum up; to models are not as widely used as they could be by companies, one reason for this is that they are seen as complicated and time consuming and it is uncertain on how they are supposed to be applied.

Purpose 1.1.

The purpose of this thesis is to compare and interpret the outcome of two decision support models, AHP and PMA. The purpose aims to investigate how the results from the models should be interpreted and which factors that are influencing the outcome of the models. The comparison is applied on a real MCDM situation on behalf of a company. The purpose with applying it on a real situation is to show if the models can be used in practice and be valuable for companies. The authors’ intention is to contribute to the research and theoretical development in the MCDM field concerning support models, how the answer should be interpreted and which factors that influence the outcome.

Research questions 1.2.

To fulfill the thesis purpose, the following research questions will be answered:

RQ no. 1: How should the outcome from AHP and PMA be interpreted? RQ no. 2: What are the main influencing factors on the outcome when using

each decision support model, AHP and PMA? Limitations

1.3.

(14)

2. Theory

This chapter consists of the theoretical framework which lays the foundation for the empirical study. It explains the basic theories of Multi-Criteria Decision Making as well as how the Analytical Hierarchy Process and Pugh Matrix Analysis work.

Decision making 2.1.

All humans make decisions, consciously or unconsciously. Information is necessary to help us understand and to make judgments; however, too much information does not result in better understanding or better judgments (Saaty 2008). The term “decision-making” was introduced in the mid 1900s to the business world and came to replace the phrases “resource allocation” and “policy making” (Buchanan & O’Connell 2006). Further, Buchanan & O’Connell (2006) mean that decision-making entails deliberation, and it is a start of achievement. In contrast to resource allocation and policy making, where both are seen as endless, decision making has a more definite line with a start and an end (Buchanan & O’Connell 2006).

Decision making is a well-known process in every organization and it has an impact on all levels including individuals, groups and the organization itself (Akdere 2011). Furthermore, Akdere (2011) states the value of having a systematic decision-making process in order for organizations to grow and get economic benefits that lead to successful outcomes. Holton & Naquin (2005) describe that:

Decisions are assumed to be made through a linear and logical sequence of steps: identification of the problem or issue requiring a decision; collecting and sorting information about potential solutions; comparing each solution alternative against predetermined criteria; ranking possible solutions; and finally, selecting the optimal alternative. The goal is to maximize rewards and minimize costs simultaneously (p.260)

Akdere (2011) mentions other reasons why organizations use decision making: the process admits members to gain ownership in the decision chosen, it eliminates, to some extent, the top-down management and it also reduces the resistance of change among the employees in the organization.

(15)

competition on the market. Further, these aspects create difficulties when choosing the right decision-making strategy in order for companies to make profit. Scheibehenne & von Helversen (2015) explain how the literature distinguishes between compensatory and non-compensatory strategies where the compensatory strategy makes decisions from a holistic perspective and weighted positive and negative values. The non-compensatory strategy has its focus on the attribute, which means decisions are compared based on their validity or importance. Moreover, both strategies are frequently used in organizations and they are dependent on the amount of information available. Compensatory strategies consider all relevant information, while non-compensatory strategies focus on specific details and exclude irrelevant data (Scheibehenne & von Helversen 2015).

Zellman et al. (2010) describe the two fundamental theories in decision making: Neoclassic theory and Bounded Rationality theory. However, there are two factors that are contested: the amount of information used by the individuals that are making the decision and how they integrate the available information. The Neoclassic theory goes through a process where the decision maker carefully considers all alternatives in order to select the best one, based on the impact of attribute and importance. It is essential that the decision maker integrates all available information about all alternatives when making the decision. Further, Zellman et al. (2010) explain some criticism to the theory whereby individuals do not have unlimited time to analyze all alternatives and fully get a deep understanding.

(16)

Multi-criteria decision making 2.2.

A way to structure and support decision makers when evaluating different alternatives/options/choices is by multi-criteria decision making (MCDM). Further, the models categorize the most favorable alternative in relation to the importance of the criteria, and also the performance of the alternative on the criteria (Thokala & Duenas 2012). MCDM was defined in the 1970’s whenever operational problems, such as production scheduling and inventory control, moved to a higher level, meaning tasks needed more planning and decision making (Stewart 1992). However, the analysis of the way people make decisions is perhaps as old as mankind. MCDM is one of the most well-known branches of decision making (Triantaphyllou et al. 2000). Stewart (1992) mentions how MCDM became a formal tool for problem solving, or in other words: mess reduction. Clearly, MCDM has developed to its own discipline with its own rights.

Today companies are complex and have to take different stakeholders’ perspectives into account when making decisions. Subjects such as risk, liquidity, social responsibility, environmental protection and employee welfare become vital. Thus, in many financial decisions, it is appropriate to have a multi-objective approach (Steuer & Na 2003). According to Mardani et al. (2015), there lies a degree of uncertainty whenever a decision will be made, which can be due to lack of information or differences in judgments between people because of different perception or personality. This uncertainty has been treated in MCDM models, and instead of using natural languages, which may not always be well defined, the uncertainty has been focused on theory and statistics.

Stewart (1992) explains the two different contexts where MCDM models can be used. In the first one, the decision maker would either be an individual or a homogenous group seeking after a decision with minor impact to other stakeholders. In those cases, the documentation does not have to be so thorough and the model can be informal. The second context, which needs a more formal model and clearer documentation, is whenever the decision maker has to make a decision in a larger group, or when the decision maker arranges a list of alternatives which further will be investigated and considered by someone else (Stewart 1992).

(17)

they need to get trained before using it. Further, the model may in several cases rely on software data or programs that need to be identified (Thokala & Duenas 2012).

Triantaphyllou et al. (2000) describe the certain aspects that many of the MCDM models have in common; more precisely, they are the notion of alternatives and criteria. The alternatives are the different choices of action available to the decision maker, which should be screened, prioritized and ranked. Criteria represent the different dimensions from which the alternatives can be viewed. If there is a large amount of criteria, they are usually arranged in a hierarchical manner. In most of the MCDM models, it is required that the different criteria are weighted by importance. Furthermore, Triantaphyllou et al. (2000) state that MCDM situations easily can be expressed in a matrix. A is

the decision-making matrix which is a 𝑚×𝑛 matrix where the element aij

indicates alternative Ai when it evaluates in terms of criteria Cj for i=1, 2, 3,

…, m, and j=1, 2, 3, …, n. The decision maker also has to determine the weight for every criteria (wj, for j=1, 2, 3, …, n), see Figure 1.

Criteria Alt C1 C2 C3 … Cn (w1 w2 w3 … wn) A1 a11 a12 a13 … a1n A2 a21 a22 a23 … a2n . . . . . . . . . . . . . . . . . . Am am1 am2 am3 … amn

Previous research comparing MCDM models 2.3.

Renzi et al. (2013) mean that there is several decision support models to utilize in the industry, but only a few of them are actually used. Further, Renzi et al. (2013) argue that comparisons of different decision support models would improve the knowledge transfer between design research and industrial practice. It would also result in efficient workload among the employees and practitioners whenever they get a broader insight about the models. The

(18)

comparison between the Fuzzy-Analytic Hierarchy Process (F-AHP) (which is a variation of AHP) and Pugh’s Controlled Convergence (PuCC) (which is a variation of PMA) made by Renzi et al. (2013) shows the advantages and disadvantages with the models, and a real situation was used in order to compare the models more thoroughly.

In cooperation with an Italian manufacturer, the models were applied to find an innovative and low-cost method to increase the life of the heel tip in women’s shoes (Renzi et al. 2013). The results drawn from this research do not show which model that is best to use; however, Renzi et al. (2013) argue that one comparison is not enough to make that statement. Instead it is valuable for practitioners to make that decision according to their problem and through knowledge transfer.

In the book Multi-criteria decision making methods: a comparative study of Dr. Triantaphyllou, Triantaphyllou et al. (2000) are focusing on some of the best known and most frequently used MCDM models. The models are compared and show a number of abnormalities, which the users often are unaware of. Furthermore, the research showed that there is no single model which outperforms all the other models in all aspects. Triantaphyllou et al. (2000) claim that to determine which model is best, the model would first need to be used by that person. Thus, a decision paradox is reached. Triantaphyllou et al. (2000) are comparing a number of the best-known models and explain abnormalities that users often are unaware of. It is concluded that even under the best circumstances, the decision maker is a rational person and the involved information is perfectly known; different models may result in different conclusions for the same problem. It is pointed out that the decision maker always should be vigilant before accepting the result of a MCDM model.

Pugh Matrix Analysis

2.3.1.

(19)

than the market leader. Further, his intention was to give people involved in the process knowledge and good understanding about why the final choice was made (Frey et al. 2007).

Cervone (2009) divides the process into seven steps. First the different criteria are chosen; Lindstedt & Burenius (2003) argue that fifteen criteria is a good amount in order to distribute a respectable overview. The second step is to select the concepts to be compared in a matrix, followed by the third step which is drawing the matrix. The fourth step consists of weighting the different criteria, see Table 1. One critical point when weighting the criteria is that the criteria will not be equally important. To approach this task, one proven technique is to allow each member in the team to assign individually his or her weight to each criterion, which then are averaged and used in the final matrix Cervone (2009). According to Cervone (2009), ten points are used to be distributed among the criteria but with larger numbers of criteria; the numbers to be distributed can be increased.

Table 1-‐ An example of the matrix when the weighting is filled in

Weight Alternative 1 Alternative 2 Alternative 3 Criterion 1 4

Criterion 2 2 Criterion 3 3 Criterion 4 1

In step five, a reference alternative is selected which all the other alternatives will be compared with, and step six consists of filling in the scores for each alternative. Each alternative is compared to the reference alternative followed by a score that is better (+1), the same (0), or worse (-1), see Table 2. Finally, the scores for each alternative are added with the weight, which is multiplied with the given score, see Table 3. Cervone (2009) claims that the alternative with the highest score is not necessarily the most important, but the alternatives with higher scores are of main concern and worth considering more carefully.

Table 2 -‐ An example when a reference alternative and the scoring are filled in

Weight Alternative 1 Alternative 2 Alternative 3

Criterion 1 4 0 +1 -1

Criterion 2 2 0 0 -1

Criterion 3 3 0 -1 +1

(20)

Table 3 -‐ Total weighted score

Weight Alternative 1 Alternative 2 Alternative 3

Criterion 1 4 0 +1*4 -1*4

Criterion 2 2 0 0*2 -1*2

Criterion 3 3 0 -1*3 +1*3

Criterion 4 1 0 -1*1 0*1

Total score: 0 0 -3

Renzi et al. (2013) claim that PMA is a good evaluation tool when it comes to handling several criteria. Renzi et al. (2013) describe that the model has a simple score when filling in the matrix, and the pairwise comparison is easy even with several alternatives. Further, the model opens up for heuristic thinking and it is suitable for group decision making in the terms of promoting communications in team discussions even though disagreement could be difficult to overcome. Another disadvantage with PMA, according to Renzi et al. (2013), is its unsuitability for software implementation.

The Analytic Hierarchy Process

2.3.1.

One MCDM model based on mathematical calculations is the Analytic Hierarchy Process (Renzi et al. 2013; Subramanian & Ramanathan 2012). Subramanian & Ramanathan (2012) state its use in different areas such as planning, selecting a best alternative, resolving conflicts and resource allocation. It is a structural process where the most important criteria are provided, and the leading design or alternative is selected among several options. The inventor of the model, Saaty (2013) states how it is a physiological way of making a decision. It is intended to put focus on the parts which influence the question instead of organizing thoughts and judgments that come with it. Further, Saaty (2002) describes the different steps in the model which can be seen as a hierarchy-structured process. The steps are described below.

Step 1. Structure a decision tree

(21)

and which criteria they shall be compared to (Subramanian & Ramanathan 2012).

Step 2. Construct a set of pairwise comparison matrices

Further, the elements in the higher level in Figure 2 are compared with the lower elements. Subramanian & Ramanathan (2012) explain how AHP provokes several extra judgments. If the AHP has n alternatives, there has to be n comparisons. There is a total need of n(n-1)/2 judgments in order to rank

n alternatives. One matrix has to be constructed for each and every

comparison.

Step 3. Pairwise comparison of the alternatives

The final step consists of weighting the priorities for every alternative. In order to make a comparison, there has to be some sort of scale that indicates how much more important one element is than another. According to Subramanian & Ramanathan (2012) and Saaty (2008), a traditional or fundamental AHP scale is favorable and goes from 1-9 and only consists of absolute numbers, see Table 4 Saaty (2008) also says that numbers from a standard scale are considered objective, however, the interpretation is always subjective. When the pairwise comparisons are made, the verbal terms are transformed into numbers as in Table 4.

Goal

Criterion 1 Criterion 2 Criterion 3 Criterion 4

Alternative 1 Alternative 2 Alternative 3

(22)

Table 4 -‐ An example of a traditional scoring scale

Scoring approach

Intensity of value Interpretation

1 Alternative i and j are of equal value

3 Alternative i has a slightly higher value than j

5 Alternative i has a strongly higher value than j

7 Alternative i has a very strongly higher value than j

9 Alternative i has an absolutely higher value than j

2,4,6,8 These are intermediate scales between two adjacent judgments

Reciprocals If alternative i has a lower value than j, example (1/5)

Subramanian & Ramanathan (2012) describe that the alternative with a higher value usually is seen as more attractive compared to the alternative with a

lower rating. Each entry, aij (the importance of alternative i over alternative j)

in the matrix is overseen by three rules: aij> 0; aij= 1/aij and aij= 1 for all i. If

the matrix is consistent, the mathematical relation has to be: aij= aik*akj.

Those numbers in the scale are used in the AHP model and inserted in a

matrix. For each criterion, a matrix is created based on the weighting values aij,

which was a result of the pairwise comparisons of the alternatives.

How to construct the matrices and complete the comparisons:

Matrix A

Criterion 1 Alt. 1 Alt.2 Alt. 3

Alt. 1 [a11 a12 a13]

Alt. 2 [0 a22 a23]

Alt. 3 [0 0 a33]

(23)

Matrix B

Alt. 1 [a11 a12 a13]

Alt. 2 [1/a12 a22 a23]

Alt. 3 [1/a13 1/a23 a33]

Now, each column in the matrix is normalized, which means that each number in the matrix is divided by the sum in each column. More clearly it can be expressed as:

b1= (a11 + 1/a12+ 1/a13), b2= (a12+ a22 + 1/a23), b3= (a13 + a23 + a33)

This gives a the new matrix: Matrix C

Alt. 1 [a11/b1 a12/b2 a13/b3]

Alt. 2 [(1/a12)/b1 a22/b2 a23/b3]

Alt. 3 [(1/a13)/b1 (1/a23)/b2 a33/b3]

An average value for each row is then calculated as follows (n = the number of the criterion):

(a11/b1+(1/a12)/b1+(1/a13)/b1)/3 = cn1

(a12/b2+a22/b2+(1/a23)/b2)/3 = cn2

(a13/b3+a23/b3+a33/b3)/3 = cn3

When the alternatives are multiplied with the weighing factor of the criteria, the results are given:

Alt. 1 v1*c11+v2*c21+…+vn*cn1 = result alt.1

Alt. 2 v1*c12+v2*c22+…+vn*cn2 = result alt. 2

Alt. 3 v1*c13+v2*c23+…+ vn*cn3 = result alt. 3

(24)

Important things to have in mind when doing the AHP

Saaty (2008) explains how the judgments can be inconsistent and therefore they have to be measured by a Consistency Ratio (CR):

𝐶𝑅 =𝐶𝑜𝑛𝑠𝑖𝑠𝑡𝑒𝑛𝑐𝑦 𝑖𝑛𝑑𝑒𝑥 (𝐶𝐼)

𝑅𝑎𝑛𝑑𝑜𝑚 𝑖𝑛𝑑𝑒𝑥 (𝑅𝐼)

If the matrix is consistent, CR=0 but if CR> 0.1, the scoring has to be rechecked by the decision maker in order to make more consistent judgments (Subramanian & Ramanathan 2012) To compute CR, the priority

vector/normalized Eigen vector (λmax) needs to be calculated. The Eigen

vector λmax is the highest value in Matrix C above. The degree of consistency is

then measured by the following formula: 𝐶𝐼 =(𝛌!"# − 𝑛)

(𝑛 − 1)

By comparing CI with a Random Index (RI), Saaty (2013) introduced the appropriate CI, named as the Random Index seen in Table 5.

Table 5 -‐ Random index (Saaty 2013)

Order (n) R.I. First order difference

(25)

Saaty (2013) mentions how the AHP becomes successful in practice whenever the structure is simple and there is an adaptation to both individuals and groups. Some observations made in the AHP process were that humans are more likely to compare the physical stimuli (a thing or an event that causes a reaction) and the response to it, instead of using logical thinking. Through the stimulus-response theory a numerical value scale has been constructed in order to link it together homogenously. As an example, Saaty (2013) explains the difficulties in comparing the size of ants with the size of melons. Since people are capable of pinning down their judgments of the dominance of one criterion compared to another, the size of melons compared to the size of ants has to be linked together homogenous.

Another observation made when it comes to the process is that complex decision making needs to be organized and creative thinking has to be used when structuring a problem. This is either done hierarchically or by a network. One last observation made was that individual and group judgments can be combined.

Subramanian & Ramanathan (2012) mention that AHP was the most popular model to use for sustainable energy planning compared to two other MCDM models. Additionally, the model strength lies in considering subjective opinions of the decision makers. However, the model has become interesting to combine with other models more specified to handle objective data (Renzi et al. 2013). Saaty (2013) states that AHP is a good model when integrating hard data together with subjective judgments. It also offers guidance and monitoring of organizational performance toward dynamic goals. A disadvantage with AHP, according to Renzi et al. (2013) is the loss of information through the judgments and furthermore the crisp numbers in the model; thus the uncertainty factor is not considered.

(26)

3. Åmotfors Energi

This chapter includes a description of Åmotfors Energi, which represents the company for this study. It also gives a deeper insight into the MCDM situation the company is facing and an explanation about waste as an energy source. The authors’ intention is to make the reader fully understand why the company is facing a MCDM situation. Further, the intention is to give a brief view about the company’s process.

Åmotfors Energi 3.1.

To complete the comparison of the outcome from AHP and PMA data, alternatives, criteria and weighting factors were needed. The task was given to the authors of the thesis from Åmotfors Energi in order to help the company find what possibilities they have in order to make a profit on their unutilized heat resource on 30 GWh per year.

Åmotfors Energi is a company producing 175 GWh of heat and electricity through waste incineration. They have, during recent years, optimized their production of heat and they are therefore able to produce 30 GWh more per year compared to when it was founded. However, there is no customer in a

closer distance able to buy the heat2_{. It would be desirable for Åmotfors}

Energi to make a profit on all of their heat production, including the 30 GWh. Åmotfors Energi have to find another solution for selling their heat, or build a business where they can use their heat for production of another product in order to make a profit. In this thesis, possible alternatives for using the heat have been produced by known techniques which further are evaluated with the AHP and PMA model to generate a good solution for the company. Furthermore, it is important to know that the 30 GWh can be delivered in a temperature and at a pressure that are desirable for the customer.

The company was founded in 2008 and it has 20 employees and a turnover on 75 million SEK per year. Åmotfors is a small town with approximately 1,500 inhabitants situated in the county of Värmland, Sweden. The Belgian supplier Keppel-Seghers was responsible for the process equipment (Hall, K. &

Larsson, K-A 2010). The Combined Heat and Power (CHP) plant is 2,200 m3

and it has a capacity to take care of 80,000 ton of households per year and

according to Hallgren3_{, approximately 70,000-72,000 tons per year are being}

used. The households mainly come from the county of Värmland and from the eastern part of Norway. Hall & Larsson (2010) state that the idea of using

(27)

households as an energy source is because of the economical benefits for

Åmotfors Energi but also for the suppliers. Further, Hallgren2_{mentions that}

the company produces approximately 155 GWh of steam and 20 GWh of electricity per year.

Energy from waste 3.2.

According to Sopor.nu (2015), waste is a material, object or substance which the holder discards, intends to discard or is required to discard. The waste can come from households, offices, schools, restaurants etc. From an environmental perspective, the Waste Framework Directive has defined a waste hierarchy, which describes how waste should be managed. The hierarchy is often drawn as stairs as in Figure 3. It would be better for the environment if the amount of landfilling decreased and the amount of recycling and energy recovery increased instead.

Both methods are beneficial since it produces either material or energy (Avfall Sverige. 2009; Sundberg 2013). Energy recovery from waste can be done by incineration with a CHP plant. This does not only decrease the amount of landfilling, but it also influences the fuel used for production of heat and electricity (Eriksson et al. 2005). Political decisions toward more sustainable waste management has resulted in that 15 percent of the European municipal waste flow needs to be redirected from landfilling to other treatments (Eriksson et al. 2005). When energy is recovered from waste for use in the district heating system, the Swedish waste management is also effected by the energy systems. The Swedish energy system is required to gradually decommission their nuclear power and, in doing so, it introduces renewable energy sources on the energy market. Waste can partly be seen as a renewable energy source. According to Avfall Sverige (2014), the European Union (EU)

Other recovery, e.g. energy recovery Prevention

Preparing for re-use Recycling

Disposal

(28)

states that energy recovering from waste is equal to energy recovering from any material recycling.

There are several different ways to extract the energy from waste through thermal treatment, different techniques have developed over time. Renhållningsverksföreningen (2005) has given an example of how a CHP plant can work today. The example is illustrated based on a CHP plant in Malmö,

see Figure 44_{. The process begins with trucks that collect the waste and dump}

it in a large container (1). A crane moves the waste from the container to the hopper (2). The waste slides through the funnel into the oven (3). In the oven the waste incinerates as it moves down while it is supplied with combustion air. The temperature in the furnace is about 1000°C, which means all combustible material will incinerate. The hot flue gases rise and continue into the boiler.

The slag from the combustion falls down a trough filled with water and cools down (4). The slag will be sorted and recycled.

Inside the pan (5) there are tubes where the water flows and is being heated up by the hot flue gases. The water is heated into steam and continues to a purifying step. From the high temperature and high pressure, both electricity and heat can be produced. The superheated steam is passed through the turbine (6) which is powering the generator. In the generator (7) the turbine’s rotation is converted into electricity. The steam that has passed through the turbine still holds a lot of energy (heat) that will be used as a district- heating network. For this, a large heat exchanger is used (8).

4_{The figure is taken from Sysav 2003, which is referred to in Renhållningsverksföreningen 2005.} Permission to publish the figure is given to the authors of the thesis from Sysav.

46

Undersökningar visar att med den tillämpade deponeringen av rökgasreningsrester från avfallsförbränning minimeras utlakningen. Därmed förhindras spridning av tungmetaller och organiska föroreningar. När det gäller t.ex. dioxiner är dessa hårt bundna till partiklar. Under förutsättning att rester som innehåller dioxiner inte blandas med hushållsavfall och andra avfall är risken för utlakning mycket liten. För mer information om dioxiner se RVF Rapport 01:13. Förbränning av avfall. En kunskapssammanställning om dioxiner.

Arbetet inriktas nu mot att ytterligare öka säkerheten vid deponering och minska utlakningen av farliga ämnen. Karbonat/koldioxidstabilisering är en metod som är under utveckling. Andra alternativ, som används i bl.a. Tyskland och Norge, är slutförvaring i salt- och kalkgruvor.

4.7 Ett modernt kraftvärmeverk baserat på avfall som bränsle

SYSAV:s kraftvärmeverk i Malmö

För att ge en så konkret och verklighetsnära bild som möjligt av dagsläget inom

avfallsförbränning med energiproduktion illustreras uppbyggnaden av och funktionen hos Sysav:s nya linje för kraftvärmeproduktion i Malmö. Denna redovisning baseras på en information som företaget presenterat i broschyrform under 2003. Anläggningen är ett kraftvärmeverk som med 200 000 ton avfall som bränsle årligen producerar 540 000 MWh fjärrvärme och 145 000 MWh el.

Avfall blir energi

Processen startar med att sopbilarna som samlat in det brännbara avfallet tippar detta i en stor bunker (1), se figur 3. Med en gripskopa som kan lyfta fem ton tar traversföraren sedan upp avfallet och släpper ner det i påfyllningstratten (2). Avfallet glider sedan ner genom tratten och in i ugnen. Avfallet brinner medan det rör sig nedåt på en rörlig bädd, en s.k. roster (3) samtidigt som förbränningsluft tillförs. Temperaturen i eldstaden är ca. 1000 grader, vilket gör att allt brännbart brinner upp. De heta rökgaserna stiger uppåt och fortsätter in i pannan. Det som blir kvar efter förbränningen är slagg som faller ner i ett vattenfyllt tråg och kyls ner (4). Slaggen transporteras sedan bort för att sorteras och återvinnas.

Figur 3. Principskiss över energiproduktionen i Sysav:s kraftvärmeverk

(29)

4. Methodology

This chapter starts with an explanation of the authors’ choice of decision support models. It continues with a philosophical standpoint, which lays the foundation for the research methods. It also gives the reader a deeper insight into the empirical study concerning the research strategies, the data gathering and the collection of alternatives and criteria for the AHP and PMA model. Lastly, the validity and reliability are described and criticized.

Business Research: Philosophy, Approach & Design 4.1.

Research is a systematic way for people to find out things based on logical relationships and not only beliefs. Furthermore, the research needs to have one or several clear purposes (Saunders et al. 2009). Saunders et al. (2009) mean that when using different techniques to collect and analyze data in research, it is the philosophical choices which lay the foundation for how the research will be conducted. Further, Saunders et al. (2009) say that it is important to be aware of the philosophical commitment through the choice of research strategy. When the research philosophy is adopted, it gives an understanding of which direction the research approach is running. Saunders et al. (2009) distinguish between the deductive and inductive approach used in research. The first one aims to develop a theory and hypothesis and to design a strategy to test the hypothesis. In the latter one, the data is collected and the theory is developed as a result of the data analysis. After the philosophy and the approach comes the research design, which is a plan on how to answer the research questions.

This thesis consists of a positivism philosophy, which means that the standpoint is followed by the natural scientist and the research is undertaken in a value-free way. Further, the authors’ emphasize empirical data and scientific methods. The thesis theory is reflected through a deductive approach and the design holds an exploratory research where the problem has not been clearly expressed and there are opportunities to find new insight (Saunders et al. 2009). According to Saunders et al. (2009), using multiple methods could provide better opportunities to answer the research question or questions. The purpose of this thesis consists of a quantitative research with influences from experimental design.

Literature study 4.2.

(30)

know all the theories that are being studied are relevant for the chosen study (Yin 2007). In this thesis, it became vital to study theories about decision making, decision-making processes, AHP and PMA in order to conduct the study. To find relevant scientific literature, scientific databases have been used: Scopus, Science direct and Springer link.

The keywords used were: Multi-criteria decision making, decision-making steps,

decision-making process, evaluation methods, Analytical Hierarchy Process, Pugh Matrix Analysis.

Research process steps 4.3.

RQ1 and RQ 2 are investigated by testing the models and analyzing, discussing and compare the outcome of AHP and PMA. This means that the empirical findings will be the result from the testing. To give the reader a clearer picture of how the research has been structured, the process steps are illustrated in Figure 5.

Choice of models and the real situation 4.4.

Choosing which models to compare the outcome from was the starting point of this thesis. There are several tools and models to choose from when it

Choice of models and real situation

Model testing and comparison

Collection of alternatives, criteria and weighting factors to use when testing the models

Further development of two of the highest scored alternatives – see chapter 6

(31)

comes to idea evaluation such as: SWOT analysis, checklists and different types of matrix analyses (Rebernik & Bradač 2008). To enable to test the models, the authors decided that the models should be of the same type and therefore, two matrix models were chosen. The reason for choosing matrix analysis was because of its strength in evaluating several alternatives and criteria (Renzi et al. 2013). The Analytical Hierarchy and Pugh Matrix Analysis are two commonly used models able to handle the same alternatives, criteria and weighting factors (Cervone 2009; Renzi et al. 2013; Rezaei 2015). Another reason for choosing those models was because of their differences in complexity. PMA is a simpler model compared to AHP, which is more complex (Renzi et al. 2013). However, they have a similar structure, but the workload when using them differs.

Since the models need data to be compared, it was found favorable to test them with support of a real situation. Since the authors found out that Åmotfors Energi is facing a unique MCDM situation, it became challenging and interesting to cooperate. The MCDM situation includes economic, environmental and social aspects which influenced the authors to find a sustainable alternative.

Method used to collect alternatives and criteria 4.5.

Before being able to test the AHP and PMA model, alternatives needed to be collected and selected for the trial. This part describes the pre-work done in to generate a beneficial alternative for Åmotfors Energi to make use of their 30 GWh excess heat.

The first step in a MCDM situation is to define a set of alternatives and criteria (Triantaphyllou et al. 2000). This has been done in several steps, which is illustrated in Figure 6. In the first step, in the idea generation, the intention was to generate all sorts of possible ideas that Åmotfors Energi could do with their excess heat; this is what Linde (1977) describes as the fantasy part, see Appendix I. In the idea generation phase, alternatives were collected from the

experts5_{and people with limited technical knowledge. The reason for}

combining different technical knowledge in the idea generation was to get a variety of ideas. It common that people with high technical knowledge produce ideas technically feasible but less original and incremental. However, people with less technical knowledge are more likely to produce radical ideas,

(32)

ideas that may not be technically possible but more original and have a higher user value (Magnusson 2009).

Collection of alternatives from secondary sources

4.5.1.

The people with technical knowledge from which alternatives were collected in the idea generation process were: the authors of this thesis, engineer students in their second year at Karlstad University and a group put together by Åmotfors Energi comprised of business developers, consultants, people from Karlstad University and former politicians who all together had committed the mission to generate ideas for Åmotfors Energi. The group was named “the Thinking Tank” and consisted of twelve people. The participants in this group were given a brief introduction about Åmotfors Energi’s situation, then the process was divided into two steps: the creative process and the

workshop process. The group was split into two in the creative process. The

mission was to provide concrete suggestions for possible activities that the Idea generation

Brainstorming Alternatives from experts Alternatives from students

Idea selection Rough filtering PMA AHP Idea development Further development of the selected ideas

(33)

excess heat from Åmotfors Energi could be used for in the near future. This process continued for three months. Participants had regular sessions with concrete tasks to develop realistic industries. The process was documented and the participants had to deliver reports. The workshop process was based on the reports that were produced during the creative process. However, the workshop process consisted of developing the alternatives produced in the creative process a bit more. Further, the participants developed a communication plan, which was prioritized with a specific geographic area. Finally, the process resulted in a list with different alternatives for what could be possible solutions for Åmotfors Energi, see Appendix II.

The alternatives selected from engineer students in their second year at Karlstad University were accomplished in groups of 4-5 students. Their task was to present a solution for Åmotfors Energi to use the company’s excess heat. The project resulted in a four-sided report for each group where they developed an alternative together with a simple dimensioning. A total of fourteen different alternatives were developed, see Appendix III.

Brainstorming sessions

4.5.2.

(34)

Rough filtering

4.5.3.

After the idea generation process, a total of fifty alternatives were collected. These alternatives were divided into five different groups by the authors: Green

Nutrition, Wood and Timber, Food Production, Tourism, Drying Processes and Others.

In order to evaluate which alternatives were worth investigating further, a rough filtering was completed. Thus, the fifty alternatives were divided into categories of what Åmotfors Energi wants to do, can do and what they should do. This way of categorizing was described by Anders Linde, see Appendix I, in the 1970s and it is a way for companies to gain understanding about different alternative actions they can be faced with in their strategic planning (Linde 1977). The authors of this thesis ought to call it “Linde’s model”. To structure the alternatives into these categories, a deeper insight in the company’s goals and vision were needed to find out. A questionnaire was sent by email to Åmotfors CEO, Jan Hallgren, including a description of Linde’s model, see Appendix VI.

Method used to compare the outcome of AHP and PMA 4.6.

In this section the methods and strategies used to compare the outcome of the AHP and PMA model are presented.

Experimental design

4.6.1.

Experiments are used in all different kinds of areas by investigators to discover something about a process or a system (Montgomery 1997). The design is valuable in the engineering field when it comes to improving the performance of a manufacturing process. Montgomery (1997) states that the use of experimental design early in the process can lead to improved process yields, reduced development time, overall costs and variability. Further, experimental design plays an important role in the area of engineering design where new products are developed and existing ones improved. Montgomery (1997) also mentions evaluation and comparison of design products as one application of engineering design. Perry et al. (2007) explains how there exits several first-order and second-first-order experimental designs for the investigators and in first-order to select the suitable experimental design, the problem has to be identified thoroughly together with its constraints.

(35)

aspects are to be carried out, which is done by identifying the independent and dependent variables. Further, Kirk (2003) explains the principle replication where an investigation is made of two experimental units under the same conditions. The two units in this study are the AHP and PMA model. Moreover, the chosen criteria, the produced alternatives and the experts testing the models are the dependent variables held constant. The differences in configurations between the models are seen as the independent variables.

Data gathering

4.6.2.

The collection of data is evidence to answer or prove the research question or questions (Rowley 2002). There are two ways to gather data: from primary or secondary sources. Primary data is information that has not been through any analysis before (Dahmström 2005). Qualitative secondary data entails already produced data to be used again in order to develop new methodological and/or social scientific understandings (Irwin 2013). Secondary data is used and carefully evaluated throughout this thesis. The data is gathered from Åmotfors Energi in terms of statistics, graphs and business information. Further, information and facts about the business area decision making, as well as information about AHP and PMA, are gathered from trusted sources. To collect primary data to answer RQ 1 and RQ 2, the two models were tested. Three experts with different work experience were designated by the authors of the thesis to score and fill in the models. The three experts all had sufficient knowledge in the area of heat and energy and they were familiar with the excess heat at Åmotfors Energi. The group of experts consisted of an operation technician working at Åmotfors Energi, an associate professor of energy, environmental, and construction technology at Karlstad University and

an energy and sustainability consultant at Sweco. According to Sukhov6_{, it is}

necessary to let experts with technical knowledge evaluate ideas if the alternatives are technical. The experts were doing the evaluation separately without any influence on each other. They started with PMA and received the matrix as an excel file together with a simple explanation of each alternative and one explanation of how the models worked and where they were supposed to be filled in. They were told they could ask questions and take what help they felt necessary to make a reliable evaluation. After the results from the PMA were gathered, the experts were given the AHP matrix, the same explanation of the alternatives, and an explanation of how the AHP

(36)

model worked. After receiving the results from the AHP from all three experts, the data could be compiled.

Trustworthiness: Reliability & Validity 4.7.

In order to gain trustworthiness, reliability is one important parameter and refers to which extent the thesis fulfills consistent findings in data collection techniques and procedures (Saunders et al. 2009). Therefore, the data collected in the thesis have been carefully selected meaning it comes from original sources. Furthermore, several people have previously cited those sources. Another aspect of reliability is the selection of the participants conducted to run the AHP and PMA model. The people were chosen based on their expertise in the area of heat and energy and also because of their different work experience. This may have resulted in spread empirical findings which minimized bias. Furthermore, having three different people testing the models, three different results can be analyzed which made it more reliable than if only one would have done the test.

(37)

achievable. Even if the generalizability is noticeably low, the authors find the systematic approach of how to deal with a MCDM situation useful for companies in the context of handling criteria and gaining and filtering alternatives.

(38)

5. Empirical results RQ 1 and RQ 2

This chapter gives the reader an understanding about the upcoming discussion and conclusion of RQ 1 and RQ 2. It gives an explanation on how the models were used. It also presents the total result from the experiment.

Result from the idea generation and selection process – 5.1.

alternatives, criteria and weighting factors

The result from the models will be used to answer RQ1 and RQ2. After the process of collecting and selecting ideas, nine alternatives were chosen to be further evaluated in AHP and PMA. These nine alternatives are presented in Table 6. According to Cervone (2009) one reference alternative has to be selected and it has to be one of the alternatives that will be evaluated. Therefore in this research Alternative 1, district heating, was chosen as the reference alternative.

(39)

The criteria chosen for the evaluation is presented in Table 7 together with the weighting factors.

Table 7 -‐ Criteria and weighting factors

Category Criteria Weighting factor Ec on omy C1 Investment cost 1 C2 Profitability 3 C3 Decommissioning cost 1 C4 Economical risks 2 C5 Maintenance 1

C6 Risk for damages that will need reparation 2

C7 Reparation costs 1 E nv ir on m en

t _C₈_{Noise from the business} ₁

C9 Disturbance on the wildlife 1

C10 Impact on the environment 2

Soc

ial C11 How much it favors the countryside 1

C12 Easy to hire staff with right skills to run the

business 2 Phas es and pr oces s step s C13 Subsystems 1 C14 Implementation time 2

C15 If the business have reasonable dimensions

for the amount of heat 2

C16 Difficulties to get the business authorized

by laws 1

C17 Is the alternative reasonable for usage of

the heat 3

(40)

PMA 5.2.

The Pugh matrix was constructed as in Figure 7 where the criteria (Cm) are

shown in Table 7 and the alternatives in Table 6. All three experts conducted one PMA each where they filled in their score for all alternatives relative to Alternative 1 (District heating). The score was either -1, 0 or 1. For example, if they considered the first criterion, which was investment cost, the experts gave

alternative 2 (A2) -1 if they believed that the investment cost for A2 was higher

than A1. Further, the experts gave A2 0 if they believed that the investment

cost was about the same and they gave the score 1 if the investment cost for

A2 was lower then the investment cost for A1. This procedure was done for all

the criteria and all the alternatives. The scoring was always judged in relation

to the reference alternative A1.

𝑊𝑒𝑔ℎ𝑖𝑡 𝐴! 𝐴! 𝐴! 𝐴! 𝐴! 𝐴! 𝐴! 𝐴! 𝐴! 𝐶_! 1 0 𝐶! 3 0 𝐶_! 1 0 𝐶_! 2 0 𝐶! 1 0 𝐶_! 2 0 𝐶_! 1 0 𝐶_! 1 0 𝐶_! 1 0 𝐶_!" 2 0 𝐶!! 1 0 𝐶_!" 2 0 𝐶_!" 1 0 𝐶_!" 2 0 𝐶_!" 2 0 𝐶_!" 1 0 𝐶_!" 3 0

Figure 7 – Construction of the Pugh matrix used in the experiment

Comparing the Outcomes of Two Decision Support Models: The Analytical Hierarchy Process and Pugh Matrix Analysis