Performance in maintenance; Methodological approach to deal with the Performance Killers and Cost Drivers.

(1)

Performance in maintenance; Methodological approach to deal

with the Performance Killers and Cost Drivers.

Master thesis

A Case based research in cooperation with Sitech Services.

Abstract:

Plant availability is essential in providing value within maintenance. Performance killers and cost drivers reduce the potential value that can be achieved. This research has, based on empirical evidence, developed a methodological approach for analyzing the performance killers and cost drivers. A conceptual framework is proposed in which the identification and analysis of performance killers and cost drivers can be classified. The framework is based on the results from the case study at Sitech Services and is supplemented with what is known from literature.

Key works:

Performance Killer; Cost driver; Conceptual Framework; Methodological approach; Plant availability; Logarithmic scatterplot.

University: Rijksuniversiteit Groningen

Study: Technology and Operational Management (TOM) Document: Master thesis

Topic: Performance in maintenance; Methodological approach to deal with the Performance Killers and Cost Drivers.

Author: Lesley Houkema (s2402912) Advisor: dr. ir. W. H. M. Alsem Co-advisor: prof. dr. R. H. Teunter

Date: 6.23.2014

(2)

2 Master Thesis: Performance in maintenance; Methodological approach to deal with the Performance Killers and Cost Drivers.

Preface.

This master thesis has been written as the final project for my Master of Science in Technology and Operations Management. This research has been initiated to find a methodological approach for deriving and analyzing the various potential Performance Killers and Cost Drivers in the field of maintenance. This research has been conducted by the use of a case study in cooperation with Sitech Services.

I would like to thank Sitech Services as a company for providing the opportunity to perform this research. It would not have been possible to perform this research without their cooperation.

I would like to thank everyone at Sitech Services that contributed to this research via interviews and feedback on my thesis. This provided me with the necessary data and information to be able to construct the methodological approach provided in this thesis. I hope that this research has also assisted them in the further optimization processes of their plant availability.

I thank my supervisor for sharing his experience in asset management and maintenance, and the provided feedback throughout this research. This feedback provided me with new insights and ideas on how to perform this research, gather data and stay on track during this research.

(3)

Contents.

List of Acronyms. ... 4 Practical implications. ... 4 1. Introduction. ... 5 2. Research framework. ... 7 2.1 Research questions. ... 7

2.2 Conceptual model: Performance killers and Cost drivers. ... 8

3. Theoretical framework. ... 9

3.1 Value of asset maintenance. ... 9

3.2 Measuring maintenance performance. ... 10

3.3 Performance Killers (PK) and cost drivers (CD). ... 12

4. Methodology. ... 14 5. Case description. ... 15 5.1 Sitech Services. ... 15 5.2 Problem definition. ... 16 6. Analysis. ... 17 6.1 Initial analysis. ... 17

6.2 Case: Current situation. ... 20

6.3 Case: Desired situation. ... 22

6.4 Case: Gap analysis. ... 23

7. Results. ... 25

7.1 Conceptual Framework PK’s and CD’s. ... 25

7.2 Stage 1: Comparability... 28

7.3 Stage 2: Data gathering and analysis. ... 30

7.4 Stage 3: Improving. ... 33

8. Discussion. ... 36

9. Conclusions and further research. ... 39

10. References. ... 40

Appendix. ... 42

1. Interview 1: ... 42

2. Interview 2: ... 45

3. Initial steps to analyze performance killers and cost drivers ... 46

4. Workshop summary ... 47

5. Work order information items ... 49

(4)

List of Acronyms.

BSC = Balance score card CD = Cost driver

KPI = Key performance indicator MTBF = Mean time between failure MTTR = Mean time to repair

PK = Performance killer

PI = Performance indicator RCA = Root cause Analysis SD = Shut down

SHE = Safety, health and environment VDM = Value Driven Maintenance

Practical implications.

This research has shown that it is necessary to rethink some of the current processes performed in maintenance. To be able to compare various PK’s and CD’s across multiple plants, one has to derive the similarities at the different plants in order to determine where they can be compared. Next, it is important to define what the standards should be regarding the targets / goals for performance and costs. This provides the ability to determine if a plant is performing as desired. The data to be gathered should be determined beforehand, according to the analytical approach that is appropriate. By establishing this beforehand, the data gathering process can be tailored towards the requirements for analysis and the validity of the obtained data remains more accurate. Finally, to determine the impact of a potential PK or CD, three dimensions should be analyzed, namely Downtime / Cost for PK’s and CD’s respectively, Frequency of events, and Failure type per event. Improvements should start at the most critical aspects found in these analysis.

Resulting from this concept, a framework has been established in Figure 8. This framework considers the obstacles that have been found within the case and shown in Table 1. This framework takes into account a possibility for reducing the time period for gathering data by combining assets into failure types, providing visual insights into PK’s and CD’s via logarithmic scatter plots, reducing the amount of RCAs initiated to focus more on the successful implementation of findings within the RCA’s, and the effort required to derive the various PK’s and CD’s possible.

(5)

1. Introduction.

Today’s maintenance managers are constantly balancing between higher machine availability (asset utilization) and lower maintenance costs (cost control) (Jonker et al. 2006). Two terms related to these topics, are Performance Killers (PK’s) and Cost Drivers (CD’s) (Stenström et al. 2011, 2013; Kumar et al. 2007), but little has been written in literature regarding PK’s and CD’s (Stenström et al. 2011, 2013; Kumar et al. 2007). Due to the current paradigm shift in maintenance, from a necessary evil towards creating additional value (Parida and Kumar, 2006), the need to prove their value has become more and more important. By being able to determine what the PK’s and CD’s are, what causes them to occur and what can be done to prevent them from occurring, reliability engineers can construct strategies to reduce their impact in order to increase the value derived from maintenance. The need for a methodological approach to derive PK’s and CD’s, and the ability to determine their causes has grown significantly. Thus, this research makes a first attempt at creating a framework that provides a guideline for organizations that want to deal with their PK’s and CD’s.

Recent studies in literature in the field of maintenance focus on the value, where the term value driven maintenance (VDM) has gain popularity among more and more organizations. The continuous pressure to enhance the organizational capabilities to provide value causes organizations to reevaluate processes. This is primarily done through improving the cost effectiveness of their operations (Jonker et al. 2006; Kumar et al. 2013) including maintenance. But, since maintenance is inherited to production (Kumar et al. 2013), the availability of plants, and indirectly profits, will suffer if maintenance is not executed properly. Especially, in industries like the process industry where continuous production and economies of scale are the most important aspects (Fransoo and Rutten 1993; Kallrath 2002), low utilization and high expenditure are not acceptable. The goal of maintenance, therefore, is to provide the maximum possible availability of assets, in order to justify their expenses. Reducing the impact of PK’s and CD’s, to optimize plant availability, aids in optimizing this process.

(6)

This research will be conducted in co-operation with Sitech Services. Sitech Services is a maintenance organization based in Geleen, the Netherlands. They perform maintenance for the plants of i.a. DSM, OCI and Lanxess. Within their new corporate strategy, Sitech Services’ aims at increasing the availability of the multiple plants located at their site even further. In other words, they want to become more efficient by being able to determine the root causes behind their PK’s and CD’s. They believe that there should be similarities in causes spread around the site, which influence the availability throughout all the different plants across the site in a similar way.

(7)

2. Research framework.

This thesis attempts to create a framework to determine PK’s and CD’s via exploring the field of maintenance and through determining the obstacles that can be found within the various PK’s and CD’s that are known both in a practical setting as well as in literature. This section will describe the research approach together with the research questions and conceptual framework regarding the relations of PK’s and CD’s.

2.1 Research questions.

PK’s and CD’s are phenomenon’s within the field of maintenance that negatively influence the value produced in maintenance. In order to minimize their impact or improve the value of performance of maintenance, an organization needs to be able to recognize them. This research will focus on the requirements that should be set in order for an organization to gain the ability to analyze and compare PK´s and CD´s. This research will be conducted in order to answer the following main research question:

What are necessary steps to take in an organization in order to effectively analyze and approach the various performance killers and cost drivers across multiple facilities?

To decompose and structurally approach the various topics in this main research question, various sub-questions are necessary. The following set of sub-questions will be treated within this thesis in order to attempt to answer the main research question.

What are performance killers and cost drivers and how to identify them?

Which data is required in order to analyze the various performance killers and cost drivers?

Which obstacles can an organization find while trying to analyses the performance killers and cost drivers?

(8)

Figure 1; Conceptual model

2.2 Conceptual model: Performance killers and Cost drivers.

Value Driven Maintenance (VDM) focuses on four main value drivers as defined by Jonker & Haarman (2006). These are; (1) Asset Utilization; (2) Resource Allocation; (3) Cost Control; and (4) Safety, Health and Environment (SHE). These four factors combined provide the value within maintenance and have a relation to the performance of maintenance activities. Since this research primarily focuses on the PK’s and CD’s, which are components of asset utilization and cost control, the other value drivers from VDM will be excluded from this research.

(9)

3. Theoretical framework.

The aim of this section is to provide an understanding of the context within this research and describe explore what is known regarding PK’s and CD’s. This section will therefore handle a few different subjects. First, it will discuss the value and importance of correct maintenance. Second, it will explore the different ways to measure the performance of maintenance. And finally, it will continue with a discussion of the concepts of performance killers and cost drivers and the relevant literature around these topics.

3.1 Value of asset maintenance.

As VDM describes, value produced in maintenance is derived from delivering maximum asset availability at a minimal cost (Jonker & Haarman, 2006). The four value drivers (Jonker & Haarman, 2006), (1) Asset Utilization; (2) Resource Allocation; (3) Cost Control; and (4), Safety Health and Environment (HSE), play a central role in increasing the companies value obtained from maintenance. As maintenance is necessary for sustaining asset availability and reliability in order to ensure productivity, product quality, on-time delivery, and a safe working environment (Tran & Yang, 2012), determining the right activities to perform in maintenance is essential in order to provide additional value for an organization.

A common approach is to determine where to perform maintenance can be derived from the various performance indicators (PIs) inherent to maintenance. Kumar et al. (2013) describe a PI as; “a performance measure as a number and a unit of measurement where the number

provides the magnitude and the unit provides the meaning”. Parida et al. (2006) on the other

hand, describe a performance indicator as “a measure capable of generating a quantified

value to indicate the level of performance, taking into account single or multiple aspects”.

(10)

narrow down the amount of PI’s available to only analyze the critical indicators for their operation.

Determining these KPI’s specific to an organization is possible in different ways. One approach suggested in literature is the balance score card (BSC) approach of Tsang (1998). This approach aims at translating organizational visions and strategies into key performance measures or indicators. This approach uses the organizations strategy and translates them into long-run (strategic) objectives of which its attainment is determined by appropriate performance measures with their related target. In this way, a company is able to determine the KPI’s specific to them, which enables them to determine the value of maintenance. Another approach is taken by Chamorro et. al. (2003). They developed a framework, which starts by mapping the business processes within an organization and looking at the knowledge management system within the organization. Though, they also focus on the translation from strategic level towards operations level. Interestingly to note, is that while literature proposes common lists of KPIs, it lacks an agreed-upon methodological approach to selecting or deriving them (Kumar et al. 2013).

3.2 Measuring maintenance performance.

Maintenance performance is, most commonly, measured by cost of maintenance (Idhammer, 2011; Parida and Kumar, 2006). But other methods are available as well. For example, Idhammer (2011) state that costs of maintenance is an outcome of the degree of reliability. High reliability results in low maintenance costs. He therefore proposes to use reliability as a measurement of maintenance performance. Parida and Kumar (2006) on the other hand state that the efficiency and effectiveness of the maintenance system play a pivotal role in the organization’s success and survivability, and therefore, the system’s performance needs to be measured using a performance measurement technique (Parida and Kumar, 2006). For this research, the latter technique deems to be the most appropriate one.

(11)

Lagging indicators, on the other hand are; “measures of the outputs of a process, giving indication of events that have already taken place” (Tsang, 1998; Stenström et al. 2013; Idhammer 2011; Kumar et al. 2013). This classification created between the various PI’s or KPI’s according to what one would like to analyze of determine. Stenström et al. 2013 have developed an overview of the indicators involved within the process as can be seen in Figure 2. Note that, Performance Drivers as well as the PK’s and CD’s are positioned within the

leading indicators and require a leading indicators in order to prevent them from occurring. But to do so, one first has to determine, from recorded data, what the potential treats for performance and costs are, by analyzing the lagging indicators as this will indicate what the known PK’s and CD’s within the system are.

Muchiri et al. (2011) also applied the concept of leading and lagging indicators in an attempt to create a framework to identify the correct maintenance PI’s. Focusing on the leading indicators, Muchiri et al. (2011) created four elements in which the different KPI’s can be categorized. These are; (1) Work identification; (2) Work Planning; (3) Work Scheduling; (4) Work Execution. Keeping track of the KPI’s in these four elements, as they all influence the maintenance performance, allows an organization to identify sources of potential threat to the performance or costs.

(12)

Next to developing a new set of KPI’s, one can look at the standards created by institutions like the British Standards Institution (EN 15341; 2007, Maintenance key performance indicators) or the Society for Maintenance and Reliability Professionals Best Practice Metrics (Kumar et al. 2013). As the authors point out, there are several benefits while using such standards. For example, some of the benefits include:

Maintenance managers can rely on a single set of predefined indicators supported by a glossary of terms and definitions;

The use of predefined indicators makes it easier to compare maintenance and reliability performance across borders;

When a company wants to construct a set of company indicators or a scorecard, the development process is simplified by access to predefined indicators;

The predefined indicators can be incorporated into various computer maintenance management systems software and reports;

The predefined metrics can be adopted and/or modified to fit a company’s or a branch’s specific requirements; and

The need for discussion and debate on indicator definitions is ended and uncertainties are eliminated.

Other authors, like Stenström et al. (2011); Muchiri et al. (2010), had also recognized that such standards could provide multiple benefits to organizations.

3.3 Performance Killers (PK) and cost drivers (CD).

While performance killers and cost drivers are the central point of this research, only a few authors mention them. Though, in practice the terminology performance killer (PK) and cost driver (CD) are commonly used (Markeset et al. 2005; Stenström et al. 2011; 2013). Stenström et al. (2013) have written a review about these terms in an attempt to provide more attention to these value killers. In doing so, they used the following definitions to indicate what PK’s or CD’s are. Performance Killers are defined as those factors that reduce the

production performance without being big enough of an impact to stop the process and Cost

(13)

the ability to reduce the revenue of an asset. The review from Strenström et al. (2011) has also depicted a small overview of things to consider as PK’s or CD’s. These are:

performance killers in maintenance; Equipment that forms a potential risk with respect to uptime, health, safety and environment; Bottlenecks in capacity, administration and inventory; Incompetence; Lack of proper tools and facilities; Faulty procedures and checklists; Inadequate information and communication flow. cost drivers in maintenance; Unplanned maintenance; Process bottlenecks; Equipment

with high energy requirements; Potential liability issues; Operational and/or maintenance costs; Training costs; Facility costs; Disposal costs.

One drawback within the definitions provided by literature is the limiting factor that is proposed within the definition of PK’s. Especially the requirement that the process cannot be stopped narrows the concept to the point where certain aspects might be missed. If a plant needs to be shut down (SD), then it can no longer perform, meaning that the performance suffers. The more time an SD consumes, the more the performance has to suffer, which should be seen as a PK. Using the definition stated above, this event should be excluded, from the list of PK’s that are possible. This research will therefore adopt a broader definition for PK’s namely, “those factors or events that affect the output of the production process, by

(14)

4. Methodology.

This research provides a case based approach within the company Sitech Services. Sitech Services has interest in a methodological approach to deal with Cost drivers and Performance killers across multiple plants located at their location. Case studies are most commonly applied in settings where exploration, theory building, theory testing or theory extension/refinement is required (Karlsson, 2009). This type of research is applicable in settings where; why, what and how questions are important. Furthermore, this research is explorative focused, thus a case based method has been selected as the most appropriate method. In order to answer the stated research question, a few steps have been taken. As the proposed research questions are sequential of nature, this research started by exploring the first question which has been done via a literature research regarding the PK’s and CD’s. This has served as the foundation during the research on which the further research has been build. The next step has been to gather and analyze data to gain insight in the process and find PK’s and CD’s that are interesting to analyze. Finally, the case has been analyzed to find the elements that are important in order to obtain the ability to identify the critical PK’s and CD’s.

The scope of this research has been limited to one case study in which different plants of interest are analyzed for their PK’s and CD’s. Furthermore, this research has been limited to the asset utilization and cost control factors from the value drivers in maintenance as described by Jonker & Haarman (2006). The other aspects, Quality and SHE (safety, health and environment) are not explored to improve focus and due to the time restrictions within this research.

(15)

5. Case description.

In this research project, a case based research to gather empirical evidence has been chosen as the focal point. In cooperation with Sitech Sevices, it was possible to construct the case described below. This section will focus on providing a description of the case and the outlining of the research scope. This will be done by a description of the company in general which is followed by a problem definition.

5.1 Sitech Services.

Sitech Services is a company active in, among others, the field of maintenance. Their current goals focus around achieving a world class maintenance status, which means that they aim at being one of the leading companies in maintenance. But in order to get there, various obstacles have to be dealt with.

Sitech Services currently performs its maintenance on various plants across their site. Each of these plants is related to the process industry, where one of the key aspects is to keep the plants fully utilized. With current plant utilization targets of 97.5%, this provides various challenges around performing maintenance. For instance, if maintenance has to be performed on an intrusive asset, the production has to be shut down. This decreases the technical availability of the plant, causing it to produce less products, resulting in a loss of potential profit. With their current utilization targets, there is relatively little time available to stop production and perform maintenance, so effective strategies need to be developed to perform all the maintenance in this small timeframe.

Currently, Sitech Services is in the middle of a major organizational change. In this change, one of their focus points builds around gaining additional efficiency and effectiveness (improved performance) and reducing the costs (reduction in cost drivers). This means that they aim at increasing the availability rates of their factories while simultaneously decreasing their expenses.

(16)

possibility to rethink and rework processes in order to gain more efficiency and effectiveness and reduce the expenses around maintenance for their customers.

5.2 Problem definition.

As a contribution to a larger improvement theme within Sitech Services, the various Performance Killers and Cost drivers that could be found around their plants are explored within this research. Sitech is interested in finding similarities within PK´s and CD´s across multiple plants to derive a common strategy to deal with them. Especially when it comes to the analyses on the various reasons why these aspects exist and how they can be prevented. Their aim for this research is to find an effective and efficient approach to identify the various PK’s and CD’s that exist across multiple factories so they can develop an approach to deal with them and reduce their impact on the performance and cost. Their focus therefore, is mainly on the various tools and techniques that are available in order to analyze them and what is required in terms of data before these can be applied.

In the analysis of the initial starting situation and by further interviews with experts, the obstacles encountered that hinder the effective analysis of the various PK’s and CD’s have been identified. An overview of these obstacles can be found in Table 1.

Obstacles in the analysis of PK’s and CD’s Work is labor intensive.

Work is inconsistent.

Requires a lot of transformation in data.

The PK’s and CD’s are not comparable across multiple sites. Too many different paths to reach a desired conclusion.

Performance measurements aren’t uniform or comparable. Root cause analyses (RCA) are not efficient enough. Underlying reasons for a PK or CD are hard to trace back.

No clear definition on when certain costs or downtime are tolerated. Table 1; Obstacles in the effective analysis of PK's and CD's

(17)

6. Analysis.

This section focuses on the further analysis of the case described in section 5. The analysis is done according to an initial analysis of the data provided by the firm in chapter 6.1, followed by an exploration of the obstacles found. A comparison between the current and the desired situation will be established in chapter 6.2 and chapter 6.3 respectively. Then, a gap analysis in chapter 6.4 will describe what an organization should improve on to effectively analyzing the various PK’s and CD’s. With the findings in the Gap analysis and the established elements a step is taken towards a generalized methodology.

6.1 Initial analysis.

In an attempt to compare different PK’s and CD’s, graphical representations have been made from the data provided by the firm regarding costs and downtime. Graphs, as displayed in Figure 3, have been constructed form data of the different plants in order to determine where abnormalities within the data occur, which should indicated that either performance or costs has suffered from undesired events. Events in this scenario are the registered maintenance activities that have been performed according to the order numbers. Note that, one order number can have multiple activities, but is seen as one single event. Figure 3 shows an overview of the accumulated costs per month in a time frame from January 2010 to August 2013 for one of the plants on which the firm performs maintenance. Interesting here is the

(18)

costs displayed in this graph are very turbulent and differentiate year by year. An extreme point here can be found in the seventh period of 2012. This indicates that, in this example, a noticeable rise in costs has occurred.

This extreme point, obviously, is not in line with the rest of the costs for this plant which could indicate that there are certain cost drivers in this month that cause these costs to be so extreme. Closer examination is done by comparing the number of events that occurred within this time frame to the costs. Figure 4 shows the same costs plotted in Figure 3 but compares them to the number of events within the month. This figure shows that even though the number of events within a single month remains high, the average costs per event do not drastically increase. The accumulated costs in months where the number of events is around 400 is about €400,000.- whereas the costs around 650 events lay around €600,000.-. But, as is evident from this graph, the seventh month of 2012 still has a higher average and can be seen as an abnormality.

Further investigation of a plant is than achieved through analyzing the data according to asset allocations (functional placement and location of an asset) in an attempt to point at the asset which causes these costs to be deviate. But, the data registration for such events is not accurately specified within the plants. This poses difficulties in the further analysis of these

(19)

abnormalities within costs. Figure 5 shows the costs divided among different functional locations within the plant, where most of the costing aspects have been formulated under UF2, ME2, and ME4. UF2, ME2, and ME4 are terms that are used as general indicators within a plant and do not indicate more than that e.g. electrical wiring or pipelines in total that have produced these amount of costs. Further explorations of data in depth were not possible within the current data set. A different approach should be taken in order to determine which data is required to analyze the PK’s and CD’s especially when trying to determine causes for events. This also means that it was not possible to determine causes associated with these abnormalities in costs.

The initial analysis has shown that, with the current data and analyzing techniques, it is not yet possible to determine the causes for, in this example, CD’s. The same thing applies to the PK’s in which case the dimension of costing is replaced with downtime. There is a need to determine what needs to be accomplished before it can be attempted to derive effective measures against PK’s and CD’s. It is not possible to simply compare the various PK’s and CD’s and try to find relations between them. There is a need for additional knowledge around PK’s and CD’s that involve e.g. causes, targets or standards before this can be accomplished. In order to do so, a current situation, desired situation and a gap analysis will be performed.

(20)

6.2 Case: Current situation.

Various obstacles have hindered deriving the PK’s and CD’s, because e.g. the accessibility and especially the causes of PK’s and CD’s do not exist or are difficult and unknown. To be able to map the elements that require further exploration within Sitech Services, the case will be analyzed according to the model in Figure 6.

The various obstacles shown in Table 1 provided an overview of where the difficulties within the analysis of PK’s and CD’s can be found. Based on the concept of Figure 6, and the obstacles in Table 1, a format for interviews (see Appendix 1) has been developed to sketch the current and desired situation. The findings from these interviews for the current situation can be categorized in three common causes for difficulties in analyzing PK’s and CD’s, namely;

Differences in managements of the different plants.

Reliability of the analysis is too inconsistent and not sufficient.

Reconfigurations in key performance measurements system (KPI system).

As would seem obvious, these three aspects pose there limitations on the comparing, identification process and analysis of PK’s and CD’s. Each of them increases the amount of labor required in order to derive them.

6.2.1 Differences in management.

In order to build a reliable system to compare the various PK’s and CD’s across multiple sites with one another, there is little room for differences in management. Differences in management cause the operational process and therefore also the data gathering processes to be less comparable. For example, consider the data registrations process within plants. If the management registers data regarding the labor performed and material used, in one plant as events where labor and materials are combined, but in another plant they separate the labor from the materials, the recorded data per event will be different. Another obstacle caused by differences in management is the approach taken in maintaining a plant. Similar failures might occur within different plants, but due to possible differences in management, they are

6.2; Current situation 6.3; Desired situation 6.4; Gap analysis

(21)

solved differently. This causes the materials and labor used, to differentiate. Thus, differences in management disrupt the validity of the data which is gathered and ultimately the validity of analysis.

6.2.2 Reliability of analysis.

Besides validity, the differences in management of the plants pose their limitations on the reliability of analysis. Using unreliable data to analyze the PK’s or CD’s can lead to improper decisions which will not yield the desired solutions. But, in the current situation there are certain facts to note. First, most of the data collection is done by the operators / technicians that work within the plants as they are the first to observe failures or identify potential risks within the plant. Though, most of these employees are not aware of the reason why certain data is gathered. Furthermore, operators are not really interested in gathering data. Combined with the complexity of the current data management system, inaccuracy in data is unavoidable. Second, different interpretations of the options are possible within the current system. Multiple options to report a failure or defect currently exist and cause more inaccuracy in terms of data reliability. If common or similar errors are reported under different names, one could easily miss out on the severity of that error when analyzing the available data. Also see example in chapter 6.2.1. Finally, in depth analysis are not easily performed. Difficulties arise around tracing back where data originated and in determining the cause for a certain cost figure. Furthermore, the standard set for costs associated with certain cost figures remains vague. For example, a registered event might show that the replacement of a bearing has cost €5,000.-. This might seem expensive, but since there is no target costs given for replacing a bearing or registered events around the replacement of the bearing, it is impossible to determine if these costs are normal or not, and if they can be prevented the next time this bearing needs replacement.

6.2.3 Reconfigurations of performance measurements.

(22)

CD’s. Though, analyses to find causes for PK’s or CD’s or similarities in causes, will remain difficult even after establishing these alterations. Further developments within the process of gathering data, related to performance measurements, and the analytical approach are still required.

6.3 Case: Desired situation.

PK’s and CD’s are phenomenon’s which should be dealt with in order to progress towards increased efficiency and effectiveness in an organization. In the desired situation the aim will therefore be to acquiring the ability to efficiently deal with various PK’s and CD’s across the site as to reduce the amount of downtime and costs within the different plants. As pointed out in Table 1, the obstacles have already been identified e.g. work is to labor intensive, inconsistency in data, root cause analyses are not effective, etc. In order to reach the desired situation the focus should be on methods that aid in solving these obstacles. The interviews, appendix 1 and 2, that have been conducted provided an overview of the abilities the organization wants to gain. The findings have been combined in three desired abilities the organizations wants to achieve, which will be described hereafter. These are:

- The ability to simply and directly create an overview of the various PK’s and CD’s

across multiple sites.

- Knowledge about the PK’s or CD’s should be easily traceable.

- Ability to draft up and execute effective measures to deal with PK’s and CD’s.

6.3.1 Overview PK’s and CD’s.

(23)

6.3.2 Knowledge about PK’s and CD’s.

Simply providing an overview of the various PK’s and CD’s will not enable an organization to reduce their impact. Especially if the causes for these events occurring are unknown. Therefore, the degree of knowledge that is required should be specified together with the analytical approach. If one defines how to analyze, the requirements for that analytical approach should become evident and the data gathering process can be altered to accommodate this. More importantly though, is to determine the purpose of an analyses. As there are different root cause analysis techniques available to demonstrate where or what caused a certain event to happen, the data still has to be gathered. Knowing the potential benefits of the analysis you want to perform, aids in convincing the importance of proper data gathering. In some situations, though, it might be unclear what the analysis should be. For such cases, one can look at standardized measures. Parida and Kumar (2009) have constructed such a list of measures that are useful to generate understanding of the maintenance process and for the analysis of maintenance data (see appendix 6). Though, these measures are general and of high level, they do provide an overview of the status at of plant. Furthermore, most of the current and commonly applied analytical tools, such as the Overall Equipment Effectiveness (OEE), use these measures as inputs.

6.3.3 Improvement measures.

Finally, only determining what the PK’s or CD’s are or what causes them, does not generate value for a company. Effective measures to limit their impact are a final step in achieving a more optimized availability within a plant. Thus, after knowing where and why, you want to determine what to do. This could be done in various ways, but should always be tailored to the situation at hand. For example, a possible way to deal with a found PK could be to adjust the maintenance approach from corrective maintenance to preventive or condition based maintenance so to prevent a fraction of the downtime due to this phenomenon. This way the impact of the PK should be reduced and the result will be an improvement in availability.

6.4 Case: Gap analysis.

(24)

These difficulties arise around targets to compare performance or costs against, the comparability of the various plants, the validity of the analysis that are performed, and the analytical approach according to which the data is analyzed. Furthermore, there is a circular principal present in the process of analyzing PK’s and CD’s. In order to determine which data should be gathered, the potential benefits of the data should be made clear. But, in order to show the potential benefits that can be achieved with the analytical approach, the required data should first be gathered or available. To map the problem of where to start and to get a more linear approach, the different theoretical steps that are required for analyzing PK’s and CD’s have been developed in appendix 3. This has been done based on the priority within the possible steps, the obstacles identified throughout the research and in the interviews with experts, starting from the desired situation working back to the current situation.

These theoretical steps have be generalized and categorized in 5 different elements, namely; (1) Organizational processes; (2) Creating targets; (3) Data management; (4) Performance measurements; and (5) Improvement plan.

To further analyze the gaps that arise within this process, and in order to establish possible sub components that require development, a workshop has been organized where these elements have been presented. A summary of this workshop can be found in appendix 4. An overview of the resulting elements and their sub components is provided in Figure 7.

The elements and sub components displayed in Figure 7 provide the basis for the development of a methodological approach to obtain the ability to identifying the similarities in causes for PK’s and CD’s across multiple plants, and will be further developed and explained in chapter 7. - Management - Operations - Knowledge - Comparability - Realistically - KPI’s - Knowledge Requirements - Validity - Collection - Motivation - Training - System - Tools - Techniques - Measurements Detail - Performance Killers - Cost Drivers - Improvement Process Organizational processes Creating targets Data management Performance measurements Improvement plan

(25)

7. Results.

The different obstacles an organization encounters in analyzing PK’s and CD’s have been determined according to the case analysis and the interviews with experts. This has resulted in 5 different elements and a number of components within the elements that are necessary to construct a methodological approach to deal with PK’s and CD’s. This chapter will focus on the development of a conceptual framework that provides organizations with such a methodological approach after which the different elements and stages within the framework will be explained and supplemented with existing literature. Finally, the tools and root case analysis techniques applicable for analyzing PK’s and CD’s will be discussed.

7.1 Conceptual Framework PK’s and CD’s.

The analysis in chapter 6 has shown, that the difficulties in analyzing PK’s and CD’s arise around targets to compare performance or costs against, the comparability of the various plants, the validity of the analysis that are performed, and the analytical approach according to which the data is analyzed. Furthermore five different elements and related subcomponents have been found in Figure 7 that should be incorporated within a framework. As the framework to develop aims at improving the current situation and work on reducing the obstacles found in Table 1 the goal within each element should be clear. Currently, the descriptions of the five elements is to general and can contain a wide variety of additional components that are not important within this approach. In order to provide better focus on what needs to be accomplished and the context in each element, the names of the elements have been altered to accommodate their context and goal. See Table 2.

Element Name Goal

1 Alignment of the organizational processes Standardization & Comparability 2 Defining performance and cost targets Standardization & Comparability 3 Data management and system Increase validity in data

4 Performance measurement and analysis Identify PK’s and CD’s + causes

5 Optimization process Improvement approach PK’s/CD’s

Table 2; Overview of elements altered to their context and their goals.

(26)

determine which elements relate to one another. And finally, different stages have been introduced to separate the elements and their sub components to allow for checkpoints in between the different elements. The resulting framework is shown in Figure 8.

Figure 8; Conceptual Framework for analyzing PK's and CD's

(27)

Stage 1 aims at creating comparability within different plants by using the elements “alignment of organizational process” and “defining the performance and cost targets”. Fundamental to the entire question of comparing various PK’s and CD’s is the question if you can actually compare the assets. For example, two similar pumps both use the same kind of valves, but in one case it is either always open or closed, where in the other is opened for 80% to 90%. Would it make sense to compare these two valves if they are used differently and are exposed to different conditions? These two elements, therefore, also received the highest priority in the framework. To achieve this comparability, the sub components of each element are to be used. For alignment of organizational process the subcomponents to consider are (1.1) Management, (1.2) Operations, (1.3) Standardization, (1.4) Benchmarking and (1.5) Knowledge. Defining performance and costs targets has (2.1) Information requirements, (2.2) KPI’s and (2.3) Targets, as its subcomponents. Further explanation of these subcomponents will be provided chapter 7.2. At the end of this stage, the degree of comparability should be checked, and if one concluded with enough confidence that comparability has been achieved, the next stage can be initiated. If not, these elements have to be reconsidered to gain additional information regarding the comparability of plants.

(28)

The final stage, stage 3 of the framework, aims at improving and optimizing the failure types for both PK’s and CD’s. After determining what, where and why PK’s and CD’s form, an optimization process can be initiated. Strategies can be developed to either reduce their impact, or prevent them from occurring and effort can be put into implementing the desired changes. The sub components within the element “Optimization process” are (5.1) PK’s and CD’s, (5.2) Improvement approach, and (5.3) Monitoring. Further explanation of these subcomponents will be provided chapter 7.4.

7.2 Stage 1: Comparability.

As has been determined in chapter 6.1, PK’s and CD’s cannot simply be compared to one another. The variations due to differences in the organizational process can already be a cause for differences in performance or costs. By working on the sub components within this stage, the current situation can be modified to prevent the results from later analysis to be invalid. The components of (1.1) management and (1.2) operations play a pivotal role in this process. As Bendoly & Cotteleer (2008) state that in every operational process where there are a number of individuals responsible for the process, variation is inevitable. Decisions made by management, in a situation where each plant is managed autonomously, will therefore cause the operational procedures to be different. As would seem evident, the sub component (1.3) standardization is a key component to create a comparable situation. Following the same set of procedures and rules eliminates a source of variability that can exist within the data gathering process later on in the model. Especially when one attempts to make multiple plants comparable, it is important that the same formal approach or methodology in the data gathering process is taken and is based on a comparable situation.

(29)

but unknown in another. The process of obtaining knowledge should be shared across the multiple plants to work on the development of standardized operational processes.

After the organizational processes are explored for comparability, one can start defining performance and costs targets for the different plants. While doing so, one first needs to determine what needs to be known. This can be achieved through looking at Information requirements (2.1). Out of all the possible data that can be gathered, not everything will be as useful. Furthermore, gathering all the possible data can be very time consuming, especially on employees who are not hired for this purpose. By determining which data is necessary for the analytical approach, the list of possible data can already be reduced significantly. Furthermore, standards have already been created which include useful data in the field of maintenance. For example, the EN 13460: 2002 standard, Appendix 5, shows a list of common data that is useful in the field of maintenance. Though, not everything will be useful, and it should be tailored to the organization. Knights (2001) states that, experienced maintenance and operating personal are indispensable to this process, since familiarity with the machine, the operating environment and with maintenance and operating practices is required. Thus, through involving operators and technicians in the process on is able to further determine which information is useful and required to analyze PK’s and CD’s.

(30)

hand, are a bit more complicated, as they do not have a clear limit like time per day. In this case, the available budget can be used, which can be divided among the number of events that on average occur. For instance, there is a budget of €10.000,- and 20 events on average within a period of time. This means that each project can spend up to €500,- for it to be considered normal. Deviations above this limit can be considered as CD’s and can then be dealt with. These targets are used later on in the process when identifying the impact of the different failure types that can occur.

To prevent unnecessary changes while trying to achieve comparability in the process, and to keep track of the improvements, one should not aim at changing an entire process radically. An approach to prevent radical changes can be achieved through the use of Lewin’s model of change (Cawsey et. al. 2012). They describe the process of changing processes in 3 steps; Unfreeze, Moving and Freeze. Here unfreeze means that the process is opened for changes. Moving enables changes to be implemented. And freeze closes the entire process to additional changes and should be followed as is. This same principle should be applied to create and maintain cohesion between the various plants.

7.3 Stage 2: Data gathering and analysis.

(31)

is usually not supported by ERP systems (Labib, 2004). The data gathering process should require more attention than the systems, which should be altered towards what is necessary for storing the defined data.

Within the process of gathering data, two more components have been found that are important to increase the validity of data. These are (3.3) Motivation and (3.4) Training. Complexity in systems, or improper understanding about why data should be gathered can lower the motivation of employees to keep registering the data. Naturally, validity of the data than suffers due to missing or inconsistencies in data caused by lack of motivation to gather data. Keeping motivation high and providing the correct training to the collect data, therefore, aids in the process of increasing validity. For example, by providing feedback from the results achieved through an operators data collection, his or her motivation and pride in their work can increases. Doing so increased the validity but also the accuracy in the data.

(32)

different failure types (1, 2, 3, 4, and 5) have been plotted where failure type 3 is the most critical PK or CD as it lays within the “Acute and Chronic” segment.

This principle by Knights (2001) has been applied in other studies as well. Karim et. al. (2006), for example have applied the logarithmic scatterplots in order to find where quality issues come from within the field of construction. By creating failure types for various components, they were able to derive the main issues within their research. Another tool, that can be applied to determine the criticality of failure types, is the one described by Labib (2004). Even though the same principle of plotting frequency versus downtime is applied, different criteria to base decisions on is presented. As can be seen in Figure 10 the downtime (or costs for CD’s) and frequency are divided into a High, Medium and Low labels. Failure types can be divided across the different segments of this table in order to determine what approach or decision should be taken. Note that, again the aim in this figure is to move failure types towards the low frequency and low downtime segment.

The proposed tool by Knights (2001) allows an organization to determine which failure types require the most critical to improve at within the different possibilities. Further RCA techniques (4.2) can be applied hereafter to determine the root causes for a failure type. Examples of such techniques are (1) Failure elimination method, (2) FMEA (failure mode and effect analysis), (3) Tripod analysis, (4) Pro-act, (5) 5 Why, (6) RCA, or (7) Human factor analysis. Though, a common issue with most of these analysis, is that they are initiated to

(33)

easily and the results from the root cause analysis, therefore, are often not fully implemented before another analysis will say that it has to be done differently again (Isenhour, 2014).

By focusing on the most important failure types for PK’s and CD’s first, the amount of e.g. RCA initiated can be reduced which increases the changes for successful implementations. Finally, the amount of detail in the measurement (4.3) should be clear. Failure types can be further explored towards the different assets that have shown that these failure occur, where priority for improvement can be given towards the most impactful asset.

7.4 Stage 3: Improving.

At this stage, improvement plans can be developed to improve / optimize potential PK’s or CD’s that have been found within the analysis. If, the process before this stage has been followed, the impact of a failure type and the root causes for the various Performance Killers and Cost Drivers (5.1) should be known. Following Knights (2001), a list of possible root cause for failures can be generated together with an action plan to deal with each root cause. In their example, they have divided the failure types within 6 categories, namely; (1) inspection, (2) maintenance, (3) operation, (4) design, (5) material and (6) resources. In combination with a system for possible actions to take, for example, improve inspection

(34)

frequency or change component, a clear and visible improvement approach (5.2) can be developed. Improvement should be made according to the RCA that have been performed. Isenhour (2014) has developed a categorization in the different types of root causes where he differentiates between physical, human, systematic and organizational causes. Figure 11 shows an overview of these different causes.

Isenhour (2014) also states that the amount of effort required to improve differs between the categories. Obviously, the first category requires the least amount to identify as well as solve. As the causes for failures arise deeper into the framework of Figure 11, the amount of effort increases significantly. Each category, therefore, requires a unique approach in order to optimize the process. Note that the proposed methodological approach in this research (Figure 8), already starts at the organizational causes within the analysis of PK’s and CD’s.

Effect (failure)

Mechanical failures in equipment are the main cause for failures. In this level it is relatively easy to determine where and why failures occur.

1. Physical Cause

Failures do to human errors are the main cause for failures. As it is more difficult to gather data for failures coming from the human aspect, more effort is required compared to the

physical cause. 2. Human

Cause

System (people and physical) causes are failures coming from errors made due to deviations from the performance standards, work procedures or work processes.

3. Systematic Cause

The lowest level regards the organizational aspect, where a lack of experience or awareness might be the root cause for failures.

4. Organizational Causes

(35)

The final component in this stages, (5.3) monitoring, has been introduced to keep track of the changes that are made, and provide a visual overview of the progress. As important as identifying potential PK’s or CD’s are, without ensuring that improvements have been made and keep being made, the danger of reverting to the old procedures may limit the potential benefit. Using the logarithmic scatter plots again, one can keep track of the improvements by considering failure types over multiple periods. An example for this methods is illustrated in Figure 12, where failure type 1 has been plotted for 3 different periods.

Figure 12; Logarithmic scatter plot, monitoring failure type improvements

(36)

8. Discussion.

This research has been conducted to find a methodological approach to identify the potential PK’s and CD’s, including the causes for them to occur. As a guideline throughout this research, various research questions where developed in chapter 2. So, naturally, these questions need to be answered. The main research question, “what are necessary steps to take

in an organization in order to effectively analyze and approach the various performance killers and cost drivers across multiple facilities”, has been approached through the different

sub questions that will be answered here.

Q1: What are performance killers and cost drivers and how to identify them?

Literature regarding PK’s and CD’s is relatively limited, and no clear approach has been provided on how to identify them. Stenström et al. (2013) has made an attempt to obtain more awareness towards PK’s and CD’s in literature, and provided an review of current literature regarding these topics. Stenström et al. 2011, 2013; Kumar et al. (2007) proposed definitions for both PK’s and CD’s, but have been very limited regarding their definition on PK’s. Their definition of PK’s limits the amount of PK’s that can be found due only including the factors that are not big enough to stop the process. Rather, a more broad definitions is required to include different factors that can limit performance. This research has proposed the following definitions for PK’s. “Performance killers are those factors or events that affect the output of the production process, by either limiting or fully preventing the production of an asset”. Cost drivers on the other hand are “factors that causally affect costs over a given time span”. PK’s and CD’s can be identified through determining their impact in the performance and costs. This can be accomplished by plotting the frequency of events versus the downtime or costs for PK’s or CD’s respectively. Criticality can be assigned according to their deviation from the average downtime or costs allowed.

Q2: Which data is required in order to analyze the various performance killers and cost drivers?

(37)

Furthermore, logarithmic scatter plots can then be created, combined with a set target. This enables one to determine if PK’s or CD’s are “normal”, “acute”, “chronic” or “acute and chronic”. Differentiating between these four impact types, allows one to determine a more accurate maintenance strategy to reduce their impact. Critical though, is to start collecting the data as soon as a failure occurs or maintenance needs to be performed. Valuable facts, like for example temperature, can disappear relatively quickly after an asset stops working, possibly preventing one from determining the real cause of the failure. This is one of the reasons why condition based maintenance can be very useful. By proactively thinking about the necessary data you want to collect to determine where potential risks can occur, data will automatically be gathered to analyze the causes for failures.

Q3: Which obstacles can an organization find while trying to analyses the performance killers and cost drivers?

Various obstacles have been found in this research that hinder an organization from analyzing PK’s and CD’s. An overview of the obstacles found during this research is shown in chapter 5, Table 1. Most of these obstacles have shared causes for them to occur. The difficulties in analyzing PK’s and CD’s arise around targets to compare performance or costs against, the comparability of the various plants, the validity of the analysis that are performed, and the analytical approach according to which the data is analyzed. Pareto analyses, commonly applied within the field of maintenance to determine where performance or costs come from (Knights, 2001), are very limited. They only consider performance or costs per asset, and provide no insight on the impact or cause(s) of PK’s or CD’s.

Q4: Which elements can be used to construct a framework and a methodology to identify the PK’s and CD’s?

(38)

additional problems might occur within this framework. It therefore remains a conceptual framework which needs to be tested in order to validate this approach.

Next to the application of the framework are the desired results one expects to get. As has been discussed previously, when utilization gets closer to the 100%, with targets of e.g. 97,5% utilization, the additional benefits that can be gained from finding PK’s and CD’s in the technical aspects reduce significantly. If one looks at leading industries in maintenance, such as the aircraft industry and the nuclear industry, one will find that technical reliability of assets are almost never the cause of a PK or a CD. Rather, the human component or organizational procedures should be explored in order to determine where, what and why PK’s and CD’s arise. And as Isenhour (2014) has shown in his framework for root cause failures (Figure 11), the further one has to explore the causes for failures, the more effort is required in order to gain additional benefits in optimization. Significant improvements will most likely come from standardization processes and reduction in variability. One has to keep this in mind, as this framework has been developed to find causes within the technical side by using a failure coding system to determine why assets malfunction.

(39)

9. Conclusions and further research.

Cost drivers (CD’s) and Performance killers (PK’s) are not easily comparable to one another. An Pareto analysis is a common tool used in maintenance to find PK’s and CD’s, but does not provide sufficient information to determine their impact and cause (Knights, 2001). This research has been conducted to provide a methodological approach to analyze PK’s and CD’s. The different elements found in this research are important to explore in order to obtain the ability to compare various PK’s and CD’s, to construct a common maintenance approach to deal with them, and in order to make a next step to improve plant availability. These elements have been combined in the form of a framework presented in Figure 8. To be able to determine the causes of PK’s and CD’s the stages in this framework should be followed. First, one should establish a situation where the various plants are comparable by creating standardized work procedures and defined targets to compare against. Next, data should be gathered about the plants following the format presented in appendix 5. This data can then be categorized and analyzed according to failure types which should eliminate the needs for an excessive time frame to collect the data. Criticality in the failure types should be formulated according to the impact of each failure type. Once this is done, further techniques such as RCA, Tripod, FMEA etc. can be applied. Finally, once enough information is gathered about the PK’s and CD’s and their causes are determined, an improvement stage can be initiated to reduce their impact.