The role of new technologies in providing smart home energy solutions

The role of new technologies in providing smart home energy solutions

Using demand response as a tool

to engage customers in the new world of smart energy

Master Thesis

By

Marthe Hanna Posthumus University of Groningen Faculty of Economics and Business

MSc Marketing Intelligence & MSc Strategic Innovation Management Jan 30rd, 2017

1st Supervisor MSc Marketing Intelligence: Maarten Gijsenberg

Management summary

The energy market is facing some major challenges. Decreasing margins and new technologies force energy retailers to change their conventional way of doing business from selling energy as a commodity to providing new smart home energy services. This requires energy providers to search for new ways to integrate new technology based services and transform them into future business profits. Demand response is a development which may offer energy retailers the opportunity to add new value to the customer and realize a competitive advantage. Nonetheless, demand response requires a more active role of the consumer, which is difficult to achieve in a low involvement market such as the energy market. The current energy market is to a great extent designed on changing behavior based on financial incentives and based on economic reasoning. Nonetheless, literature in different contexts suggests, emphasizing the environmental benefits might be more effective in behavioral change, especially when financial incentives are low.

This current study was divided into two parts. The first part of the research aimed at analyzing whether environmental incentives have more effect on behavioral change for new services in comparison with financial incentives. This was examined by conducting a field experiment. The results of the field experiment indicate environmental incentives have more effect on behavioral change over time than financial incentives. This suggests companies may consider using environmental incentives instead of using financial incentives, in engaging customer in new services such as demand response.

The second part of the research aimed at answering the question how firms could use the insights from behavioral change strategies to integrate new value adding services in their business strategies. This may help them to transform from an energy supplier to a service provider. This was examined by conducting 11 depth interviews with diverse players in the energy market. The results from the in-depth interviews, suggest energy retailers should experiment with services such as demand response in the short term, to gain more customer knowledge and engage the customer in the new world of energy, which enables them to establish a more favorable position in a later stage. In this way, firms can engage in organizational learning to improve their services. Moreover, the results of the interviews suggest investing in the customer relationship is key to be able to add new services to the customer. Demand response will require a more active role of the customer, whereas this role is now far from active. For now, firms should focus on establishing a strong relationship with the customer to establish a favorable position in a later stage. In this relationship, it is important to be customer centric. The results also state the service should be tangible. Thus, firms can integrate services by building a stronger relationship with customers, realizing organizational learning, visualizing their services by using tools such as smart thermostats and develop services that respond to the individual needs of the customer.

Keywords: demand response, behavioral change, household energy consumption, servitization, business innovation, service orientation, organizational learning, customer engagement, energy market,

Preface

The master thesis that lies in front of you is the final product of a last semester of hard work towards a master in Marketing Intelligence & Strategic Innovation Management. About 1,5 years ago I started at the Energy Academy and start to learn about how interesting the energy market is. From the moment I was approached by my manager at the Energy Academy Europe, I fell in love with the dynamic and complex nature of the energy industry. So many questions lay to be answered, which you can observe from so many disciplines. Moreover, I could connect this perfectly to my ideological (and even maybe a little bit) naïve mission: improving the world. The last years a certain question often crossed my mind: “How can you motivate people to act sustainable?”. I find this question very intriguing. And from this interest, I started my research. But having this inner drive, does not mean the road is easy. On beforehand the project seemed undoable and many people warned me, I was taking a high risk. And I agree, I took a high risk and it certainly was not the easy way. But, the final product is here, and I am thankful I could choose this path to end my masters with.

What I really learned from this master thesis starts way back in time. I always have been very studious and interested in why things are the way they are. From reading books of Stephen Hawking or Paulo Coelho to even study books of other studies. However, I also wanted to do something with this knowledge and translate it to things that add value to the world. I was a walking encyclopedia of knowledge, but failed in really applying this knowledge to practice in a structured way. In order to gain the skills to conduct research in a professional way, I choose to study MSc Marketing

Intelligence. To translate the insights that can be derived from these questions into sustainable

solutions, I chose to study MSc Strategic Innovation Management. But overall, the eventually the end goal was to get a more abstract and structured way of thinking. Now in this last month, I finally came to realization I reached this higher way of thinking and am able to link constructs to each other in a structured way. The road has been long and unstructured, and I came across myself several times. I learned how to set up an experiment and how to work with programs such as R, Excel and SPSS by heart and discovered functions I did not know they exist. But most of all, I learned more about myself, and to finally trust on my own decisions. And that is something even more valuable for me than learning all these new skills.

I first want to thank my parents for always supporting me unconditionally. Moreover, I also would like to thank Maarten Gijsenberg and David Langley for trusting me on this unordinary project which in the first place (and even at the end) brought some large risks and challenges with it. Both of them helped me structuring the millions of ideas I had in mind, and helped me to reach this end product. Moreover, I would like to thank Erik Boels, for helping me to together conduct this beautiful

experiment and asking me practical business questions, which triggered my thinking constantly. It has been a very interesting journey from starting with a unique opportunity to set up an experiment myself and speaking to interesting people in the energy market, to a tough end period, with days from 9am till 12pm where the biggest challenge was to concur myself. But here it is. The end product of 6 months. I hope you enjoy reading it.

Table of content

1. Introduction ... 1

2. Context ... 3

3. Theoretical framework ... 5

3.1 Theoretical framework part I ... 5

3.1.1 Strategies for behavioral change ... 5

3.1.2 The role of moral incentives in encouraging behavioral change ... 6

3.1.3 Effect of intervention over time ... 7

3.2 Theoretical framework part II ... 8

3.2.1 Value adding services ... 8

3.2.2 Becoming a service provider ... 8

4. Research design and methodology ... 9

4.1 Experiment ... 9

4.1.1 Research design ... 10

4.1.2 Data collection and structure ... 10

4.1.3 Procedure ... 10 4.1.4 Variables ... 11 4.1.5 Model specification ... 13 4.1.6 Plan of analysis ... 14 4.2 Interviews ... 16 4.2.1 Research design ... 16

4.2.2 Data collection and case selection ... 17

4.2.3 Procedure ... 19 4.2.4 Analysis ... 19

5. Results ... 20

5.1 Experiment ... 20 5.1.1 Data cleaning ... 20 5.1.2 Descriptive statistics ... 21

5.1.3 Overall fit and significance model 1 ... 22

5.1.4 Effect intervention in the short and long term ... 23

5.1.5 Validation of model 1 ... 24

5.1.6 Overall significance and fit stage model 2 ... 24

5.1.6.1 Direct effect intervention (model 2a,2b) ... 24

5.1.6.2 Validation direct effect intervention ... 25

5.1.6.3 Long term effect intervention (model 2c,2d) ... 25

5.1.6.4 Validation long term effect intervention ... 26

5.1.6.5 Overview hypotheses ... 26

5.2 Interviews ... 26

5.2.1 Within case analysis ... 27

5.2.2 Cross-case analysis ... 48

5.2.2.1 Overall vision demand response ... 48

5.2.2.3 Shift in business strategy ... 52

5.2.2.4 Changes in the energy market ... 55

6. Discussion ... 56

6.1 Main findings from the experiment and contributions to literature ... 57

6.2 Main findings from the interviews and contributions to literature ... 58

7. Reference list ... 61

8. Appendix ... 68

Appendix 1: Letter experiment “CO2” group ... 68

Appendix 2 mobile version and web version of e-mail ... 69

Appendix 3: Results model 1: estimates intervention ... 70

Appendix 4: Non-normal distribution of the residuals ... 72

Appendix 5: Interview questions ... 72

Appendix 6: Codebook axial codes ... 76

Appendix 7: Overview codes ... 80

1

1. Introduction

At the current moment, the energy market is facing some major challenges. Decreasing margins through increasing competition on low costs and the development of technologies force energy suppliers to change their traditional role as energy supplier to service provider, offering smart home energy solutions. This requires energy providers to search for new ways to integrate new technology based services and transform them into future business profits.

Demand response is a new service development in the energy retail market, which can be used by energy suppliers to diverse their product portfolio. By providing customers the opportunity to shift their electricity consumption to favorable supply conditions, customers can reduce their electricity bill and contribute to the environment by using cleaner energy. Demand response requires households to shift (automatically, manually or both) their current electricity consumption in response to a price signal or other stimuli (Darby & Mc Kennar, 2012) towards favorable supply conditions. Favorable supply conditions are e.g. when power from renewable sources is available or when price is lower during off peak demand periods (Moura et al., 2009; Giordano et al., 2012; Broeer et al., 2014; Toft et al., 2014).

Changing energy consumption for new value adding services such as demand response, requires a more active role for customers in the regulation of their energy consumption (He et al.,2013; Geelen et al., 2013; Wonsink, 2010) and therefore more effort due to a change of their current behavior. Furthermore, in the current energy market, customers show a lower level of loyalty toward their energy suppliers over the last years (ACM, 2015). Since changing habits is difficult, especially when consumers are not motivated and show a low level of involvement (Bagozzi 1982, Verplanken and Wood, 2006) it is challenging for companies to engage households in demand response. Therefore, energy retailers still have a struggle to create a support base for new services due to low involvement (Apajalahti, Lovio & Heiskanen et al., 2015). Nonetheless, due to the increasing competition, energy retailers need to make a shift from energy supplier to service provider (Bonnemaizon and Batat, 2011; Hannon et al, 2013). The present research examines how households can be persuaded to switch their consumer’s electricity demand change in response to financial incentives and incentives that emphasize environmental benefits. In this way, this research examines whether customers are more willing to change their behavior when they contribute to the environment by reducing co2emissions, or when they are provided with the opportunity to gain small financial benefits. Moreover, this research will shed light on how energy retailers could integrate new services such as demand response in their current strategies in order to pursue a competitive advantage in the long term.

2

found that pricing is ineffective as a demand management instrument in general (Thorsnes et al., 2012). This indicates there are contradictory results of the role of financial incentives in demand response. Moreover, what has received far less attention in research on demand response, is the role of incentives which emphasize environmental benefits rather than financial benefits (Toft et al., 2014; Delmas et al., 2013). Feelings of a moral obligation or responsibility towards the environment and having a positive contribution to society have proven to play an important role for pro-environmental behavior (Dogan., 2013; Bolderdijk., 2012; Toft et al., 2014). Incentives who emphasize the environmental benefits of behavior, have been actually even more impactful than financial incentives in encouraging pro-environmental behavior (Hargreaves et al., 2010; Asensio & Delmas, 2015). Moreover, when potential financial benefits are too small to justify the effort, it may be more effective to emphasize the non-monetary (e.g., environmental) benefits associated with sustainable behaviour (Bolderdijk et al., 2012). Spense et al. (2015) confirmed the effect of environmental framing regarding the intention to adopt smart appliances and found that individuals concerned about climate change are more likely to accept demand response, whereas those concerned about energy costs are actually less likely to accept. This might indicate that appealing to a feeling of moral responsibility might be a stronger driver to change energy consumption behavior in comparison to financial incentives. Therefore, this research focuses on a comparison of the effectiveness between financial incentives and incentives who emphasize environmental benefits in demand response.

This research fills the void in two ways. Firstly, this research fills a gap for marketing literature by examining whether environmental incentives or financial incentives have more impact on the change of energy consumption behavior. Secondly, this research fills a gap for strategy & innovation management literature with providing insights in how energy retail companies can integrate new services such as demand response into their business strategies to change their traditional form as energy supplier to becoming an energy service provider.

The scientific contribution is two-fold. This research contributes to marketing literature by examining whether incentives that emphasize environmental benefits outweigh financial incentives in behavioral change. Moreover, this research will contribute to strategy & innovation management literature by studying how firms can use strategies for behavioral change to integrate new services such as demand response into their business strategies to change from a traditional form (e.g. energy supplier) to becoming a service provider.

The practical contribution is also two-fold. This research provides insights for companies on how the effect of interventions could be maximized in persuading customer to change their current behavior while using a low involvement ‘product’. Additionally, this research will provide energy retailers with strategic implications on how they can change from a traditional form as energy supplier to become an energy service company.

Research question

3

capture value from demand response. To achieve this research objectives, the following two research questions and sub questions were developed:

1. ‘Does an environmental incentive has a stronger effect on behavioral change* than a financial incentive?

2. ‘How can firms use behavioral change strategies to integrate new value adding services** in their business strategies in order to transform from their traditional form to a service provider?’

*The behavioral change in this study is the % of energy consumption that is used during off peak hours, after the intervention, in comparison with a pre-measurement period.

**In the current research demand response is considered as a new value adding service.

The structure of this paper is as follows. First a short overview is provided of the current energy market and future challenges. Secondly, the theoretical framework is discussed. The theoretical framework is divided into two parts in order to provide a theoretical context for both research questions. In the first theoretical framework strategies for behavioral change are discussed, resulting into a conceptual model with corresponding hypotheses. The second theoretical framework elaborates on the relevant facets of a business strategy and business models and how companies can develop new technologies for services such as demand response and integrate these technologies to pursue a competitive advantage. The third chapter discusses the two methodologies which are used to answer both research questions. Finally, the last chapter will start with a discussion where the results will be compared with prior research, then discuss directions for future research and end with a conclusion with managerial implications. Note that this study is two-folded. All sections are divided into two parts, by first elaborating on operational strategies to change behavior of consumers, followed by how to integrate new services such as demand response, which require a behavioral change of customers, in the business strategy of an energy retailer.

2. Context

4

Mahajan, 2003). Because consumers are now allowed to freely choose their supplier of electricity and gas, companies need to distinguish themselves from the competition to be able to retain their customers. Therefore, companies need to invest more into the relationship with their customers, as the costs to maintain existing customers are lower than the costs to acquire a new customer (Berry, 1995). To be able to compete, companies aim at providing service quality to customers, in addition to improving efficiency and reducing costs (Sirin & Gonul, 2016). Currently, the main focus of energy service business models in the energy market is on solutions based on energy efficiency and demand response, consumer information services and distributed generation (Hamwi & Lizarralde, 2016). This research focuses on demand response.

In addition to the increasing competition in the energy market and decreasing margins, the energy market is undergoing a transition phase, where conventional energy sources such as fossil fuels and nuclear energy, need to be replaced by cleaner. An increasing awareness regarding the negative impact of fossil fuel generation on the environment, ensures energy utilities are coming under political pressure to change their generation mix from electricity generated from conventional sources, to electricity generated from renewable sources. Hence, the Dutch government aims for increasing the percentage of renewable energy in The Netherlands to 14% in 2020 (ECN, 2015) to reduce co2 emissions. Nonetheless, in 2015, there still was only 5.8% percentage of renewable energy in The Netherlands (CBS, 2015) available in the electricity grid. This means a major part is still generated from conventional energy sources such as coal plants. To realize the government goals for CO² reduction in the energy market, energy retailers need to change their conventional business models, which have been used by energy utilities for over a decade (Hamwi & Lizarralde,2016). One of these new services, which could contribute to the co2 reductions is demand response.

Demand response can contribute to the reduction of CO² by stimulating households to use energy during favorable hours when renewable energy sources are available. Moreover, due to the rising penetration of renewable energy sources and the increasing electricity demand, the electricity grid will cope with some challenges. To create a balance between supply and demand, a two - way communication needs to be established (Lin et al., 2013; Gangale et al., 2013; Karlin, 2014). Demand response can be used to match supply and demand more accurately by encouraging households to shift their electricity consumption (automatically, manually or both) in response to a price signal or other stimuli (Darby & Mc Kennar, 2012) towards favorable supply conditions (Moura et al., 2009; Giordano et al., 2012; Broeer et al., 2014; Toft et al., 2014). Nonetheless, this will require a more active role in the regulation of energy consumption (He et al.,2013; Geelen et al., 2013; Wonsink, 2010). Because in the current energy market, customers show a lower level of loyalty toward their energy suppliers over the last years (ACM, 2015) it is challenging for these companies to realize demand response. Consequently, energy retailers still have a struggle to create a support base for demand response due to low involvement (Apajalahti, Lovio & Heiskanen et al., 2015). The current research will provide more insights in how to motivate people to shift their behavior to off peak hours.

5

3. Theoretical framework

This section elaborates on the theoretical context the research questions are related to and helps to provide a guidance to answer these research questions. The first part of the theoretical framework of this section focuses on strategies for behavioral change, resulting into a conceptual model with corresponding hypotheses. The second part of the theoretical framework elaborates on the concept of value adding services and links this to how firms can integrate these new values added services in their business strategy to become a service provider. Due to the explorative nature of the second research objective, the second framework briefly elaborates on the topics that are related to the second research question and does not result into hypotheses.

3.1 Theoretical framework part I

3.1.1 Strategies for behavioral change

To change behavior, it is important to examine what are the drivers behind this behavior. Prior literature suggests values play an important role in behavioral change (Steg et al., 2014). Values determine on what specific element a person focuses on in a certain situation and are related to the goal the person wants to pursue and the extent to which these goals are chronically accessible and salient in a given situation (Steg et al., 2014). Moreover, values are believed to transcend situations and affect a wide array of beliefs, attitudes, norms, intentions and behaviors (Feather, 1995; Gardner & Stern, 2002; Rokeach, 1973). Values vary in importance and are desirable goals that serve as guiding principles in people’s lives (Schwartz, 1992). Goal framing theory (Lindenberg & Steg, 2007) suggests that consumers have three different types of goals: hedonic goals, gain goals and normative goals. Whereas hedonic goals refer to improving feelings in a certain situation such as seeking direct pleasure, seeking excitement or avoiding effort, normative goals prompt consumers to focus on the appropriateness of actions and make them especially sensitive to what they think they are ought to do, such as contributing to a clean environment. Moreover, gain goals lead individuals to focus particularly to be sensitive to changes in their personal resources, such as status and money. The three goals Lindenberg & Steg (2007) suggest in their goal framing theory, influence which information people detect and determine where consumers draw their attention to and how consumers will act in a specific situation (Steg et al., 2014). Finally, values that are proven to be particularly relevant for understanding environmental beliefs and actions are (1) self-enhancement, existing out of hedonic and egoistic values, and (2) self-transcendence values, consisting out of altruistic and biospheric values (Steg, Perlaviciute, Van der Werff, & Lurvink, 2014).

6

case (Steg et al.,2014). Nonetheless, by strengthening normative goals rather than hedonic and gain goals, consumers will focus on acting appropriately, such as providing benefits for other people, the environment, future generations, which will encourage pro-environmental actions, even when the specific action requires some costs or effort.

3.1.2 The role of moral incentives in encouraging behavioral change

When the potential financial benefits of a certain action are too small to justify this effort, it may be more effective to emphasize the non-monetary (e.g., environmental) benefits associated with sustainable behavior (Bolderdijk et al., 2012). Prior research found that feelings of a moral obligation or responsibility towards the environment and having a positive contribution to society play an important role for pro-environmental behavior (Dogan et al., 2013; Bolderdijk., 2012; Toft et al., 2014). Spense et al. (2015) confirmed the positive effect of targeting the moral obligation in relation to the intention to adopt smart appliances and found that individuals concerned about climate change are more likely to accept demand response, whereas those concerned about energy costs are actually less likely to accept.

Moral obligation is triggered by a consumer’s desire for a stable, competent and morally good self-concept, and consistency between their behavior and this self-concept (Aronson, 1992; Dunning, 2007). Engaging in morally good behavior such as demand response may foster the positive self-concept (Khan & Dhar, 2006; Sachdeva et al., 2009), and thereby provoke positive feelings (Kasser, 2002; Dunn et al., 2008), whereas immoral behavior such as harming the environment, may threaten the positive self-concept, and thereby provoke negative feelings (Elliot & Devine, 1994; Stone & Fernandez, 2008). Although the extent to which consumers attach value to environmental quality differs per person (De Groot & Steg, 2007), many consumers tend to feel a moral responsibility to look after the environment (Thogersen & Olander, 2006; Lindenberg & Steg, 2007). Therefore, it could be assumed that acting ‘green’ may elicit a positive influence on the self-concept (Mazar & Zhong, 2010), which results in positive feelings (Carter, 2011). In contrast, pursuing personal monetary gains in general does not result in moral benefits (Ariely, Bracha & Meijer, 2009), which may result in negative feelings (Dunn et al, 2010). Lindén et al., (2006) confirm this and found that providing information concerning the cost of energy use does not always reduce energy consumption among households. Additionally, other research also found that pecuniary incentives might be counterproductive for energy conservation because they might crowd out more altruistic or prosocial motivations (Bénabou and Tirole, 2005; Bowles, 2008). This might explain increasing energy usage in over thirty years of experimental field studies dating back to the 1970s (Delmas et al., 2013). Moreover, pecuniary strategies might not be effective if the financial incentives are negligible (Delmas et al., 2013; Wolak, 2011; Bolderdijk et al.,2012).

7

H1: Households who are exposed by an incentive that emphasizes the environmental benefits of a new service, show a direct higher change in behavior than households who are exposed to financial incentives.

3.1.3 Effect of intervention over time

When feedback is given over a longer time and therefore provided more frequently, this may result into more persistent effect, due to the formation of new habits during that time (Fischer, 2008; Hutton, Mauser, Filiatrault, & Ahtola, 1986; Van Houwelingen & Van Raaij, 1989). In this way, households can monitor the effectiveness of their efforts (Abrahamse, 2007). Nonetheless, changing habits is difficult, especially when people are not motivated (Bagozzi 1982, Verplanken and Wood, 2006). Moreover, the content of the feedback where a person is exposed with, may induce different ways of thinking and therefore have a different effect on the motivation to change behavior (Heyman and Ariely 2004; Lindenberg and Steg, 2007). These different ways of thinking, also known as decision frames, determine which element of a situation people consider they focus on and also how they evaluate these different elements (Lidenberg and Steg, 2007). Moreover, these decision frames determine how much effort consumers are willing to invest in a certain action (Heyman and Ariely, 2004). Therefore, it is expected that feedback that emphasizes financial benefits and environmental benefits may trigger different decision frames and therefore have a different effect on the motivation to change behavior on a longer period of time. Hence, it might affect whether people consider the behavior as worthwhile and influence how much effort they are willing to invest in the certain behavior. This might indicate the willingness to change behavior for a new service, is strongly related to whether households are motivated. Because demand response requires effort due to habit change and potential financial benefits might be too small to justify the effort, it may be more effective to emphasize the non-monetary (e.g., environmental) benefits associated with sustainable behaviour (Bolderdijk et al., 2012). Moreover, consumers are more likely to be intrinsically motivated to engage in pro - environmental behavior when they have strong biospheric values, while the opposite effect seems to hold for those with strong egoistic values and are motivated by extrinsic factors (De Groot & Steg, 2010). Thus, since values are responsible for shaping a major part of our intrinsic motivation (Kollmuss, A., & Agyeman, J. (2002), emphasizing environmental benefits might have a more persistent effect over time than incentives emphasize financial incentives. Therefore, it is expected that consumers who are exposed by incentives which emphasizes environmental benefits may be more willing to change their energy consumption behavior over a longer time in comparison with financial incentives. From this theoretical reasoning, the following hypothesis is formulated:

H2: The effect of the incentive emphasizing environmental benefits will have a more positive effect on behavioral change in the long term, than an incentive that emphasizes financial benefits

FIGURE 1 Conceptual model

Incentive via letter

Behavioral change

Incentive via e-mail

Environmental vs financial incentive

H1,2 +

+

8

3.2 Theoretical framework part II

3.2.1 Value adding services

Before elaborating on how to integrate new services such as demand response in a business strategy, the concept of a new value adding services will be discussed. What services distinguishes from products, is that services are intangible, inseparable, heterogeneous and perishable (De jong & Vermulen, 2003). Services are in that sense unique, and have to be customized to the individual customer needs (Pine & Gillmore, 1999). The integration of new services is therefore closely related to adding value to the individual customer. To integrate new services which are based on new technologies, consumers need to see the added value of the service. By knowing how to integrate value adding services, it might be helpful to take a closer look at what determines a value. In prior research on behavioral change, values are considered as important (Steg et al., 2012; Steg et al., 2014). Thus, in changing behavior to engage customers in new services, it is important to take into account customer values, as already mentioned in the first part of the theoretical framework.

According Rokeach (1973) a value is ‘an enduring belief that a specific mode of conduct or end-state of existence is personal or social preferable to an opposite or converse mode of conduct or end-state of existence’. Additionally, the role of adding value to the customer is considered as important in establish a competive advantage and realizing a long-term success (Naumann, 1995). Hence, to become a service provider, it is essential to integrate value-adding services to add value to the customer (Berry & Parasuraman, 1992). In the energy market, energy suppliers can add value to the customer by providing them with insights and advice about their energy consumption, which can be used to minimize or change energy consumption and therefore e.g. reduce costs. This can be realized by e.g. a new service such as demand response.

3.2.2 Becoming a service provider

In an industry where margin decreases due to commodisation and lower entry barriers encourage increased competition focused on low costs strategies, companies need to seek for new ways to extend their business with new offerings (Reinartz and Ulaga, 2008). Often these new offerings include to a high extent service content (Matthysens and Vandebempt, 2008). Focusing more on service offerings, also provides firms the opportunity to broaden their market scope. The shift from offering products to offering services and solutions through products is also known as servitization (Vandermerwe and Rada, 1988). For many industries, this means firms are not only extending their product portfolio with services, but also may change their value proposition, or even of its business model (Kindstrom, 2010).

9

pursuing a dynamic service offering portfolio that is adaptive to the dynamic customer needs.

Moreover, a highlighted concept which is considered as crucial importance for servitization, includes intangible input factors such as workforce, innovative capabilities, and customer orientation (e.g., Baines et al., 2009; Kindström, 2010).These factors have played a minor role in the energy sector, until now (Helms, 2016).

Concluding, literature on servitization has mainly focused on the manufacturing sector (Robinson et al., 2002). Nonetheless, even greater challenges could be expected in the context of the energy market, where commodity suppliers need to transform from an energy supplier to an energy service provider (Baines et al., 2009; Kindstrom). Prior literatures emphasizes energy retailers still struggle to create a support base for new energy services (Apajalahti, 2015). Therefore, this research gains more insights in how to use strategies for behavioral change to integrate new services into the business strategy to transform from an energy supplier to a energy service provider.

4. Research design and methodology

To create a complete overview of consequences of the implementation of a strategy for behavioral change, quantitative and qualitative research are both conducted. First, quantitative research is conducted to test whether households will be motivated to switch their energy usage to certain incentives. This is done by setting up an experiment. Secondly, exploratory qualitative research is used to get a deeper understanding of the vision regarding demand response and explore how energy retailers can use behavioral change strategies to integrate services such as demand response in their business strategy. Thirdly, the results of the experiment are discussed in a short follow-up semi- structured interview. By making use of two different types of collecting data we can provide a stronger substantiation of constructs and propositions (Eisenhardt, 1989). Moreover, observing the research phenomenon through both quantitative and qualitative data collection will ensure that, with qualitative research a deeper understanding of the theory underlying the relationships will be derived and quantitative data relationships are indicated which may not be salient to the researcher (Jick, 1979). The following paragraphs will elaborate further on both methodologies. First the methodology will be discussed for the experiment, followed by the methodology of the in -depth interviews.

4.1 Experiment

10

4.1.1 Research design

The field experiment is conducted by testing a 2 (environmental vs financial) between-group and within-group design on a sample of 100 households across the Netherlands. A pre- measurement was used to verify the behavior before the intervention and compare this with the period when the households are exposed by the intervention. All participants first received a letter with information about the incentive, followed by 7 weekly emails with information about the condition they got assigned to. The households in the environmental condition group, received an email which contained a) information on their switching behavior of the previous week and b) the number of times they switched to off peak hours during the current week and c) a statement regarding in how the number of times that a person switched, contributed to the environment. Households in the financial condition group, received an email which contained a) information on their energy consumption of the previous week and b) for the number of times they switched to off peak they receive a message that they received a financial incentive including the difference between APX price during off peak hours versus peak hours, which is ranging from 1ct to 3ct per time. This e-mail can be found in the appendix.

Additionally, all participants were exposed with two extra incentives. First, all participants received a sticker, which can be considered as a prompt, with the message (co2 vs euro) that belongs to the condition they got assigned to. Prior research has found that prompts (reminder) have an effect on pro- environmental behavior (Luyben, 1982). The current study does not test this effect separately. Moreover, to maximize the likelihood of switching, the performance of the entire group is also mentioned in the e-mail. This means participants will receive weekly feedback about their own behavior as well as the total result of the behavior of the rest of the group. Prior literature (Steg & Vlek, 2008) has found role models and social support can be provided to strengthen social norms and informing actors about the perceptions, efficacy and behavior of others. Hence, pro-environmental behaviour can be supported by modelling and providing information about the behaviour of others (Schultz, Nolan, Cialdini, Goldstein, & Griskevicius, 2007; Abrahamse et al., 2005; Lehman & Geller, 2004).

4.1.2 Data collection and structure

The data for the experiment was gathered, in collaboration with a technical aggregator and a large energy retailer, among 100 Dutch households over a period of 11 weeks from the 12th _{of October till} the 1st of January. The households were selected from a group of customers of a large energy retailer in The Netherlands that currently have a smart thermostat. On beforehand, households were selected on the basis of whether they had a timer on all three white good appliances, so they had the opportunity to program their white good appliance during the off -peak hours in the night. Data was collected by using smart plugs.

The data from the households are divided into two datasets. The first dataset includes time series data concerning energy usage of three white good appliances on a daily basis from 100 households, over a period of ten weeks from the 12th _{of October till the 1}st _{of January in 2016. The second dataset has a} cross- sectional structure, and contains all 100 households, and the group they got assigned to.

4.1.3 Procedure

11

On the 20th of September, respondents were asked via email to participate in an experiment. Participants were told in the email the experiment aimed at improving the quality of deriving data from smart plugs. Three weeks later, on the 12th of October the participants that agreed upon participating in the experiment, received a package with a smart plug with information about this smart plug and how to install it to their white good appliances. After installation of the smart plug, current energy usage could be analyzed as input for the pre-measurement. During the pre-phase, the participants also were assigned equally to two groups, based on two criteria. In a prior survey, that was send out by the large energy retailer two months before, participants were asked whether they perceived barriers, which inhibited them to not use their white good appliances during the night. The participants that were divided and assigned equally to one of the two groups. Furthermore, the participants were also equally assigned on the basis whether they had a contract taking into account a day/night tariff, also known as day & night meter. A day & night meter offers consumers the possibility to use electricity at lower cost during night hours. Hence, participants that already have a day/night tariff might be more willing to switch their behavior. Within the both groups, participants were randomly assigned to either the financial condition of the moral condition group. On the 1st of November, the participants received a letter with information about the experiment and a sticker, that they could stick on their white good appliances as a reminder for participating in the experiment. Furthermore, in the last week of September a short survey was conducted among a different sample than the experiment sample group including a number of statements concerning moral obligation on a different sample of 20 consumers, to examine which statement evokes the strongest feelings of moral obligation. This was used as a manipulation check.

In the during phase, participant received a letter (on the 1st of November), followed by 7 weekly e-mails in a period of two months from 22nd of November till the 1st of January. Both the letter and the e-mail included either a financial incentive or an environmental incentive. On the 22nd of November, the participants received the first email with information about their performance. The letter only contained general information about the incentive. The design of the letter and the e-mail can be found in the appendix 1 and 2.

After conducting the experiment, the data was analyzed. The appropriate analysis is determined on the basis of the type of data and the purpose of the analysis. The purpose of the current study, is to examine whether the effect of the intervention changes over time and whether this differs between the euro and the co2 group. In this way, we can identify whether an environmental incentive has more effect than a financial incentive in encouraging behavioral change. The data consists of several time-series variables, thus in order to not lose the richness of the data, dynamic variables should be included. In the following section the data a short description of the data is provided.

4.1.4 Variables

12

𝑦"#$ = (off peak count/daily total count) *100

Lagged dependent variables

To prevent autocorrelation one lagged variable was included in the model, formulated as: 𝑦"#$%& = (off peak count/ daily total count) *100

Direct effect intervention via letter

The incentive is given in the form of a letter and an e-mail. The letter, is sent on the 1st _{of November.} The effect of the letter is coded into a binary variable (1/0) on the day the participants received the letter. In this way, the direct effect of the incentive is measured.

Lagged effect incentive via letter

To capture the effect a day after the intervention of the letter, the day after the intervention by letter, was coded into a binary variable (1/0).

Long term effect intervention by letter

Takes into account the whole period after the intervention of the incentive by letter. This means that all periods after the incentive are coded into 1 and periods before the intervention are coded into 0. In this way the model examines whether the effect of the intervention maintains, decreases or increases. Direct effect incentive via e-mail

The mail is coded into a binary variable (1/0) on the days when a certain participant opened the e-mail (on the 22nd_{, 25}th_{, 30}th _{of November and the 7}th_{, 14}th_{, 21}st _{and 28}th _{of December). When the e-mail} was send but not opened, a 0 was written down.

Lagged effect incentive via e-mail

To capture the effect a day after the intervention of the e-mail, the day after the intervention by letter, was coded into a binary variable (1/0).

Long term effect intervention by e-mail

By studying the percentage off peak over time with descriptive statistics, it was detected the period after the e-mail was send, there is an increase of the percentage in off peak after the first intervention by e-mail. Nonetheless, this effect decreased over time. To capture the overall effect in the model, all days, after the first time a household opened an e-mail, were coded into a binary variable (1/0). Financial incentive

The participants that are assigned to the “euro” group will be exposed by a financial incentive via a letter that is sent on the 1st _{of November and the weekly e-mail. This financial incentive is calculated} by the difference in APX price per kWh during off peak hours in and peak hours. A kilowatt-hour (kWh) is one of the most common unit of electricity used by utilities in residential and commercial billing.

Environmental incentive

13

consequences will be communicated regarding how the reduction of co2 emissions contribute to the environment.

Weekend variable

The descriptive statistics, indicate there is a large drop of percentages using white good appliances during off peak hours during the weekends on a weekly basis. Therefore, a weekend variable is included, which is coded into a binary variable (1/0).

Holiday

The descriptive statistics show a large drop of percentages using white good appliances during off peak hours during holidays. Therefore, a holiday variable is included, which is coded into a binary variable (1/0).

No usage

Because some appliances are not used on a daily basis and show a difference in usage behavior, a no usage dummy is included in the model.

Trend variable

Because it is expected the effect of the intervention will be linear, the model includes a trend variable. Adding a deterministic trend is also a possible remedy for non-stationary variables.

4.1.5 Model specification

Both models are specified as linear additive models in order to fully comprehend the variables individually. Moreover, in the specification the five criteria of Little (1969) are taken into account: Simple, Evolutionary, Complete, Adaptive and Robust. The first model is a unit-by unit model, since all households are analyzed separately. This will result in 100 regressions, including one regression per household. Because model 1 includes a lagged sales variable to deal with autocorrelation, it is considered as a partial adjustment model (Leeflang et al., 2015). For the second model, the estimates of the betas (β) of the intervention (both mail and letter) of model 1, are used as dependent variable.

Model 1 is formulated as follows:

𝑦"$ = 𝛼"$+ 𝛽&"𝑑𝑖𝑟𝑒𝑐𝑡𝑒𝑓𝑓𝑒𝑐𝑡𝑙𝑒𝑡𝑡𝑒𝑟"$&+ 𝛽4"𝑙𝑒𝑡𝑡𝑒𝑟𝑑𝑢𝑚𝑚𝑦 "$&+ 𝛽7"𝑙𝑜𝑛𝑔𝑡𝑒𝑟𝑚𝑒𝑓𝑓𝑒𝑐𝑡𝑙𝑒𝑡𝑡𝑒𝑟"$

+ 𝛽;"𝑑𝑖𝑟𝑒𝑐𝑡𝑒𝑓𝑓𝑒𝑐𝑡𝑚𝑎𝑖𝑙"$+ 𝛽="𝑑𝑖𝑟𝑒𝑐𝑡𝑒𝑓𝑓𝑒𝑐𝑡𝑚𝑎𝑖𝑙"$%&

+ 𝛽>"𝑙𝑜𝑛𝑔𝑡𝑒𝑟𝑚𝑒𝑓𝑓𝑒𝑐𝑡𝑚𝑎𝑖𝑙"$&+ 𝛽?"𝑦"$%&+ 𝛽@"𝑛𝑜 𝑢𝑠𝑎𝑔𝑒"$+ 𝛽B"𝑤𝑒𝑒𝑘𝑒𝑛𝑑"$

+ 𝛽_&E"ℎ𝑜𝑙𝑖𝑑𝑎𝑦_"$+ 𝛽_&&"𝑡𝑟𝑒𝑛𝑑_$") + 𝜀_"$ Model 2a: 𝛽_&" = 𝛼 + 𝛽_&𝑖𝑛𝑐𝑒𝑛𝑡𝑖𝑣𝑒_"+ 𝜀_"

Model 2b: 𝛽;" = 𝛼 + 𝛽&𝑖𝑛𝑐𝑒𝑛𝑡𝑖𝑣𝑒"+ 𝜀"

Model 2c 𝛽7" = 𝛼 + 𝛽& 𝑖𝑛𝑐𝑒𝑛𝑡𝑖𝑣𝑒"+ 𝜀"

Model 2d 𝛽>" = 𝛼 + 𝛽& 𝑖𝑛𝑐𝑒𝑛𝑡𝑖𝑣𝑒"+ 𝜀"

14

𝛼 = Constant

𝑦_"#$= % off peak on a day level (times a household used energy during off peak hours/ total)

𝑡 = Period

𝑑𝑖𝑟𝑒𝑐𝑡𝑒𝑓𝑓𝑒𝑐𝑡𝑙𝑒𝑡𝑡𝑒𝑟 = Intervention by letter

𝑙𝑒𝑡𝑡𝑒𝑟𝑑𝑢𝑚𝑚𝑦 Takes into account the effect of the intervention by letter two days after the letter was send.

𝑑𝑖𝑟𝑒𝑐𝑡𝑒𝑓𝑓𝑒𝑐𝑡𝑚𝑎𝑖𝑙 = Intervention by e-mail

𝑙𝑜𝑛𝑔𝑡𝑒𝑟𝑚𝑒𝑓𝑓𝑒𝑐𝑡𝑚𝑎𝑖𝑙 = Takes into account that there was an intervention by e-mail the day

𝑙𝑜𝑛𝑔𝑡𝑒𝑟𝑚𝑒𝑓𝑓𝑒𝑐𝑡𝑙𝑒𝑡𝑡𝑒𝑟= Takes into account the effect of the total experiment after the letter was send.

𝑔𝑟𝑜𝑢𝑝 = The group a household got assigned to. (1=CO2 group,

0=euro group)

𝑖 = Household, 1=household 1, 2= household2, 3= household3,

N=100

𝑡𝑟𝑒𝑛𝑑 = Day number: 0,1,2,3…T=81

no usage= Dummy variable for no usage, coded into 1/0

𝑖 = Household, 1=household 1, 2= household 2, 3= household3,

N=100

𝑡 = Day number: t= 1,2,3…

𝑌&"$%& = % off peak in the previous period (t -1) of household i;

𝜀&"$ = Residuals of household i, week t

4.1.6 Plan of analysis

Before discussing the plan of analysis, first a short overview of the steps is provided in table 1.

TABLE 1

Overview steps in the analysis

Methodology

Description

Literature

Ordinary least squares for 100 households per white good appliance

How does the intervention of the letter and the e-mail affect the behavioral change?

x

Testing residual assumption

Testing for assumption with regard to the residuals.

x Added z test to test significance of variables which

will hold as dependent variable in model 2 (direct and long term effect intervention)

Rosenthal (1991)

Give a weight to the betas by multiplication with inverse of standard deviation (1/std dev)

Nijs et al., (2001); Gijsenberg (2013)

Perform a weighted least squares

How do the hedonic costs and the type of incentive affect the response to the intervention?

Nijs et al., (2001); Gijsenberg (2013)

Testing residuals assumptions

15

To analyze to what extent households switch and therefore change their behavior, a time series model is performed. In this way, the model sheds light on the effect of the intervention on the behavioral change over time, and tests whether the effect declines, increases or remains stable over time. To ensure the model does not have to deal with multi-level moderators, a two - stage regression method is performed, which is partly derived from previous literature (Nijs, Dekimpe, Steenkamp, & Hanssens, 2001; Steenkamp, Nijs, Hanssens, & Dekimpe,2005; Gijsenberg, 2013). In this way, we first test the effect of the intervention over time, and subsequently examine whether this effect differs per group, the usage behavior and/or the hedonic costs the households perceive. By linking these two stages together, the independent variables, which capture the effect of the intervention, will be used as dependent variables in the second stage.

The first stage consists of the estimation of a time series model, which examines on a daily basis the effect of the intervention on the percentage a household shifted to off peak hours. For this analysis, an Ordinary Least Squares (OLS) is performed. The effects of the intervention are estimated separately for each individual households, which results into 100 separate OLS regressions. These 100 separate OLS regressions will provide estimates for the βs and the standard deviations for the effects of the intervention of the letter and e-mail, which will be used as the dependent variables for the second model. The overall significance goodness of fit of the model is tested with a F test per household and a corresponding R square. Moreover, the significance of the independent variables, which capture the effect of the intervention, will be tested by the added Z test (Rosenthal, 2001). By using the inverse of the standard deviation, the Z-score of the p value will be calculated. Subsequently, the sum of the z-scores are divided by the square root of the number of households, resulting in an overall z score. This final z score, is used for the independent variables to determine the overall p-value and therefore the significance.

The second stage consists of a moderator analysis. The analysis will include the barriers, the nature of the intervention (co2 vs euro) and the usage behavior (high, medium or low) as moderators and test whether the effect of the intervention on the behavioral change, is influenced by these variables. For the second stage a Weighted Least Squares (WLS) regression is performed in the second model. This is the appropriate analysis to the data, because it will ensure the estimates of the second stage analysis will be more accuretely (Leeflang, Wieringa, Bijmolt, & Pauwels, 2015; Steenkamp et al., 2005). In a WLS regressions all variables are weighted by the inverse of the standard deviation of an equation’s dependent variable, (Nijs et al., 2001; Steenkamp et al., 2005).

Concluding, the modeling approach of the current study is chosen to capture the effect of an intervention over time in addition to comparing only means between groups and periods. By just comparing the mean before and after an intervention and the differences between groups, without accounting for trends in the data, the effect of the intervention could be over- or underestimated. Moreover, this may result in a loss of valuable information. This study takes into account effects over time and estimates the model on an individual level. Moreover, by adding a step dummy for the letter, which is coded into 1 in all periods after the intervention has taken place, the difference between the two periods is also accounted for.

16

due to the time dynamic nature. To test whether the residuals were meet the assumption, the Breusch Godfrey test Lagrange multiplier test is conducted. The null hypothesis of the Breusch Godfrey Lagrange multiplier test holds, there is no autocorrelation among the residuals. If the test shows a significant result the residuals are correlated with each other. To already account for serial correlation, a lagged dependent variable will be included in the first model (Leeflang et al., 2015) The second model is performed on cross- sectional data, and since autocorrelation is not a problem for this type of data, we do not have to test on whether the residuals are auto correlated. The second assumption tests whether the residuals the same variance in all cases over time and therefore homoscedastic (Leeflang et al. 2015). When this assumption does not hold, the residuals are violated with heteroscedasticity and adjustments should be made in the analysis. In this case, p value cannot be trusted. Heteroskedasticity could be solved by using weighted least squares. For the first model, the Breusch-Pagan (1979) and White (1980) test are used to test the residuals on hetereoscedasticity. The Breusch-Pagan

test tests

whether the estimated variance of the disturbance term is dependent on the independent variables of the model. For the second model the residuals are tested on hetereoskedasticity with the Levene’s test. If this is the case, the heteroscedasticity assumption is violated. If the independent variables from the model explain the squared residuals, a pattern in the residuals still exists. The third assumption examines the residuals are tested on non-normality. An ordinary least squares regression requires a normal distribution of the disturbances, to ensure the standard test statistics for hypothesis testing and confidence intervals are applicable. The third assumption is tested for both models. First, the residuals are examined by examining QQ and PP plot per household and conducting the Shapiro-Wilk test. Leeflang (2015) argues, especially in small data sets, OLS estimates are sensitive to outliers, because large residuals receive a lot of weight in the least squares minimization problem. Hence, when the non-normality is caused by outliers, the OLS estimates may not be unbiased (Leeflang, 2015). Moreover, when the model specification seems appropriate it is not required to follow strict rules with regard to violations of non-normality (Leeflang, 2015).

4.2 Interviews

4.2.1 Research design

17

research is reliable because the interviews were conducted at different moments in time on different days, as proposed by Van Aken et al (2012).

4.2.2 Data collection and case selection

The sample selection of interviewees was done by theoretical sampling (Eisenhardt and Graebner, 2007). The selection of organizations was conducted following the approach of Yin (2003) to ensure a wide spectrum of different visions on demand response were covered. To ensure the research is reliable, a mix of organizations was selected. This ensured the selection of respondents results to the same results as should be obtained with other respondents (Van Aken, Berends & van der Bij, 2012). For this study, 11 interviews were conducted with managers and employees within Eneco, Stedin, ACM, Netbeheer Nederland, ECN and Quby. Below the organizations are discussed. Table 2 represents the detailed information about the sample. The first interviews are conducted face to face between the 4th _{of October till the 5}th _{of December 2016. The follow up interviews are conducted} between the 2nd _{and 13}th _{of January 2017, by sending a structured questionnaire via e-mail. Only 6 of} the 11 respondents indicated they had the opportunity to answer the questions of the follow up interview. Two follow up interviews were conducted by telephone, because the interviewees indicated they preferred to discuss the experiment via telephone. The other four follow-up interviews send the written questionnaire via email.

Stedin

Stedin is a distribution system operator (DSO), who performs both the physical distribution of the energy supplied by suppliers and the administrative processes related to it. Within Stedin an interview was conducted with an internal consultant of the strategy & innovation department. The respondent is chosen to gain more insights into the vision from a distribution system operator with regard to demand response.

Netbeheer Nederland

Is the industry organization is responsible for all distribution system operators and the transport system operator in the Netherlands. Within Netbeheer Nederland an interview was conducted with the manager at the strategy department. The respondent was chosen to gain more insights into the vision from a distribution system operator with regard to demand response. The first interview is conducted on the 10th _{of November 2016 and was face to face.}

ACM

ACM is the Authority for Consumers and Markets in The Netherlands, which is responsible for ensuring fair competition between businesses, and protects consumer interests. Within ACM an interview was conducted with a senior inspector at the energy department. The respondent is chosen to gain more insights into regulatory side of the development of new services such as demand response.

Eneco

18

how this relates to the European government. The manager at the strategy department was chosen to gain more insights into how demand response can be integrated into the strategy of energy retailers. The manager of the fundamental analysis department is chosen to gain more insights into the role of demand response is in future scenarios, and how this relates to the current and future strategies of energy retailers. The analyst was chosen to gain more insights into how demand response can be integrated into the strategy of an energy retailer, and how the trading department need to be adapted to allow new services such as demand response to be a success.

Quby

Quby is a technical aggregator who provides smart thermostats for energy retailers, which enable to open up a dialogue with their customers. Within Quby two interviews were conducted with the manager Smart Energy and the innovation manager that is responsible for Demand Side Management. The manager of the smart energy department is responsible for the development of new products for energy retailers such as smart thermostats and other smart appliances and services which can be built surround this. The innovation manager is responsible for the development of services concerning automatic and manual demand response.

ECN

Energy research Centre of the Netherlands (ECN) is the largest energy research institute in the Netherland. ECN develops new technology and conducts pioneering research in various ways into innovative solutions to facilitate the transition to sustainable energy management. Interviews were conducted with one behavioral researcher and one market researcher, where he focuses on the market mechanism and regularly affairs. The behavioral scientist focuses mainly in his role on how consumers can be stimulated to participate in energy related projects and how consumers can be motivated to deal in a more sustainable with their energy consumption by e.g. energy efficiency projects and demand response projects. He is chosen to share his knowledge concerning strategies for behavioral change and how this relates to the strategy of energy retailers. The market researcher was chosen to gain insights concerning how the market mechanism and legislation in the energy market, influences the success of the development of services such as demand response.

TABLE 2 Summary sample description

Organization Education

level

Men/Women Age Department Function

19

Netbeheer Nederland (1)

University Male 55-60 Strategy Manager

ACM (1) University Male 30-35 Energy research Senior inspector

Eneco (4) University Male 30-55 Trading, Fundamental Analysis,

Strategy, Regularity affairs

Manager(3), Analyst (1)

Quby (2) University Male 35-40 Smart Energy, Demand Side

Management

Manager (2)

ECN (2) University Male 35-40 Market, Behavioral research Researcher (2)

() = nr of interviewees for that specific organization

4.2.3 Procedure

The interviewees were contacted via email with a short introduction to the research topic and question to participate. Moreover, all participants were provided with a questionnaire which guided the conversation. The specific questions, can be found in the appendix. The interviews were recorded and transcribed to prevent a respondent bias and lasted on average 120 minutes. To decrease the odds for an interviewer bias, the interviews were recorded and notes were made additionally. This increased the reliability of the study, which means that the research supplies consistent results and is replicable. The respondents were asked open-ended questions to explore underlying motivations, beliefs, attitudes, and feelings on demand response will be uncovered (Malhotra, 2012). Moreover, some probing questions were asked to establish details such as when a particular event would occur and in what context this would occur. The first interviews had four sections. First the background of the respondent was elaborated on, followed by their vision on demand response in general. The second part of the interview focused on how companies can engage customers, strategic issues and in what way demand response could be realized in a context where financial incentives are relatively low. The third part focused on particular objects that need to be changed in the industry to be able to pursue a successful strategy for demand response. The follow up interview was a more structured interview that asked respondents how the outcomes of the experiment could be related to different strategy components of companies.

4.2.4 Analysis

The goal from the data analysis of the in-depth interviews is to determine whether there are significant impacts between the constructs and variables. As is typical in inductive research, the data is analyzed by first building individual case studies and then comparing cross cases to construct a conceptual framework (Eisenhardt, 1989). After the within case analysis, a cross-case analysis will be conducted, by using the method suggested by Eisenhardt (1989a), to develop the conceptual insights, which allows to identify common dilemmas and refine the unique aspects of each particular case. After the cross- case analysis, the results will be compared by existing theories.

20

the selective code.Moreover, the themes are linked to subjects that were raised during the interview and the main research question. After structuring the major themes and concepts, smaller themes and relations and insights are identified through axial coding. In order to reduce the volume and identify core consistencies and meaning, substance analysis of the collected data is conducted. Data is discarded which was considered as irrelevant to answering the main research question (Patton, 2002). With a comparative analysis of the codes, the data is verified on how codes relate to the integration of new energy related services such as demand response in the strategy of companies. The data is illustrated in four matrixes, with for each matrix on the horizontal axis the respondents and on the vertical axis one of the four topics that guided the interviews and the axial codes which were mentioned by the interviews. With regard to the cross-case analysis compares the 11 cases with each other and discusses the differences and similarities between the cases. Subsequently, data will be summarized in tables which summarize and tabulate the evidence underlying the construct to increase construct validity (Miles & Huberman, 1984; Sutton & Callahan, 1987). As follows underlying theoretical reasons will be examined to find out why certain relationships exist, which helps establishing internal validity of the findings (Eisenhardt, 1989).

5. Results

This section provides insights with regard to the results of the experiment and the in-depth interviews. First the results of the experiment will be discussed, followed by the results of the within and cross case analysis.

5.1 Experiment

This section elaborates on the results of the two models, which were used to analyze the data and answer the research question. First, the data is descripted with descriptive statistics. Subsequently, the direct effects of the model 1 are elaborated on. Furthermore, the model is tested for assumptions and violations of the residuals, indicating whether the outcomes of the regression are biased. To solve these violations the remedies are applied.

5.1.1 Data cleaning