Business model innovation and market share

Business Model Innovation and Market Share

Master thesis – 30.06.2013

Supervision: Dr. Dipl.-Wirt.-Ing. Sebastian Kortmann

Abstract

Current study investigates the association between business model innovation and market share growth on the level of the business unit. It collects data from a period of five years. By distinguishing

between new entrants and incumbents, it investigates for which of the two business model innovations work better. The outcomes show that there is no association between business model

innovation and market share growth. However, new entrants do serve as a moderator in the relationship, but only after a period of 2 and 3 years, or over a period of 2 years.

Content

Abstract p. 2

Introduction p. 4

Theoretical framework p. 5

- Business model p. 5

- Business model innovation p. 6 - Performance: market share p. 7 - New entrants versus incumbents p. 8

- Research question p. 9 - Hypotheses development p. 9 - Model p. 12 Methodology p. 12 - Type of survey p. 12 - Sample p. 13 - Measuring p. 15

- The independent variable p. 15 - The dependent variable p. 16

- The moderator p. 17

- The control variables p. 17

- Data handling p. 19

- Description data set p. 20

- Respondents p. 20

- The business units p. 20

Analysis p. 21

- Results p. 22

- Hypotheses tests p. 23

- No lag: the basic model p. 23 - No lag: other variables p. 23 - With lag: the basic model p. 24 - With lag: other variables p. 25 - Periodic effect: the basic model p. 25 - Periodic effect: other variables p. 25

Discussion p. 26

- No lag: the basic model p. 26 - No lag: other variables p. 28 - With lag: the basic model p. 29 - With lag: other variables p. 30 - Periodic effect: the basic model p. 30 - Periodic effect: other variables p. 31

Limitations p. 32

Future research p. 32

Conclusion p. 33

References p. 34

Introduction

A business model innovation (BMI) is a change in “the content, structure, and governance of transactions designed so as to create value through the exploitation of business opportunities” for a firm (Amit and Zott, 2001 – p. 511).

It seems difficult to pull off, especially for incumbents (Johnson, Christensen and Kagerman, 2008; Amit and Zott, 2010) that are successful (Teece, 2010). Only a few do so (Yip, 2004). Outcomes are thus uncertain. The lack of study in the field of BMI and a company’s lack of understanding of the current business model (BM) are reasons for this difficulty (Johnson, Christensen and Kagerman, 2008). The consequence of this lack of knowledge and understanding is that companies do not know when they should change their BM (Johnson, Christensen and Kagerman, 2008) and in which direction (Chesbrough, 2010). To make things worse, new entrants use this inertia (Mitchell and Coles, 2003) and jump into the gaps left open. Virgin is a good example (McGovern and Moon, 2007). New entrants do not have conflicting business models (Chesbrough, 2010; Markides and Charitou, 2004), and are not bound by path dependencies (Zott and Amit, 2007).

This study tries to help resolve the vagueness and uncertainty surrounding the subject. Empirical proof on the outcomes of BMIs is lacking (Zott, Amit and Massa, 2011). Therefore this study contributes by investigating that. It does so by using the BM concept as an explanatory variable for value creation, competitive advantage and firm performance for individual business units. With panel data over a period of 5 years it looks at market share growth as the measure of performance. Subsequently it analyzes this data in relation to the degree of radicalness of the business model innovations within this period while differentiating between incumbents and new entrants. This distinction is a key focus of this paper since it is suggested that results between the two differs. The focus on market share is interesting because, although proof lacks, more than once market share losses or gains are emphasized in relation to the introduction of a BMI (Markides, 2006; 2013; Koen, Bertels and Elsum, 2011; Johnson, Christensen and Kagerman, 2008; Markides and Charitou, 2004). By doing this, the paper proofs the empirical usefulness of the BM concept, provides useful input for managers, makes causal inferences through the collection of longitudinal data, and develops a useful measure for the degree of innovativeness of business models.

The questions that guide the paper’s aim are: Is there a relation between business model innovation and market share development?; and do new entrants moderate that relationship?

To come to a structured answer the paper starts with a literature review. Subsequently it continues posing the hypotheses. After that the method section lays out how data is gathered and dealt with in

order to be able start the data analyses. That section follows after the method section. The section thereafter is the discussion followed by the limitations and future research sections. All this will lead to a conclusion which forms the last section of this paper.

Theoretical framework

Business model

More and more research on business models is performed. However, a theoretical foundation is not yet established (Teece, 2010; Shafer, Smith and Linder (2005); Zott, Amit and Massa (2011). Onetti et al. (2010) found 48 different relevant definitions. It causes the term to be opaque (Zott, Amit and Massa, 2011¹; Shafer, Smith and Linder, 2005; Onetti et al., 2010). This is not surprising (Gladwin, Kennelly and Krause, 1995; in Zott, Amit and Massa, 2011). In contrast to the lack of consensus on the definition of a BM, there is clarity on what a BM is not (Zott, Amit and Massa, 2011). That is, it is not strategy (Yip, 2004; Chesbrough and Rosenbloom, 2002; Magretta, 2002; Casadesus-Masanell and Ricart, 2010; Onetti et al., 2010). More specifically, it is not product market strategy, nor corporate strategy (Zott, Amit and Massa, 2011). Neither is it a straightforward mechanism for the creation of value or limited to the boundaries of a firm (Zott, Amit and Massa, 2011). This implies that a BM describes more than characteristics, transactions or activities of the firm.

In line with this statement, Hamel (2000) in Shafer, Smith and Linder (2005), Amit and Zott (2001), Casadesus-Masanell and Ricart (2010) and Chesbrough and Rosenbloom (2002) state that a BM illustrates a network of the value created for the firm, its partners, customers and suppliers. It demonstrates interdependence of the activities that take place within that network and thus shows that a firm’s BM spans its own boundaries (Amit and Zott, 2010)³.

With respect to the interdependence of activities, Amit and Zott (2001) see a BM as an activity system. This is an important notice. It overlaps different definitions given in the literature. This point is illustrated by, among others, Morris et al. (2005), Johnson, Christensen and Kagerman (2008), McGrath (2010), Onetti et al. (2010), Santos et al. (2009), Sorescu et al. (2011) and Teece (2010) who all (in)directly talk about activities in relation to the BM concept.

So despite the mentioned opaqueness, overlap between the different definitions is observed (Zott, Amit and Massa, 2011; Shafer, Smith and Linder, 2005; Onetti et al., 2010). In order to provide some consistency within the literature, this paper will therefor adopt the definition of Amit and Zott (2001) in combination with their “add-on” from Zott and Amit (2010). It does so because based on Zott, Amit and Massa (2011) and Sorescu et al. (2011) it appears the most accepted definition and/or similar to other definitions (e.g. Santos et al., 2009). They define a BM as follows: it is a unit of

analysis that “depicts the content, structure, and governance of transactions designed so as to create value through the exploitation of business opportunities” (p. 511). The related activities span the firm’s boundaries and are interdependent (Zott and Amit, 2010).

First of all, the content of the transactions refers to what is being exchanged and what is required for that exchange. Secondly, structure of the transactions refers to which stakeholders are involved in the exchange, how they are connected and in which sequence exchanges are made. Lastly, the governance of transactions refers to why the stakeholders would engage in the transaction (incentive), the firm’s legal form, and how the stakeholders control the flows of resources, goods and services (Amit and Zott, 2001).

Having defined BMs, two points need be made regarding research on BM. Zott, Amit and Massa (2011), for clarity reasons, say future research should choose which concept of the BM, as the foundation of the study, it uses. In the literature they find three different concepts. Those are: “e-business and the use of information technology in organizations; strategic issues, such as value creation, competitive advantage and firm performance²; and innovation and technology management”. Furthermore, for developmental reasons, future research should aim to proof the empirical usefulness of the BM concept by, for example, assessing its consequences (Zott, Amit and Massa, 2011).

¹ Zott, Amit and Massa (2011) found that of the articles on BMs they analyzed, only 44% explicitly defines BMs, 19% refers to other researchers and a stunning 37% does not define BMs at all. It does show however an increase as Chesbrough and Rosenbloom (2002) find that a definition for BMs is rarely given.

² Lambert and Davidson (2012) also found this theme as one of the most recurrent themes in the BM literature.

³ In 2009 Santos et al. critique the 2001 definition of a BM of Amit and Zott because it excludes external linkages with parties. In 2010 Zott and Amit recognize that point.

Business model innovation

For the purpose of this paper, the Sorescu et al. (2011) definition of BMI is adopted. It is a straightforward definition and practical in its applicability across a wide variety of papers. Less practical would be the Markides (2006) definition which says the innovation needs to enlarge the market. Investigating that goes beyond the scope of the paper. Sorescu’s et al. (2011) definition on the other hand implies that the innovation can be radical and incremental. Therefore it is applicable and practical to current study as it looks at the degree of innovativeness in relation to market share development. A scale that is able to identify the degree of innovativeness is developed as Burkhart et al. (2010) and Morris et al. (2005) suggest should be. Only a few studies do so (Bucherer et al., 2012).

Furthermore, performing current literature review shows that Sorescu et al. (2011) are among the few who define BMI. They do so as follows: a BMI is “a change beyond current practice with respect to” at least one of the “three core components and their interdependencies” (p. 2). In light of the BM definition by Amit and Zott (2001) and Zott and Amit (2010) this means that changes will have to affect at least the content, structure or governance of BM transactions.

As above written paragraph illustrates, literature on business model innovation (BMI), more than research on BMs, is nascent (Casadesus-Masanell and Zhu, 2013; Bucherer, Eisert and Gassmann, 2012). Mostly the focus has been on product innovation or production innovation (Casadesus-Masanell and Zhu, 2013). Attributed importance to the subject has mainly arisen from the growing interest for the internet (Santos et al., 2009; Comes and Berniker, 2008) and due to technological progress in general (Casadesus-Masanell and Zhu, 2013; Amit and Zott, 2010). In practice this technological progress also has its direct effects on BMs. It stimulates/drives BMI (Casadesus-Masanell and Ricart, 2010; 2013; Santos et al., 2009; Chesbrough and Rosenbloom, 2002; Chesbrough, 2010). Similarly, globalization (e.g. access to new markets (Santos et al., 2009; IBM, 2006; Serrat, 2012)), deregulation, (Casadesus-Masanell and Zhu, 2013; Casadesus-Masanell and Ricart, 2010) and new customer preferences drive BMI (Casadesus-Masanell and Zhu, 2013).

Though, it remains a fact BMI is triggered by internal and external forces (Demil and Lecocq, 2010; Bucherer, Eisert and Gassmann, 2012; Mitchel and Coles, 2004 (b); Pohle and Chapaman, 2006), by no means is BMI dependent on the mentioned developments (Santos et al., 2009). BMIs can very well stand alone and within this study function as the dependent variable.

Performance: market share

This study investigates the relation between BMI radicalness and market share changes. As said, BMs can be a source of innovation (Amit and Zott, 2011; Massa and Tucci, forthcoming). A BMI can even create more value than product innovation (Chesbrough, 2007, 2010; Pohle and Chapman, 2006; Comes and Berniker, 2008). Thus, BMIs can lead to competitive advantage (Teece, 2010; Chesbrough, 2010; Mitchel and Coles, 2004 (b); Massa and Tucci, forthcoming; Boons and Lüdeke-Freund, 2013; Comes and Berniker, 2008; Johnson, 2010; Demil and Lecocq, 2010; Zott and Amit, 2010) and “sustainable performance advantage”, because it is more difficult to imitate than other forms of innovation (Amit and Zott, 2010 - p. 5). In addition, it is said to be pivotal in sustaining value creation in the long-run (Achtenhagen et al., 2013). However, not all companies are capable of doing so. Identical offerings brought to the market by different BMs will result in two different outcomes Chesbrough (2010). Likewise, Baden-Fuller and Haefliger (2013) say that a business model mediates

technology and performance. But how does BMI itself affect performance? Huang et al. (2012) claim a positive relation, but have only cross sectional data. Causal inferences are therefore not so strong (Huang et al. 2012). It is interesting to show what the outcomes of BMIs are and which factors play a part in outcome differences. Existing literature has not been able to find conclusive results.

One way to look at the effects of BMI is to look at market share. Markides (2006; 2013), Koen, Bertels and Elsum (2011), Johnson, Christensen and Kagerman (2008), Markides and Charitou (2004), and Yip (2004) recognize this and state that firms look to gain market share. Revenue is the product of the price of the goods sold and the number of goods sold. It implies that the bigger your market share, the higher your revenues will be. In combination with the IBM (2006) and Pohle and Chapman (2006) studies, which show that companies engaged in BMI have a higher operating margin / cost reduction, it can be derived that an increase in market share through BMI leads to higher profits.

Unfortunately, the more conventional ways to reach that cause maybe a ten to fifteen percent increase. On the other hand, a change in the BM potentially results in doubling or tripling the market share (Yip, 2004). Note, that this doubling or tripling of the market share is a possibility. It has not been stated as a fact. This implies that there are a lot of uncertainties and perhaps is too difficult to pull off. It is riveting to find out if, on average, companies fail or not. The findings would give managers useful input in the assessment whether a BMI is worth it or not. These findings are strengthened by longitudinal data to offset the limitations of studies like Huand et al. (2012).

New entrants versus incumbents

Being the first with the BM change is not the criteria for developing an innovation. Imitation can very well still be an innovation for other business units. Casadesus-Masanell and Zhu (2013) say incumbents imitate the BMs of the new entrants (Casadesus-Masanell and Zhu, 2013). This means that they can partially skip the trial and error phase because the new entrants already show them how it works. From the incumbent’s perspective the change is then considered less radical; because new entrants proof that the new BM is successful incumbents will be less resistant towards this change. Thus it requires less effort.

In line with above standing, empirical literature indeed shows that entrepreneurial firms perform well engaging in novelty-centered (new or different transaction links, or transaction mechanisms) BMIs (Zott and Amit, 2007). Zott and Amit find that for these firms (mean age: 7 years, median age: 4.3 years) there is a positive relation for this type of innovation with market value. For efficiency-centered innovations there is not such a relation. In contrast, Brettel et al. (2012) find that a focus on efficiency is positively related to firm performance (perceived market growth and profitability). The

same applies for a focus on novelty. Interestingly they also find that a focus on efficiency works better for older firms. Although Brettel et al. (2012) do not proof that novelty-centered business models works better for young companies. It suggests that incumbents find it harder to implement radical innovations as efficiency implies more incremental innovations. Turning to product innovation literature it indeed shows that introduction of incremental product innovations by incumbents is positively related to market share (Banbury and Mitchell, 1995).

Perhaps there is another reason for incumbents not to engage in BMI sufficiently. Possibly these established firms need more proof that shows results of BMI. Although, Massa and Tucci (2013) recognize research on consequences of BMI is emerging. Current literature specifically asks for more research on the outcomes (Zott, Amit and Massa, 2011) as the study by Zott and Amit (2007) is one of the very few that does so.

Research question

First of all, fuzziness surrounding BMs is one of the shortcomings in current literature (Zott, Amit and Massa, 2011; Shafer, Smith and Linder, 2005). This study aims to add clarity. Following the suggestions made by Zott, Amit and Massa (2011) this paper focuses on the BM concept as a strategic issue explaining value creation, competitive advantage and firm performance. Furthermore it will adhere to their request to proof the empirical usefulness of the concept.

It does so by investigating the outcomes of BMI on market share development, while differentiating new entrants and established firms. Especially in light of the potential to lose market share to new entrants, it would be interesting to see whether firms are able to gain or regain market share through BMI. Although (re)gaining market share seems to be the focus of both types of firms (i.e. established and new entrants), when considering/implementing a BMI; empirical evidence on the relation between the two is absent.

This paper therefore posits following questions: Is there a positive association between BMI and market share development? And, do new entrants moderate this relation?

Hypotheses development

This study will, obviously, also draw from BMI literature in relation to performance. But, partially turns to the product innovation literature to suggest possible outcomes. Bucherer et al. (2012) say it is useful because these two types of innovation show similarities. Both innovation processes start of chaotic, but can be guided by normative processes (Bucherer et al., 2012). Next to that, the origins from the innovation are triggered by external and internal factors and face resistance by external and internal factors. Radical or incremental innovation; it accounts for both. Dedication, responsibility,

and sponsoring to make both types of innovations work are needed. Of course there are differences as well. One is that a BMI affects a company on a broader scale which usually requires restructuring (Bucherer et al., 2012).

First of all, Kleinschmidt and Cooper (1991), Deshpande et al. (1993), Baldwin and Johnson (1995) and Prajogo (2006) claim that there is a positive relation between product innovation and market share. However, there seems to be a difference between radical and incremental product innovation (Kleinschmidt and Cooper, 1991). They conclude that the relation product innovation, market share is U-shaped (considering competition differentiation on (Inter)national level). Note that Prajogo (2006) did not talk about this U-shape and distinguishes process and product innovation. He states the former has stronger influence within manufacturing companies than in service companies.

Concerning the effects of business model innovation on performance; little to no literature investigates the direct link between BMI and market share. One study, by Ucaktürk et al. (2011), states that in periods of recession BMI is the best way to increase market share and profit. However, it is unclear where this finding comes from. It is possible though to infer hypotheses from literature that talks about BMI and the effects on performance. Huang et al. (2012) investigate 189 companies (information and electronics firms). They conclude that BMI is positively associated with customer satisfaction, sales growth, overall profitability, continuous improvement, cost reduction and on-time customer delivery. Brettel, Strese and Flatten (2012) found a positive relation with perceptive performance over a period of 3 years that is. They do so in terms of, customer retainment, customer increase, customer satisfaction, value provision for customers, realizing desired growth and market share. Unfortunately, these measures are subjective. However, the findings show resemblance with Pohle and Chapman (2006). They found that over a period of 5 years that companies who focus on BMI outperform other companies without BMI focus in three areas. Those are average revenue growth, operating margin growth and historic operating margins. Similarly Lindgardt et al. (2009) finds that companies that focus on BMI score higher on average total shareholder return for periods of 3, 5 and even 10 years.

Clearly these studies show that in general BMI is positively linked to firm performance. The findings, such as increase in revenues, sales and perceptive increase in market share and customer increase leads to think that BMI leads to positive developments in market share. This thought is reinforced by the combination of the discussed product innovation literature and Lindgardt et al. (2009) who say that a BMI achieves better results than product or process innovations. Therefore this study poses that:

Hypothesis 1: Business model innovation is positively related to market share growth.

This seems a solid statement. However, not many companies are engaged in BMI (Johnson, Christensen and Kagerman, 2008)³. Koen, Bertels and Elsum (2011), Chesbrough (2010) offer reasoning for that. They say that especially for incumbents, BMI is hard to realize. The literature mentions different barriers. Barriers can be the lack of capability, lack of willingness (Chesbrough, 2010) or conflicts between a BMI and the old BM (Chesbrough, 2010; Markides and Charitou, 2004). Santos et al. (2009) clarify this by saying that BMI requires new skills and new knowledge. Koen, Bertels and Elsum (2011) as well, state the requirement of new knowledge as a clarification. It means that other knowledge and skills can become irrelevant and thus implies that BMI is destructive for current competences (Santos et al., 2009). This refers back to the conflict situation between the new and current BM. Logically, it can be derived that for entrants it is much easier to enter an industry with a BMI as there will be no conflicts present with their old BM.

Related with preceding statements is the fact that incumbents are restricted by path dependencies and inertia (Zott and Amit, 2007). Hannan and Freeman (1984) argue that inertia increases with size and age. Also past performance is said to be positively associated with inertia (Miller and Chen, 1994).

Another obstacle for incumbents is that the process of finding a good new BM is one of trial and error (McGrath, 2010; Sosna, Trevinyo-Rodriguez and Velamuri, 2010; Sosna et al., 2010; Johnson, Christensen and Kagerman, 2008; Velu and Stiles, 2013; Chesbrough, 2010). These firms might not have an incentive system to adopt this kind of attitude (McGrath, 2010). Although BMI is said to be less costly than product innovation (Amit and Zott, 2010), making mistakes can be costly and might thus prevent these firms even from trying.

In contrast, new entrants do not seem to be bothered by trial and error stages as much as incumbents (Johnson, Christensen and Kagerman, 2008). Literature shows that they can disrupt markets/industries (Markides, 2006; Johnson, Christensen and Kagerman, 2008) and are able to capture market share (Koen, Bertels and Elsum, 2011; Johnson, Christensen and Kagerman, 2008; Markides and Charitou, 2004). Examples are Xerox’ spinoff 3Com (Chesbrough, 2010), Easy Jet, Ebay (Teece, 2010), Wal-Mart (Johnson, Christensen and Kagerman, 2008) and Apple with its iTunes and iPod (Amit and Zott, 2010). This leads to hypothesize that:

Hypothesis 2: New entrants moderate the relation between business model innovation and market share development.

³ “a recent American Management Association study determined that no more than 10% of innovation investment at global companies is focused on developing new business models.” (Johnson, Christensen and Kagerman, 2008, p. 60)

Model

Methodology

This study chooses a deductive approach to investigate its hypotheses. It collects data on business model innovation by means of an online survey. The survey is constructed to measure not only the variables for this study, but 4 other studies as well.

The aim for this specific study is to collect longitudinal data. Cross-sectional data will not be the focus of the survey because it does not suffice as the research question addresses change (i.e. market share development). Cross-sectional data is not suitable for that (Bono and McNamara, 2011). Furthermore, longitudinal data allows the establishment of causal inferences better (Huang et al., 2012). Plus, important for current study; it helps the development of models of change and growth (Kimberly, 1976). Following paragraphs will explain how data is collected, plus the reasoning, and for what purposes specific variables and corresponding measures help to realize the study’s aims.

Type of survey

The strengths of online surveys are that it allows quickly sending out surveys and gathering data from all over the world (given an internet connection). It also allows a much easier follow-up. Next to that, respondents can participate at their convenience in contrast to, for example, telephone surveys. It allows participants also to stop the process and come back to it later. Also, filling in the answers or

Business Model Innovation Market Share Development New entrants Control variables:

- Business unit age - Resource abundance - Environmental dynamism - Sector

processing the answers is much easier than paper based surveys. Plus, people can send back the survey with only a click of the mouse. This is less troublesome than sending back an envelope, which may result in a less prompt return (Kwak and Radler, 2002). It also reduces costs on both sides (Evans and Mathur, 2006). Besides advantages, there are also weaknesses attached to online surveys (Evans and Mathur, 2006).

First of all there is the risk that people respond for whom the survey is not meant (Evans and Mathur, 2006). The makers of current survey blocked that by requiring respondents to fill in a password. Only people with the password can access the survey. To a certain level this will prevent undesirable responses. Another downside is that, as the developers of this study directly experienced, sometimes it can be very difficult to get the software logic behind the survey correct. This requires a lot of valuable time that could have been used for data collection itself.

General concerns with surveys exist as well. The researcher is familiar with the language used, but the respondent may not be. Extreme clarity needs to be assured for the respondent (Evans and Mathur, 2006). Tests with laymen were needed for optimum question tweaking. In line with this issue, there is the trouble that in-depth questions are difficult to gain (Evans and Mathur, 2006). A fourth possible issue is about privacy (Evans and Mathur, 2006). To take away any doubt, the researchers stated three times that all data is treated confidentially and cannot be linked back to the individual companies.

Sample

To test the hypotheses posited in this paper, data from companies based in different parts of the world is collected (table 1). Cases are approached based upon proximity within the professional and social network of 5 researchers. The joint collection of data is chosen to make up for low response rates for online surveys. Using the networks of the researchers is done because access is easier and it increases the willingness to participate (as it is the case with convenience sampling) (e.g. Teddlie and Yu, 2007). It offsets a lower response rate to internet surveys in comparison to mail surveys, as Kwak and Radler (2002) claim is the case. The downside is that it causes too much bias (Lucas, 2014) and thus the population is not well represented. The reason for this bias can be due to volunteers (Lucas, 2014) or too much similarity within the networks.

One way to possibly offset the volunteer bias is by requesting members of LinkedIn fora to participate. The LinkedIn group New Product Development, Innovation and Growth was one of them. This group contained at that specific point in time 15 315 members. It resulted in an unknown number of responses since they did not receive a separate link. The same applies to another group:

Business model, innovation and agile strategy (2463 members). However, no comments have been placed under the request. Surprisingly though, in the days after a reminder had been posted the researcher’s profile received 3 extra profile views that were likely to be connected to the request. Still it is assumed that the LinkedIn requests resulted in no responses. And thus, the bias has not been offset.

All respondents are approached either by face-to-face contact, email, Facebook messages or LinkedIn messages. These messages contain a personalized message accompanied by the official invitational letter (appendix 7). After a couple of days (varying between 5 and 10) a reminder is sent to invitees. Respondents are also asked to refer other to potential other participants. There is only knowledge from one participant that forwarded the survey. Further results on this enquiry are not known. In total the 5 researchers send out 257 invitations. In the end, 54 respondents filled in the survey. This is a response rate of 21%. It appears a good percentage. An explanation is found in the idea that people from the researchers’ social circles display a favorable attitude towards a specific researcher. This willingness is illustrated by a quote of one respondent: “we all know what kind of effort it takes Table 1: Sample

Sector # Type of BU # Country # Age Employees

Service 33 Incumbent 31 Netherlands 29 Mean 29.93 Mean 326.76

Manufacturing 8 New entrant 10 Germany 4 Medain 22 Medain 30

Total 41 USA 3 Mode 8* Mode 2*

Competing on France 2 Min 1 Min 1

Local level 3 Columbia 1 Max 87 Max 2300

National level 25 Finland 1 1st quarter 5.50 1st quarter 7

International level 13 Sweden 1 3rd quarter 49.50 3rd quarter 215

Total 41 Total 41 Total 41 *Multiple modes present.

to conduct this type of studies”. Possible other factors influencing the willingness to respond to surveys are: the subject, the survey provider’s reputation, and the sponsoring corporation´s reputation (Fang et al., 2012). Drawing from Fang et al. (2012) possible explanations for lower response rates are a low reputation of the UvA (the sponsor), the researcher (the provider) or both.

Measuring:

The measures in the survey are answered through self-administration. This means reliability is in the hands of the respondent him- or herself. In order for the researches to know to what degree to ‘trust’ respondents’ expertise, two questions about the importance of their position in relation to the business unit’s strategy are asked.

-The independent variable

Business Model Innovation. To measure the degree radicalness of BMI, this study chooses an approach based upon the Boston Consulting Group (BCG) business model framework. This framework is adopted by Kiron et al. (2013). With the questions from the model they investigate how the business models of companies have changed. Lindgardt et al. (2009) recognize two elements as the essence of a BM. That is the operating model and the value proposition. Each of these two elements is built up by three sub elements.

Table 2: Business model elements (Lindgardt et al., 2009)

Element Sub elements Questions

Value Proposition Target Segment(s) -Which customers do we choose to serve? -Which of their needs do we seek to address? Product or Service Offering -What are we offering to the customers to satisfy

their needs?

Revenue Model -How are we compensated for our offering? Operating Model Value Chain -How are we configured to deliver on customer

demand?

-What do we do in-house? -What do we outsource?

Cost Model -How do we configure our assets and costs to deliver on our value proposition profitably? Organization -How do we deploy and develop our people to

This typology shows resemblance with the constructs of Chesbrough and Roosenbloom (2002); Demil and Lecocq (2010); Baden-Fuller and Mangematin (2013); Boons and Ludeke-Freund (2013); Chesbrough (2007) and George and Bock (2011). It also covers the definition by Amit and Zott (2001). The value proposition covers the content of transactions, while the operating model covers the structure and governance of transactions. Achtenhagen et al. (2013) very well translate these issues into possible business model changes (appendix 1). From this, subsequent questions to measure the degree of BMI are derived. Questions 1 and 2 cover the value proposition and thus the content of transactions. Questions 3, 4, 5 and 6 cover the operating model and thus the governance and structure of transactions.

1. We have significantly changed our target segment…

2. We have significantly changed our product or service offering… 3. We have significantly changed our value chain…

4. We have significantly changed our revenue model… 5. We have significantly changed our cost structure…

6. We have significantly changed our way to deploy and develop our people to sustain and enhance our competitive position…

Range of possible answers goes from: ‘strongly disagree’ to ‘strongly agree’.

Post data collection analysis showed a Cronbach’s alpha of 0.846 which is higher than 0.7. Based on Nunnally (1978) the scale is considered reliable. All questions are answered per year separately (from 2009 until 2013) on a 7-point Likert scale. The higher the score, the more that specific part of the business model has changed. It allows exactly matching the market share increase or decrease with a BMI in a specific year. Concerning the 7 Likert scale; it does not seem to matter a lot how many points are used when talking in terms of validity and reliability (Matell and Jacoby, 1971). However, the more points, the more precise the intensity of the direction will be. At the same time it would be also more of a burden for the respondent to answer the questions (Matell and Jacoby, 1971). A 7 points scale seems reasonable. The choices are not too many and the degree of innovativeness can be more precisely measured than with 5 points.

-The dependent variable

Market Share Development. Market share development is defined as a change in market share in percentages. Respondents are able to indicate their market share from 2009 to 2013 with a lever. If it concerns a new entrant, then it will be based on the first years of operations. This means that the maximum of operational years for a new entrant is 5 years. Note, the market share is asked based on

the level of competition. That is local, national or international. Respondents had to indicate it themselves. The distinction is made to have more data and be better able to compare relative market share change. For example, for an Amsterdam based fitness gym it is impossible to indicate its national market share percentage. It would then mean that that data cannot be analyzed. Note that, the percentage of market share is asked. This means the percentage of change has to be calculated after the data collection.

-The moderator

New Entrant. Incumbents are distinguished from new entrants by the number of years in a specific industry. A business unit is defined as incumbent when it is 6 years or older. If a business unit is 5 years or younger it is a new entrant. This study does so because this seems a cut-off point between failure and long-term success. Cooper and Dunkelberg (1981) for example say that only a little less than 30% make it past the first 4 to 5 years. Van de Ven et al. (1984) state something similar and say that 25% reach 6 years or more (sample size of 12). In the eyes of Peña (2006) companies in this stage (after the initial 3 to 4 years of gestation) are in the consolidation phase. Gibb (1990) and Bennet (1989) actually identify these first three years as the critical years (from Littunen et al., 1998). It means that after that point they are considered good performers, even though they have known severe struggles before (Peña, 2006). Astebro and Bernhardt (2003) took the initial 4 years of a company to measure survival rates. So, it appears justifiable to call a business unit an incumbent or established when it is 6 years older.

-The control variables

The control variables make the study stronger in isolating and measuring the effect of business model innovation on market share development.

Business Unit Age. Equal to Brettel, Strese and Flatten (2012) the business unit’s age is a control variable. It is used because it influences performance (Brettel, Strese and Flatten, 2012). Age tells something about the experience level. Older units have had more time to develop their business and gain market share (Zaheer and Bell, 2005). Next to that, they could be better able to increase their market share because they have more experience in doing business as Autio et al. (2000) similarly reason. So to analyze possible effects, respondents are asked to fill in the year in which the business unit has been founded. Based upon the given answer the respondents were either referred to questions relating to incumbents or new entrants. All business units founded in 2008 or before are considered incumbents.

Business Unit Size: Resource Abundance. Aspara et al. (2010) find a difference in “average profitable growth” between large and small firms (differentiated by sales) that focus on BMI. Business unit size tells something about the unit’s resource base. It might influence performance (Brettel, Strese and Flatten, 2012). The size of the resource base in this case expressed as having resource abundance. It is measured on a scale from 1 to 7 and contains four items. Four items, adopted from Miller and Friesen (1982), measure the abundance in capital, material supplies, managerial talent and skilled labor. It is reasoned that the higher access to good employees, the higher the potential for value creation (Rogers, 2004) and thus its performance (Zott and Amit, 2007). Also, there is a positive relation between size (specifically employees) and survival rates (5 year survival period: small 67%, large 75%) of companies in their formative years (Agarwel and Audretsch, 1999). Skilled labor influences R&D which can stimulate economic growth (Shefer and Frenkel, 2005). It also has a direct on innovativeness (Rogers, 2004) and subsequently on market share (Bell and Zaheer, 2005).

Calculating the Cronbach’s alpha does not show high reliability. The value is 0.521. Pallant (2004) states that in these cases (less than 10 items) the mean inter-item correlation represents reliability better. The score of 0.216, according to Briggs and Cheek (1986) is good as it is between 0.2 and 0.4. Industry Sector. Brettel, Strese and Flatten (2012) investigate performance in relation the type of Also there seems to be a difference to what degree a service company benefits from an innovation (product or process) in comparison to a manufacturing company (Prajogo, 2006). Therefore this study adopts sector as a control variable. Respondents indicate whether they are focused on servicing or manufacturing.

Environment. Environmental dynamism has a direct effect on performance (McArthur and Nystrom, 1991). It makes a difference in capturing market share in an industry with a lot of uncertainty or not. Cooper and Smith for example say that when there is more uncertainty, competition is fiercer and requires more effort (Cooper and Smith, 1992). This means that it would be more difficult to gain market share. Also, when barriers to entry, capital intensity and concentration are high in certain industries, large firms seem to be more innovative. While small firms tend to be more innovative in less mature industries that are innovative itself, use a lot of skilled labor and contain relatively a high percentage of large firms (Acs and Audretsch, 1987). In return, this level of innovativeness can also affect market share if that is translated into a relative product advantage (Robinson, 1990) It especially influences market share in an industry where small firm play a key role in innovative activities (van Dijk, 2000).

Because these environmental circumstances can influence performance (Voelpel et al., 2004) and thus market share development, the survey asks for environmental dynamics. This scale is built up by 5 items (table 3) adopted from Miller and Friesen (1982). The corresponding Cronbach´s alpha of this scale is 0.742. It is above 0.7 but to be sure this study refers to Pallant (2004) and Briggs and Cheek (1986) again and concludes the mean inter-item correlation score of 0.369 is good.

Table 3: Environmental dynamism items from Miller and Friesen (1982 - p. 17) Our firm must rarely change its marketing

practices to keep up with the market and competitors.

Scale

Our firm must change its marketing practices extremely frequently (e.g. semi-annually)

The rate at which products/services are getting obsolete in the industry is very slow

(e.g. basic metal like copper).

Scale

The rate of obsolescence is very high )as in some fashion goods and semi-conductors),

Actions of competitors are quite easy to

predict (as in some primary industries). Scale

Actions of competitors are unpredictable.

Demand and consumer tastes are fairly easy

to forecast (e.g. for milk companies). Scale

Demand and tastes are almost unpredictable (e.g. high fashion goods).

The production/service technology is not subject to very much change and is well

established (e.g. in steel production).

Scale

The modes of production/service change often in a major way (e.g. advanced

electronic components).

Data handling

In total 104 respondents opened the survey. From them, 87 actually started answering questions. In the end there are 54 respondents that completed the survey. Reasons for this dropout may be that the survey is considered too long, or too difficult. Unfortunately 13 more cases need to be dropped. One of the responses is excluded because of very unlikely answers. The twelve other cases are deleted because no details on the market share percentage are given. For the remaining answers, variables and coding were checked.

First of all, due to wrong coding 3 years are subtracted from the indicated amount of years working at the business unit and company. Second, values for one question are recoded. Instead of a scale

from 8 to 14, they are recoded from 1 to 7. After rechecking, data entry points on new entrants, for the years in which they did not yet exist, is deleted.

The next step is to calculate to which degree the companies changed their business model in each year. So, there are 6 items that measure BMI in each year. Those are added up and divided by 6. This gives a score between 1 and 7. Next to that, an average score for the resource abundance and the degree competitiveness is calculated according to the same method. For market share the percentage of growth from to year is calculated. When a company would grow from 0.XX, or nothing, to 5%, the growth percentage was considered as 5.

The final step is to restructure the data from a wide table to a long table. It allows analyzing panel data and thus matching business model changes in one year with market share increases or decreases in the same, or perhaps another year.

Description data set:

In total 41 useful surveys are collected. Table 1 shows some of the business unit characteristics as indicated by the respondents. But first some information on respondents themselves is given.

-Respondents

Since the answers to the questions are mainly subjective, it is good to know to what degree the respondents actually are able to be knowledgeable on the questions. Therefore the degree of strategic influence is measured, as well as the degree to which they know about strategic decisions. The first showed an average score of 4.95, and the latter 6.27. Based on a scale from 1 to 7 it means that the respondent is very likely to know about business model changes, and for a large part contributed to the decision. Therefore, it can be assumed that the answers can be trusted. More details can be found in appendix 2.

-The business units

Subsequent table 4 shows statistics on the variables and thus of the business units and its environment. More details in appendix 3.

Table 4: Business Unit Characteristics and Industries Average Resource Abundance Mean score Average Competitiveness Mean score Industry # Capital 4.24 Changing Marketing Practices 4.34 Costruction 3

Skilled Labor 4.61 Product/Service

Obsolescence 3.83

Manufacturing 6

Material Supplies 4.49 Competitor’s

Predictability 3.83

Transportation & Public Utilities

3

Managerial Talent 4.49 Demand and

Taste Forecasting 3.76 Retail Trade 4 Changing Production Technology 3.98 Finance, Insurance, Real Estate 3 Agriculture, Forestry and Fishing 0 Wholesale trade 3 Services 18 Public Administration 1 Mining 0 Average Overall Abundance 4.46 Average Overall Competitiveness 3.95 Total 41

Analysis

Since this study collects panel data, the ´conventional´ correlation tests do not apply. Therefore the generalized estimating equations (GEE) method is chosen. It allows relate changes in the independent variable to changes in dependent variable at different points in time. In order to perform a GEE, like for other parametric tests, normality needs to be assumed (Twisk, 2003). Normality tests show the distribution is positively skewed and also has a positive kurtosis with most values just above zero. Plus, using other distributions within the GEE is not possible as the data also contains negative values, which are not accepted. Furthermore, the histograms show that the tails are more or less symmetrically distributed, therefore normality is assumed.

Note that as the observations in the same subject are not independent, a correlation correction needs to be applied (Twisk, 2003). In this case the most applicable structure seems to be the exchangeable one. It assumes that all within-subject correlations are equal (Twisk, 2003). Although it is best to assume the right correlation structure, differences in outcomes for the different structures are little and not considered to be very dangerous (Twisk, 2003). To check this statement test 4.1

(see table 5) is performed with an unstructured correlation structure as well. The results are exactly the same.

Since this study has collected panel data, time is considered an important factor in our data analyses. Besides comparing a business model innovation in time t with a market share change from time t to time t+1, other combinations are tested as well. One group of tests looks at possible effects taking place after a certain period. The other group of test looks at possible effects taking place over a certain period. Subsequent table 5 shows which test belongs to what model. You may go back to this table for enhanced clarity.

Table 5: Model + test number

No

lag 1 year lag 2 year lag 3 year lag period 1 year period 2 year period 3 year No interaction, no CVs 1.1 1.2 1.3 1.4 10.2 10.3 10.4 New entrant interaction, no CVs 2.1 2.2 2.3 2.4 11.2 11.3 11.4 No interaction, with CVs (no age) 3.1 3.2 3.3 3.4 12.2 12.3 12.4 New entrant interaction, with CVs 4.1 5.2 5.3 4.4 13.2 13.3 13.4 New entrant interaction, with CVs

(no age) 5.1 5.2 5.3 5.4 14.2 14.3 14.4

Sector interaction, with CVs 6.1 6.2 6.3 6.4 15.2 15.3 15.4 Resource abundance interaction,

with CVs 7.1 7.2 7.3 7.4 16.2 16.3 16.4

Age interaction, with CVs 8.1 8.2 8.3 8.4 17.2 17.3 17.4 No interaction, with CVs (no age) 9.1 9.2 9.3 9.4 18.2 18.3 18.4

Results

Before getting to the analysis, subsequent table 6 contains descriptions of the variables within the model. As can be seen the Kurtosis for the market share is indicated a highly peaked distribution with values skewed to the right.

Table 6: Variable descriptions BU age Resource abundance Environmental dynanism Degree of BMI for all years

and companies

MS change over all years

and companies N Valid 41 41 41 185 182 Missing 0 0 0 20 23 Mean 28.93 3,99 4.41 3.94 7.04 Median 22.00 4,00 4.50 4.00 0.00 Mode 8a 3,00 4.75 4.00 0 Std. Deviation 27.07 1.20 1.08 1.23 33.93 Variance 732.77 1.44 1.17 1.51 1151.33

Skewness 0.84 0.14 0.27 -0.08 3.45 Std. Error of Skewness 0.37 0.37 0.37 0.18 0.18 Kurtosis -0.52 -0.64 0.22 -0.65 32.35 Std. Error of Kurtosis 0.72 0.72 0.72 0.36 0.36 Range 86 4.60 4.75 5.33 400 Minimum 1 1.60 2.25 1.50 -100 Maximum 87 6.20 7.00 6.83 300 Percentiles 25 5.50 3.00 3.75 3.17 ,00 50 22.00 4.00 4.50 4.00 ,00 75 49.50 4.80 5.00 4.83 10.47

Hypotheses tests*:

-No lag: the basic model

First of all, the basic effects without any control variables are tested. It starts with the relation between the radicalness of business model innovation and market share changes. This test (test 1.1) demonstrates a non-significant coefficient score of 2.279. A second test (test 2.1) with the moderation effect shows only significance at 90% confidence. So, both tests support neither of the hypotheses.

The next step repeats the tests but now with control variables. Table 7 contains the correlation matrix for the parameter estimates of the variables. All in all the outcomes of the test with control variables neither show support for the hypotheses. Both tests (3.1 and 4.1) give a high p-value. To see whether this could be explained by the inclusion of the control variable “age”, these two tests are executed without age as a control variable. From test 4.1, test 5.1 results and shows indeed stronger results, but still non-significant. The other evaluation (test 9.1), resulting from 3.1, shows not much different outcomes. Again, H1 and H2 are not supported.

-No lag: other variables

However, evaluation 9.1 does display a significant coefficient for the control variable resource abundance for an alpha of 0.05. This is similar as in 3.1. For 4.1 and 5.1 it is significant for an alpha of 0.10. Therefore, resource abundance is tested as a moderator (test 7.1). The corresponding coefficient is 1.178. The p-value is 0.055. Strictly seen this result does not give support to a potential hypothesis that resource abundance serves as a moderator at an alpha of 0.05. Test 6.1 also changes the role of one control variable to that of a moderator. This study does so because this control

variable, sector, shows a high coefficients. Though significant they are not. The result of this test is small and insignificant.

As a fourth possible moderator, the control variable age is chosen. At a 10% significance level it provides small negative coefficients. This specific test, test 8.1, has a p-value of 0.055 and a coefficient of -0.032. It is a very small indirect support for H2 since the variable age and the binary distinction between incumbents and age overlap. This is because age serves as the input variable to create this binary typology of companies (incumbent – new entrant).

Table 7: Correlation matrix of parameter estimates. (Intercept) Services

sector Manufacturing sector BU age abundance Resource Environmental dynanism BMI entrant New Incum-bent (Intercept) 1.00

Sector:

Services .108 1.00 Sector

Manufacturing Red. Red. Red. Business unit age -.485 -.182 Red. 1.00 Resource abundance -.553 .233 Red. -.140 1.00 Environmental dynanism -.697 -.320 Red. .466 .150 1.00 BMI -.022 -.303 Red. -.033 -.451 -.242 1.00

New entrant -.002 -.260 Red. .513 -.344 .183 -.025 1.00

Incumbent Red. Red. Red. Red. Red. Red. Red. Red. Red.

Red. means redundant.

-With lag: the basic model

Seeing that the results for no lag give no significant values, this study looks into the situations where the effect potentially takes place after a certain period too. To accomplish that, it sets year t equal to market share change from t+1 to t+2, from t+2 to t+3, and from t+3 to t+4. The situations will respectively be called; a lag of 1, 2, and 3 years.

Starting with a lag of on year, all tests show no significant results. However, for a lag 2 and 3 years that changes. With no control variables (tests 2.3 and 2.4) both coefficients are significant and are

higher than 3.5. Adding the control variables (tests 4.3, 4.4, and 5.4) they remain significant. Test 5.3, with a p-value of 0.056 is only just not significant anymore. Therefore, for a lag 2 and 3 years hypothesis 2 is supported.

The results for hypothesis 1 are less convincing. Only with a lag of three years and without the new entrant moderation, the results display significance. So, only the results of test 1.4, 3.4, 8.4, and 9.4 support H1. At all these tests, the control variable age is absent.

-With lag: other variables

Similarly as the tests without lag, the moderation effect of certain control variables in the models with lag is tested. Age for example is an interesting one. No significant results show up and the coefficients are small. The moderator tests 8.2, 8.3, and 8.4 show no significant results. This is somewhat in contrast to 8.1, which had a p-value of 0.055. For resource abundance as a moderator only test result 7.2 is significant. For a 3 year lag, there is no significant result. There is even a small negative coefficient (test 7.4). Test 7.3 shows almost the same result. It is significant at the 10% level. A potential hypothesis for this interaction effect would thus only be very moderately supported. For sector interaction almost no support for a potential hypothesis is found. Only test 6.3 shows significance when applying an alpha of 0.1. And in contrast to when sector is used as a control variable, the coefficient is now low. Weak results are also obtained by the moderator age. There is no significance observed.

-Periodic effect: basic model

For now this study looked into a lagged effect of market share change. Following paragraphs deal with the overall effect over a bigger time period. They compare a change of market share over the years t to t+2, t to t+3, and t to t+4 in relation to a given business model innovation in t. Besides this, nothing changes. Results are stated below. Please note that the direct effects are not stated here, because they have already been discussed. The study directly looks at t+2 and so forth.

None of the 27 new test results supports H1. The interaction effect does receive some support. The main models for the moderation effect (test 13.3 and 14.3) show significant results. Tests 10.2, 10.3, and 10.4 do so too. The significance and coefficients are higher than the “lag tests”. So hypothesis 2 is supported, but only when you look at market share change over a period t to t+3.

-Periodic effect: other variables

The very first thing that pops out of the results is that at every test resource abundance shows significant coefficients. It also applies to the moderator test (test 17). As control variable it has

coefficients around 12.5 and p-values from 0.017 to 0.049. As moderator the coefficients are around 4 (tests 16.2, 16.3, and 16.4). The results would support a potential moderator hypothesis for resource abundance.

For sector it would also be supported, but only by 15.3 and 15.4. As a control variable the outcomes do not display so much significance. For an alpha of 0.05 or lower, tests 17.3, 17.4, 18.3, and 18.4 satisfy. The remainders have only significance with 90% certainty or not at all.

Lastly there is the control variable age. Evaluations 12.3, 12.4, 13.3, 15.3, 15.4, 16.3, and 16.4 all have negative significant coefficients. The same applies to the moderator effect of age for over a 3 and 4 year period. The coefficients are negative at an alpha of 0.01. This indirectly could support the H2, as is mentioned before.

Discussion

This study hypothesized that business model innovation is positively associated with market share development. It also hypothesized that new entrants serve as a moderator in this relation. In first instance an innovation in year t was compared with market share change from year t to t+1. Later on, it tested models incorporating the effect after a certain period, and models measuring the effect over a certain period. Preceding section outlays the GEE test results. Current section discusses the implications of the outcomes.

Despite the lack of data, this study does make valuable contributions to the literature. First of all, it develops a model to quantify degree of radicalness of the BMI. Secondly, it employs a quantitative method. Little empirical research has been done on the subject so far. It proves the empirical usefulness of the business model concept by evaluating its consequences as suggested by Zott, Amit and Massa (2011).

Next to that, this study takes away some opaqueness surrounding the subject by deliberately using the concept to explain strategic issues, such as value creation, competitive advantage and firm performance. Plus, by choosing the BM definition of Amit and Zott (2001) it adds consistency to the literature.

No lag: the basic model

In contrast to Yip’s (2004) claims, this study failed to proof a positive association between business model innovation and market share change. Different related reasons are applicable. First of all, the

implementation of the BMI could have only finished in the last month of the year. In that case it is not realistic to expect an immediate effect on market share.

Secondly, after implementation it is possible that the business unit needs time to make it its own. First of all, as Chesbrough and Rosenbloom (2002) and Chesbrough (2007) touch upon, people try to resist change. It takes time to overcome negative attitudes or inertia. Even when change is accepted, the business unit has to go through learning stages to benefit from it (Sosna et al., 2010). Andries and Debackere (2007) confirm a learning effect is present with implementations of business model innovations.

Similarly, Bock et al. (2012) notice certain practices first need to be unlearned. Especially if the business model innovation entails new ways to deploy and develop people it may require a long time before new practices or processes are learned.

Equivalently, changing your target segment might require the business unit to completely start from scratch again with customer acquisition and other related issues. Answers to questions like “how are we going to reach our new target customers?” take time to implement.

These explanations seem plausible; especially since this study measures business model innovation radicalness. It does so by giving a combined score from 1 to 7 based on 6 items. A very high score means that extreme changes have taken place in the majority of the items. This would mean that the more radical the innovation is, the later the change will have an effect on market share due to learning effects.

Another type of learning is going through trial and error phases during BMI implementation. Experimentation is required to get the business model right (Chesbrough, 2010; Teece, 2010; McGrath, 2010). In theory the business model might work very well, but practice is different. Business model innovations thus require trial and error (Desyllas and Sako, 2013). Even a copied business model needs to be adapted (McGracth and MacMillen, 2000) (from Zott and Amit, 2007). It is a possible explanation for why no moderation effect for new entrants is found in the no lag model. They may have skipped the internal resistance phase, but what is left is learn how to work the business model in the market.

All in all, for a model without lag, business model innovation is not proven to be associated with market share growth. It is neither proven that new entrants function as a moderator in this relation. Therefore, under current discussed circumstances, H1 and H2 are rejected.

No lag: other variables

Originally resource abundance was not intended as a control variable. The chosen control variables were revenue and number of employees. Unfortunately, the majority of all respondents did not fill that in. Therefore it does not allow work with that control variable properly. As an alternative resource abundance is chosen because it contains the item capital. Abundance in capital implies the relative size of the company. Next to that, the scale contains the item managerial talent and skilled labor. Again abundance in the two indicates relative size. Therefore the control variable number of employees is replaced by resource abundance as well.

Resource abundance shows some significant results as a control variable. It raises questions like: does resource abundance strengthen the ability to gain market share through business model innovation; or is resource abundance even a prerequisite to gain market share through business model innovation? Therefore this study also checks its status as a moderator.

The outcome is too weak to state so. At the same time if it would be significant, it is not entirely in line with the reasoning that the business unit has to go through a learning phase. That means if that is the case then either the theorizing about the learning effect is not correct, or learning is stimulated by abundance in resources. Remember that the scale is made up by capital, skilled labor, material supplies, and managerial talent. So, it is not beyond imagination that if the labor forces or the talented managers are of a high level, some learning steps can be skipped. Cognitive abilities for example are important factors in learning (Bachmann, 1985, from Morrison and Brantner, 1992). In that sense, lack of talent inhibits progress and lengthens the learning phase.

Another issue the result touches upon: Since resource abundance does not seem to function as a moderator within model 7.1, it implies, depending on which items exactly score high, that not a lot of capital has to be available for the introduction of a business model innovation. This would confirm what Amit and Zott (2010) say in their paper.

Next to resource abundance, the sector variable is tested as a moderator. The foremost reason is because the control variable shows coefficient values of around 4. This seems to indicate a strong effect. However, the evaluation of model 6.1 implies a business model equally works for service and manufacturing firms. However, this study does not make a difference in what kind of innovation has been implemented. It can for example be that the manufacturing sector and service sector are equally focused on innovation but that they choose different areas. Maybe for manufacturers cost structure innovations work better, while for service providers, revenue model innovations do.

Thirdly, business unit age is evaluated as a moderator because it overlaps a little with the binary distinction between new entrants and incumbents. However, the results imply that there is not much difference between older and younger firms. It implies that BMI is equally easy or difficult to perform for old and young forms. It contrasts with Zott and Amit (2007) that say incumbents are hindered by inertia. These specific results, however, do not confirm that. A post hoc independent-samples t-test shows no significant difference between incumbents (M= 3.89, SD= 1.21) and new entrants (M= 4.23, SD= 1.31; t (183)= 1.39, p= .167) for the average innovation radicalness. The eta squared score is .01. It means that the magnitude of difference according to Cohen (1988) is small. So since both types of firms seem to innovate to the same degree, inertia does not seem to make the difference. Perhaps it means that the lack experience in business of new entrants is just as strong as the inertia in incumbents.

With lag: the basic model

Because of the potential presence of a learning effect, the same tests are performed, but then for an effect taking place after 1, 2, and 3 years. The 1 year lag tests show no different results than without lag. For a lag of 2 and 3 year though, that changes. New entrants seem to be better able to capture market share through business model innovation then incumbents. It signals that indeed the reasoning about the learning phase can be true. However, it is a learning phase on another level. Since they are new to the market it is more likely that the learning is meant to understand the market better rather than the business model (McGracth and MacMillen, 2000 in; Zott and Amit, 2007). Interestingly enough, once the moderator effect is removed (tests 3.4, 8.4 and 9.4) explanatory value gets transferred to a general positive association between business model innovation and market share change. It indicates that only after 3 years the business model innovation could have effect for all type of firms.

All in all it can be said that hypothesis 2 for the main model (tests 4) is accepted for a lag of 2 and 3 years. With a lag of 3 years the results show that for each point in radicalness of innovation in, for example, 2009, the market share will grow with 4% from 2012 to 2013.

Hypothesis 1 can only be accepted after 3 years when the moderator “new entrant” is removed. A plausible explanation for failing to proof H1 is that the results for incumbents are so disappointing that only new entrants are able to achieve market share growth through business model innovation. Or, incumbents simply just face too much competition from new entrants and thus see their market share deteriorate, no matter the counter moves. Maybe it matters in which kind of industry incumbents apply BMI. Perhaps it is better for them to apply it only in industries without new entrants.

With lag: other variables

Getting back to the other variables for tests with lag, it is seen that resource abundance scores significantly at an alpha of 0.1 except for a lag of 3 years. It is a small indication that certain abundance might help in realizing market share growth. The negative BMI coefficient affirms that a bit more. It makes sense to say that abundance in resources provides more power and tools to gain market share (e.g. Rogers, 2004). It is what the outcomes for 7.2 (p-value 0.041) and 7.3 (p-value 0.058) show.

Explanatory value is derived from test 7.2, 7.3, and 7.4. The resource abundance coefficient in 7.2 goes from 1.372 to 0.978 in 7.3, and to -0.007 in 7.4. It could indicate that once the business model innovation has emerged into the minds and behavior of the employees (i.e. the learning phase is over), abundance in for example labor skills or managerial talent is not needed anymore, or has no effect anymore on gaining market share through the innovation.

Above discussion triggers an intriguing question. That is: why would resource abundance not be of use anymore to realize growth through business model innovation? Maybe the answer can be found in the environmental dynamism scale. At the moment resource abundance does not play a significant role anymore, the environmental dynamism coefficient jumps up more than 2 integers. This occurs in all models going from a 2 to 3 year lag. A possible explanation by environmental dynamism as moderator is post hoc not found. If this moderation effect would be significant, this study might have found evidence that the S-curve of product innovation maybe also applies to business model innovations. This reasoning is in accordance with Achtenhagen et al. (2013) and Mitchel and Coles (2003) who say that a company needs to keep on changing their business model in order to sustain their performance.

For sector and age as moderators, the same line of reasoning applies more or less as stated under “No lag: other variables”.

Periodic effect: basic model

As mentioned, H1 is not supported under these circumstances. It is difficult to withdraw an explanation from current data set. However, the last decade is marked by economic downturn and recessions. This can be a factor. It could have exerted a negative influence. If this study’s sample would have been bigger, it would be less a problem. Because in that case the chance that someone’s loss is compensated by someone else’s gain, is bigger. Still the data could be skewed because of course bankrupt companies are usually not incorporated in the surveys. Perhaps they went bankrupt due to a badly designed business model innovation.