Heterogeneity of Value Creation Processes: A Taxonomy of Social Enterprise Business Models

Abstract

Taking a business model perspective, this study investigates the heterogeneity among the value creation process of social enterprises. We employ a dataset of 437 Dutch social enterprises to conduct a cluster analysis, resulting in a taxonomy of seven distinctive business models. It goes beyond existing classifications by more comprehensively capturing value creation using 25 variables. Therefore, it opens up opportunities for further research through more nuanced insights into the core organisational processes of social enterprises. The taxonomy can also assist practitioners and policymakers in understanding the implications for different social enterprises when making decisions.

Keywords:

• Social enterprise business models
• social enterprises
• business model perspective
• taxonomy
• cluster analysis

Introduction

With their goal of positively impacting society, social enterprises are emerging globally to tackle salient environmental and social problems through engaging in commercial activities (Ahmadsimab and Chowdhury 2021; Battilana and Lee 2014; Doherty, Haugh, and Lyon 2014). Compared to businesses and charities that focus on either profit maximisation or social mission, social enterprises incorporate both aims at their core (Battilana and Lee 2014; Smith and Besharov 2019). Embracing plural value logics can shape novel ways of doing business (Laasch 2018; Ocasio and Radoynovska 2016). For example, Ecover, a Belgium ecological cleaning products manufacturer, uses only recycled plastic and sustainable ingredients in its products to reduce pollution (Ecover 2021); Fairphone, a Dutch smartphone maker, promotes fair trade and prevents modern slavery throughout their supply chains (Fairphone 2021); and Brownies&downieS, a chain of restaurants in the Netherlands, aims to create job opportunities for people with intellectual disabilities (Brownies&downieS 2021). While they are all social enterprises that aim to create joint social and commercial values, they differ from each other in terms of business scope, final products, and targeted problems. In other words, the value creation process can vary across social enterprises.

As the research field of social enterprises has become progressively mature (Gupta et al. 2020), researchers are increasingly aware of the heterogeneity among social enterprises (Defourny, Nyssens, and Brolis 2021). Several studies have hinted at the necessity of exploring such heterogeneity and classifying social enterprises based on conceptual framing (e.g. Defourny, Nyssens, and Brolis 2021; Neck, Brush, and Allen 2009; Zahra et al. 2009) or empirical exploration (e.g. Erpf, Tekula, and Neuenschwander 2019; Mair, Battilana, and Cardenas 2012). However, many previous papers classified social enterprises by focusing on aspects of the dual mission (e.g. Defourny, Nyssens, and Brolis 2021; Erpf, Tekula, and Neuenschwander 2019; Neck, Brush, and Allen 2009; Zahra et al. 2009), without addressing different approaches social enterprises opt for creating value with such a dual mission. Moreover, they differentiate social enterprises mainly on very few aspects, which may mask pivotal nuances in what they do, how they address a social problem, and for whom they create impact (Mair, Battilana, and Cardenas 2012): creating value through their business models. In this study, we aim to overcome these limitations and to provide a more fine-grained classification with a focus on the dual value creation processes of social enterprises. Therefore, our research question is: ‘How do social enterprises differ in their value creation processes?’

To capture social enterprises’ value creation process, we employ a business model perspective to guide our research (Osterwalder and Pigneur 2010; Zott, Amit, and Massa 2011). This perspective regards an organisation as a sophisticatedly designed system composed of multiple elements (Morris, Schindehutte, and Allen 2005). It therefore provides us with a holistic approach to explaining how an organisation functions to achieve its goals and the origins of the heterogeneous value creation processes (Cosenz and Noto 2018; Täuscher and Laudien 2018; Zott, Amit, and Massa 2011). Based on a thorough review of existing business model frameworks (e.g. Osterwalder and Pigneur 2010; Morris, Schindehutte, and Allen 2005), we identified three prime dimensions of value creation processes (i.e., value identification, value delivery, and value capture) (Teece 2010), which are further specified into 28 variables. Empirically, we rely on data collected from 437 social enterprises in the Netherlands to conduct a cluster analysis. Our results revealed seven discernible business models of social enterprises, which are labelled as: Contract Manufacturing Work Integration Social Enterprise (WISE), Green Service WISE, Green Product WISE, Green Innovation Social Enterprise, Fair-Trade Social Enterprise (FTSE), Platform-Based Social Enterprise, and Community-Based Social Enterprise.

Our research contributes to the existing literature in three ways. First, we extend insights into the fundamentally different value creation processes among social enterprises. The hybrid nature and intrinsic need for innovation have pushed social enterprises to apply a variety of approaches to survive and grow (Mair, Battilana, and Cardenas 2012; Seelos and Mair 2005). Based on an extensive set of value creation variables, our taxonomy identifies the sources of the observed heterogeneity and shows seven paths that social enterprises can choose to create value. In this way, we enrich existing classifications in terms of clarity and comprehensiveness. Second, our taxonomy can serve as a common ground in the field that encourages comparative works on aspects including management approach, impact scaling, and strategic effectiveness of different social enterprise business models (Erpf, Tekula, and Neuenschwander 2019; Nicholls 2010). Without adequate insights into the heterogeneity of social enterprises, empirical work may repeatedly draw on the same set of representative cases or generate conclusions that are largely disconnected (Mair, Battilana, and Cardenas 2012; Nicholls 2010). Therefore, the detailed value creation framework of our taxonomy opens up opportunities to refine and extend the existing literature on social enterprises through more nuanced insights into their organisational processes. Finally, our study also contributes to the literature on business model research. Our results suggest two approaches that organisations can use to reconfigure their designs and achieve business model innovation: learning from each other (e.g. green service WISEs & green product WISEs) and incorporating external elements, such as digitalisation (i.e., Platform-Based Social Enterprises). Accordingly, our taxonomy can serve as a starting point to explore how social enterprises evolve to create more viable business models and how commercial organisations can adopt to become more socially and environmentally sustainable. Given the increasing visibility of social enterprises in many economies and the trend towards marketizing the social service sectors in developed economies (Doherty, Haugh, and Lyon 2014; Zahra et al. 2009), our findings also have important managerial implications for practitioners and policymakers to recognise the differences among social enterprises in decision-making processes to assist the success of this emerging organisational form.

Literature Review and Theoretical Background

Theoretically, our research builds on three pillars: the social enterprise literature, the business model literature, and the value creation literature.

Social Enterprises and Their Characteristics

Following the seminal work of Battilana et al. (2015) social enterprises are defined as ‘organizations…[that] primarily pursue a social mission while also engaging in commercial activities to sustain their operations through sales of products and services’ (1658). This definition clearly distinguishes social enterprises from commercial organisations or nonprofits and highlights the focus of social enterprises. That is, profit generation is of secondary importance for social enterprises, while their essential aim is to trigger systematic changes in a broader context (Austin, Stevenson, and Wei–Skillern 2006; Mair, Battilana, and Cardenas 2012).

As a promising alternative to philanthropic initiatives and corporate social responsibility, social enterprises present two unique features: hybridity and innovativeness (Doherty, Haugh, and Lyon 2014; Huybrechts and Nicholls 2013; Pullman, Longoni, and Luzzini 2018; Santos, Pache, and Birkholz 2015). First, social enterprises straddle the non-profit and commercial sectors and combine the potentially conflicting goals of profit generation and public welfare creation into their design (Battilana and Lee 2014; Doherty, Haugh, and Lyon 2014; McMullen 2018). While this combination facilitates systematic changes in a more sustainable manner, it also makes social enterprises vulnerable to tensions of the conflicting institutional demands (Pache and Santos 2010; Wry and York 2017). Second, social enterprises are innovative when pursuing their dual mission (Battilana and Lee 2014; Pullman, Longoni, and Luzzini 2018). Social enterprises often step in when public and commercial organisations fail to effectively mobilise resources to tackle difficult social problems (Austin, Stevenson, and Wei–Skillern 2006; Pullman, Longoni, and Luzzini 2018). As highlighted by Thompson and MacMillan (2010), ‘if the problem were tractable, some profit-seeking enterprises would already be making profits resolving it’ (292). To assuage competing institutional demands and be successful in both sectors, social enterprises have to generate innovative solutions for managing and structuring, rather than simply replicating the practices of existing businesses or charities (Battilana and Lee 2014; Doherty, Haugh, and Lyon 2014; Wry and York 2017).

Current Classifications of Social Enterprises

The ingenuity of social enterprises in problem-solving has sparked increasing interest from academia (Bolzani, Marabello, and Honig 2020; Gupta et al. 2020). Several studies have tried to differentiate social enterprises conceptually (i.e., typologies) and empirically (i.e., taxonomies) (Kilic et al. 2015). So far, previous efforts mainly classify social enterprises according to the extent of prioritising their social mission, without considering the underlying value creation process. For example, Zahra et al. (2009) classify social enterprises into three types (i.e., social bricoleur, social constructionist, and social engineer) based on two dimensions: social opportunity discovering approach and significance of their social impact. However, this typology is heavily driven by the scope and scale of the social issue, while ignoring the actual problem solving process of social enterprises. Neck, Brush, and Allen (2009) identify the degree of hybridity and market impact as the two crucial factors to form a typology, but their key focus is to distinguish social enterprises from commercial organisations. More recently, Santos, Pache, and Birkholz (2015) develop a typology through the social value spill-over condition and the overlap between beneficiaries and customers. Similarly, Defourny and Nyssens (2017) suggest a typology based on the identity of beneficiaries (i.e., shareholders vs. stakeholders vs. the entire community) and the dual mission priority. Still, they ignore how value can be created for and obtained by the beneficiaries.

Instead of proposing classifications through theoretical deduction, other scholars try to understand the diversity of social enterprises from empirical data. Specifically, Spear (2006) categorises six social enterprises on their business activities through a multiple-case design. While useful, the limited sample size prevents us from getting a comprehensive picture. With a larger sample size, Mair, Battilana, and Cardenas (2012) developed a taxonomy of 200 well-established social enterprises according to their social issue domains and required resources. However, it does not explain the bridging process between the inputs and outcomes. Finally, Erpf, Tekula, and Neuenschwander (2019) inductively classify social enterprises into social service providers, social change makers, and social philanthropists based on 70 social entrepreneurs’ perceptions of the social orientation, attitude towards profit, market orientation, idea innovation, impact scale, and motive of their organisation. Sharing many similarities with Zahra et al.’s work (2009), this taxonomy also focuses on the ontology of social enterprises without considering other prominent factors that differentiate social enterprises, specifically the value creation. Thus, to better understand the specific nature of social enterprises and to overcome the limitations of previous work, we aim to systematically uncover the heterogeneity of social enterprises through their value creation process.

The Business Model Perspective

To better understand the heterogeneity among social enterprises, a business model perspective (Massa, Tucci, and Afuah 2017; Osterwalder and Pigneur 2010; Zott, Amit, and Massa 2011) is used as the theoretical lens. The concept of business model combines notions from the value chain concept with those of strategic positioning (Morris, Schindehutte, and Allen 2005; Porter 1996), and is regarded as the key to understanding organisations (Seelos 2014). A business model is defined as ‘the design or architecture of the value creation, delivery, and capture mechanisms [a firm] employs’ (Teece 2010, 172). It constitutes a firm’s architectural backbone, including its organisational boundaries, value proposition, and main operational activities (Fjeldstad and Snow 2018; Massa, Tucci, and Afuah 2017; Morris, Schindehutte, and Allen 2005). Since a business model clarifies the fundamental decisions a firm has to make, it helps to form systematic comparisons across firms (Morris, Schindehutte, and Allen 2005).

The business model as an organisation design tool has been widely applied in commercial organisations (Laasch 2018). The many novel business ideas, approaches and start-ups under the umbrella of the ‘new economy’ have further accelerated diffusion in the use of business models (Cosenz and Noto 2018; Oerlemans and Knoben 2010). However, a business model is not exclusively shaped by the utilitarian logic of the commercial sector (Laasch 2018). It can also be helpful for understanding a wider range of organisational forms, including non-commercial and even partially commercial organisations, such as social enterprises (Laasch 2018; Randles and Laasch 2016). Specifically, by highlighting the most important elements of organisational design, the business model perspective would help articulate the complex configuration of social enterprises (Cosenz and Noto 2018; Massa, Tucci, and Afuah 2017). Subsequently, by comparing the similarities and differences among those elements, patterns can emerge that identify a new taxonomy of social enterprises (Meyer, Tsui, and Hinings 1993; Morris, Schindehutte, and Allen 2005).

Value Creation of Social Enterprise Business Models

The process of value creation is central in identifying the essential elements of a specific business model (Freudenreich, Lüdeke-Freund, and Schaltegger 2020; Short, Moss, and Lumpkin 2009; Teece 2010). Whenever a firm is established, it will explicitly or implicitly adopt a business model that explains the value creation logic behind its processes (Massa, Tucci, and Afuah 2017). Following previous literature (Morris, Schindehutte, and Allen 2005; Rohn et al. 2021; Teece 2010; Zott, Amit, and Massa 2011), we build our framework along three dimensions of the value creation process: (1) value identification that expresses an organisation’s orchestration of resources and its value proposition to satisfy customers’ needs (Osterwalder and Pigneur 2010; Täuscher and Laudien 2018); (2) value delivery that describes how an organisation transfers value to the target group (Freudenreich, Lüdeke-Freund, and Schaltegger 2020; Täuscher and Laudien 2018), and; (3) value capture that illustrates how an organisation extracts value delivered to the target group in the form of social, environmental, and economic gains (Kumar and Reinartz 2016; Täuscher and Laudien 2018). As there is no general agreement on elements representing each dimension (Hartmann et al. 2016), we identified ten relevant elements based on a review of six highly cited business model frameworks (Table 1), including value proposition, key resource, transaction content, distribution channel, transaction type, geographic scope, market participant, revenue stream, key activity, and cost structure. Following the recommendation of Hartmann et al. (2016), cost structure is discarded, as collecting reliable data is challenging. Hence, nine elements remain for further operationalisation. Table 2 presents definitions of the nine elements and 28 analytical constructs under these elements.

Table 1. Summary of the different business model frameworks.

References	Value identification		Value delivery					Value capture			Other elements	No. of citations (As of Google Scholar 2021-10-05)
	Value proposition	Key resource	Offering: Transaction content	Offering: distribution channel	Offering: transaction type	Offering: geographic scope	Market participant	Revenue stream	Key activity	Cost structure
Chesbrough (2002)	√		√	√			√	√		√	Competitive strategy	6,572
Hedman & Kalling (2003)	√	√	√	√			√		√		Competitors, management processes	1,410
Morris, Schindehutte, and Allen (2005)	√		√	√		√	√	√			Investor factor, economic factor, internal capability factors	3,793
Johnson, Christensen, and Kagermann (2008)	√	√						√		√	Key process	4,162
Osterwalder and Pigneur (2010)	√	√	√	√	√		√	√	√	√	Key partnerships	15,615
Baden-Fuller and Haefliger (2013)	√		√	√			√	√		√	N/A	1,215
This study	√	√	√	√	√	√	√	√	√	√	N/A	N/A

Table 2. Key elements and analytical constructs of a business model.

	Elements		Analytical constructs
Value identification	Value proposition	Aim of the business: In addition to generating commercial value to sustain their operations, social enterprises may specifically focus on environmental and/or social value creation at their core (Freudenreich, Lüdeke-Freund, and Schaltegger 2020; Santos, Pache, and Birkholz 2015).	Environmental value creation: Efforts that aim to address environmental issues, such as preventing green gas emission, reducing the use of plastic, and saving natural resources (Landy, Roberts, and Thomas 1990) Social value creation: The creation of desired social change or social impact through the resolution of social issues (Singh 2016).
		Target Group (Customer-Beneficiary relationship): Customers of the commercial value creation and beneficiaries of its social value spill-overs. Following the conceptualisation of Santos, Pache, and Birkholz (2015), we further differentiate between three mutually exclusive situations for the overlap between customers and beneficiaries.	Customers = Beneficiaries: situation in which customers who pay for a social enterprise are the same group of people who directly^1 benefit from its impact. For example, people who use environmentally friendly products will benefit from less environmental pollution. Customers ≠ Beneficiaries: situation in which customers and beneficiaries are two different groups. For example, people who purchase fair-trade products will benefit local farmers but not themselves. Customers < Beneficiaries: situation in which customers are just one subset of the beneficiaries of a social enterprise. For example, customers shop for fair-trade products that are also environmentally friendly.
	Key resources: The most important resources for a social enterprise to offer the aimed value (Osterwalder and Pigneur 2010).		Physical resources: Physical assets a company processes such as machines, lands, factories, vehicles, and facilities (Osterwalder and Pigneur 2010) Intellectual resource: Proprietary knowledge and knowing capability such as technology, proprietary knowledge, patents, and copyrights of a social collectivity (Nahapiet and Ghoshal 1998; Osterwalder and Pigneur 2010). Human resource: Learned skills, knowledge, and experience that require some investment in education, training, or learning from individuals (Adner and Helfat 2003; Helfat and Peteraf 2003) Network resources: Social ties and relationships with external market players (Helfat and Peteraf 2003) Financial resources: The funds an organisation obtained to finance its investment activities (Osterwalder and Pigneur 2010).
Value delivery	Transaction content: Object being exchanged by an organisation, which can be products, services, or a mix of both (Casadesus-Masanell and Ricart 2010)		Product: items being exchanged by a business entity that are highly tangible, manufactured, and can be stored (Nijssen et al. 2006). Services: items being exchanged by a business entity that are highly intangible, perishable, and co-produced with customers (Nijssen et al. 2006)
	Transaction type: The format of transaction, which can be online or offline (Täuscher and Laudien 2018)		Online: situations in which customers interact with employees in a virtual environment through the website interface (e.g. Internet, online platform, or mobile APPs) to perform transactions (Broekhuizen 2006). Offline: situations in which customers interact with employees in a physical space (e.g. restaurants, shops, or supermarkets) to perform transactions (Broekhuizen 2006).
	Marketplace participant: The way in which products or services are delivered to the end customers, which includes direct and indirect distribution (Eggert, Henseler, and Hollmann 2012)		Individual customer (B2C): direct transactions of products or services between businesses and their customers who are the end-users of those products or services (Kumar and Raheja 2012) Business organisation (B2B): transactions of products or services between businesses, such as between a manufacturer and a wholesaler, or between a wholesaler and a retailer (Kumar and Raheja 2012).
	Geographic scope: The customer geographic dispersion of an organisation (Morris, Schindehutte, and Allen 2005).		Global: situations in which the geographical scope of a business covers multiple countries across different continents. Regional: situations in which the geographical scope of a business covers multiple neighbour countries within the same continent. Local: situations in which the geographical scope of a business covers the domestic country or a specific local community within the country.
	Distribution channel: The way in which products or services are delivered to the end customers, which includes direct and indirect distribution (Eggert, Henseler, and Hollmann 2012).		Direct distribution: situation where goods or services are directly delivered from the manufacturers to the end customers without any intermediaries (Eggert, Henseler, and Hollmann 2012). Indirect distribution: situation where goods or services are relied on the intermediaries/distributors to pass from the manufacturers to the end customers (Eggert, Henseler, and Hollmann 2012).
Value capture	Key revenue stream: The major revenue sources through which an organisation generates profits from its operations (Osterwalder and Pigneur 2010).		Asset sales: revenue stream that derives from selling ownership to a physical product; Usage fees: revenue stream that derives from the provision of a particular service; Brokerage fees: revenue stream that derives from the intermediation services performed on behalf of two or more parties in terms of commission fee (Osterwalder and Pigneur 2010).
	Social outcome: The most significant social impact that can be achieved in fulfilling social objectives (Doherty, Haugh, and Lyon 2014; Santos, Pache, and Birkholz 2015).		Social cohesion: practices that promotes the solidarity and connectedness among groups in a society, such as reducing wealth gap, reducing poverty, preventing modern slavery, improving working conditions, and improving the cohesion of a community (Kawachi and Berkman 2000). Labour participation: practices that contributes to the employability and the long-term work opportunities of people (e.g. the disabled, refugees, and migrants) who have been marginalised from the main job market. Environmental protection: practices that protect the natural environment, minimise harmful environmental impact, and limit secondary wastes (Panwar, Kaushik, and Kothari 2011). Physical and/or mental well-being: individual perception of happiness and life satisfaction (i.e., health) (Helliwell and Putnam 2004). It shows the extent to which individuals can use their psychological, social, and physical resources to meet a particular challenge (Dodge et al. 2012).

Methodology

Sample and Data Collection

The context of this study was the Netherlands because of the long history, large numbers, and relatively mature status of the social enterprise landscape (Defourny and Nyssens 2008; Doherty, Haugh, and Lyon 2014; Short, Moss, and Lumpkin 2009; Spear and Bidet 2005). According to the European Commission (2014), the Netherlands is home to around 4,000 to 5,000 social enterprises with an average growth rate of 10% per year. Moreover, since Dutch social enterprises closely reflect the wide range of activities possible within any economic entity (European Commission 2014), the Netherlands offers a promising field with adequate sources of data.

The sample set came from Social Enterprise NL, the most prominent social enterprise support organisation in the Netherlands. This sample was appropriate as Social Enterprise NL’s definition for social enterprises matches Battilana et al. (2015) and it admits members accordingly based on a conscientious selection process (Social Enterprise NL 2021). To become a member, a social enterprise needs to specify its impact areas and business sectors. Social Enterprise NL will only recognise an enterprise as a social enterprise when it is over 50% financially independent. Once membership has been granted, a profile page of the selected social enterprise would become publicly available on the Social Enterprise NL website. Hence, it is evident that all enterprises admitted by Social Enterprise NL are appropriate for our study adding to the validity and reliability of our sample.

The database contains 437 social enterprises. Cases that did not disclose adequate information to interpret their business models or that encountered bankruptcy during our data collection process were eliminated. Additionally, we excluded social enterprises with a mix of product and service provisions where multiple business models existed, as our goal is to distinguish business models of social enterprises in a concise manner. In doing so, the resulting sample for further analysis consisted of 357 social enterprises from eight business sectors (Table 3). Over 86% of the social enterprises were founded after 2005, and their average number of employees is 67 (Range 1 ∼ 3800, standard deviation = 296, missing value = 24). It is important to note that this is not a representative or random sampling process. Instead, we collected data from the entire population of qualifying social enterprises in order to identify a sufficient range of existing social enterprise business models.

Table 3. Descriptive statistics of business sectors in the sample (N = 357).

Size (Mean = 67)	1-10	146	40.9%
	11-50	133	37.3%
	51-200	39	10.9%
	200+	15	4.2%
	Unknown	24	6.7%
Year of founding (Mean = 2011)	Before 2000	29	8.1%
	2000-2005	21	5.9%
	2006-2010	57	16.0%
	2011-2015	150	42.0%
	2016-2020	90	25.2%
	Missing	10	2.8%
Sector	Construction	5	1.4%
	Manufacturing	128	35.9%
	Transportation & Public Utilities	12	3.4%
	Wholesale Trade	3	0.8%
	Retail Trade	58	16.2%
	Finance, Insurance, & Real Estate	15	4.2%
	Services	129	36.1%
	Public Administration	7	2.0%

Based on the approach in similar work (e.g. Hartmann et al. 2016; Mair, Battilana, and Cardenas 2012; Täuscher and Laudien 2018), we collected data for the 357 social enterprises from their profile pages in Social Enterprise NL, official websites, and social media accounts (e.g. Facebook, LinkedIn, and Instagram). Relevant texts and images that describe the background, mission, and business strategy of each case were gathered. This type of data is sufficient for delineating business models and for ensuring the descriptive validity of our findings (Hartmann et al. 2016).

Data Analysis

We combined qualitative and quantitative approaches to analyse the data in four steps (Creswell 2017; Creswell and Plano Clark 2007; Denscombe 2008). First, we coded text materials to make valid inferences about the objects based on our theoretical framework in Table 2 (Gerring 2004; Weber 1990). Next, we used several clustering techniques to create the taxonomy (Brusco et al. 2017). Then we applied split-sample cross-validation (Everitt et al. 2011; McIntyre and Blashfield 1980) and analysis of variance (ANOVA) to check the robustness of the identified clusters. Finally, we provided seven case descriptions of representative social enterprises selected from each cluster and two additional analyses across clusters to facilitate interpretation and bolster our taxonomy (Section ‘Results’).

Coding Process

The coding process was performed manually by two independent coders, one author of this paper and a research assistant familiar with social enterprise research. We used Excel to record, organise, and sort data. We started the coding process by conducting multiple readings of the documents for each case. To ensure coding reliability and internal consistency, we developed a coding scheme (Appendix A, supplementary material) that defines each element based on a pilot coding of 20 randomly selected cases after several rounds of discussion to reach agreement. We treated each element as a binary variable and codified whether a particular characteristic is present in a case or not (0 = no; 1 = yes). The two coders met regularly to discuss divergent interpretations, ambiguous texts, and unclear cases. As a result, we achieved a high degree of inter-coder reliability of the coding matrix at the variable level (Cohen’s Kappa = 99.16%) and case level (Cohen’s Kappa = 84.09%) (Neuendorf 2017; Perreault and Leigh 1989). To finalise the coding process, all 49 deviating cases were discussed with another author to reach a final decision. This step resulted in a binary matrix of 357 rows (number of cases) and 28 columns (number of variables).

Cluster Analysis

In the next stage, we conducted a cluster analysis to explore the underlying data structure. Since redundant variables may deteriorate the validity of a solution, we started this process with variable selection. After an initial variable screening, we excluded geographic scope as it turned out to be mainly related to organisational size rather than to any aspect of the value creation process. Following the approach by Mair, Battilana, and Cardenas (2012), we further excluded variables with restricted distribution (i.e., <10% of the population, and >90% of the population having a characteristic), as they may be either too common or too rare in a population. Consequently, the variables of financial resources (2.2% of the distribution) and the direct distribution approach (95.9% of the distribution) were removed. To avoid multicollinearity issues that may obscure the underlying constructs (Everitt et al. 2011; Ketchen and Shook 1996), we conducted a Jaccard-Tanimoto similarity coefficient test (see Appendix B, supplementary material) to examine the correlation between binary data where Pearson correlation is not applicable. This led to the exclusion of the variables ‘the aim of business – environmental & social’, ‘transactional content – product & service’, and ‘transaction type – offline’. Finally, we randomly dropped one of the three variables that measure the customer-beneficiary relationship (i.e., ‘Customers < Beneficiaries’), since bringing any two of them into the analysis would present all possible situations. The screening process resulted in 17 variables for subsequent analysis.

We then followed the commonly suggested two-step procedure of conducting cluster analysis, which combines hierarchical and non-hierarchical cluster algorithms to increase the validity of solutions (Brusco et al. 2017; Hair et al. 1992; Milligan 1980; Steinley and Brusco 2007).

The hierarchical clustering algorithm was applied first to determine the number of clusters, followed by the non-hierarchical method to identify the centroids and the optimal assignment of objects to each cluster (Brusco et al. 2017; Hardy 1996; Yeung, Chan, and Lee 2003). We used the recommended Jaccard’s coefficient to measure the distance between binary objects (Everitt et al. 2011). In hierarchical clustering, we used Ward’s method to calculate the agglomeration coefficient. The derived dendrogram suggests a classification of six to eight clusters. To further determine the most appropriate number, we checked the C-Index (Hubert and Levin 1976) and Point Biserial (Milligan 1980, 1981), which are demonstrated as the best stopping rules (Milligan and Cooper 1985). Both indices suggested that seven clusters would be the optimal solution.

In the second step, we applied k-medoids to form clusters, where k is equal to seven. K-medoids was chosen over the commonly-used k-means, which is not appropriate for clustering binary data and is sensitive to outliers (Brock et al. 2008; Han, Kamber, and Pei 2012; Hartmann et al. 2016). Table 4 shows the final results of the seven groups represented by their medoids (or centroids) and the percentage of cases sharing the corresponding characteristics within the seven clusters.

Table 4. Final cluster results, representative medoids, and percentage of total sample.

Cluster	1	2	3	4	5	6	7
Number of cases	59	52	39	50	42	42	73
Physical resources	1 (96.6%)	1 (100%)	1 (94.9%)	0 (80.0%)	1 (64.3%)	1 (92.9%)	1 (84.9%)
Human resources	0 (66.1%)	1 (98.1%)	0 (84.6%)	0 (68.0%)	1 (95.2%)	1 (90.5%)	1 (98.6%)
Intellectual resources	1 (59.3%)	0 (96.2%)	1 (76.9%)	0 (94.0%)	0 (81.0%)	0 (92.9%)	0 (94.5%)
Network resources	0 (96.6%)	0 (73.1%)	0 (89.7%)	1 (96.0%)	0 (90.5%)	0 (81.0%)	0 (90.4%)
Customers ≠ Beneficiaries	0 (76.3%)	0 (73.1%)	0 (100%)	0 (92.0%)	0 (100%)	0 (100%)	1 (100%)
Customers = Beneficiaries	0 (98.3%)	0 (100%)	1 (100%)	0 (78.0%)	1 (78.6%)	0 (95.2%)	0 (100%)
Online	1 (81.4%)	1 (80.8%)	1 (76.9%)	1 (90.0%)	0 (73.8%)	0 (88.1%)	0 (63.0%)
Indirect Distribution	1 (61.0%)	1 (44.2%)	1 (53.8%)	0 (96.0%)	0 (97.6%)	0 (95.2%)	0 (100%)
B2B	1 (93.2%)	1 (67.3%)	1 (84.6%)	1 (82.0%)	1 (64.3%)	1 (92.9%)	1 (97.3%)
B2C	1 (76.3%)	1 (96.2%)	1 (84.6%)	1 (68.0%)	1 (88.1%)	0 (57.1%)	0 (68.5%)
Asset sales	1 (93.2%)	1 (100%)	1 (84.6%)	0 (76.0%)	0 (95.2%)	0 (97.6%)	0 (98.6%)
Usage fees	0 (93.2%)	0 (100%))	0 (84.6%)	1 (60.0%)	1 (95.2%)	1 (97.6%)	1 (98.6%)
Brokerage fees	0 (98.3%)	0 (73.1%)	0 (100%)	1 (76.0%)	0 (97.6%)	0 (92.9%)	0 (94.5%)
Environmental protection	1 (72.9%)	1 (71.2%)	1 (94.9%)	1 (76.0%)	0 (85.7%)	1 (92.9%)	0 (100%)
Labour participation	0 (86.4%)	1 (100%)	0 (100%)	1 (52.0%)	0 (85.7%)	1 (88.1%)	1 (89.0%)
Social cohesion	1 (89.8%)	0 (86.5%)	0 (100%)	1 (76.0%)	0 (66.7%)	0 (78.6%)	0 (87.7%)
Well-being	0 (74.6%)	0 (96.2%)	0 (82.1%)	1 (70.0%)	1 (71.4%)	0 (85.7%)	0 (94.5%)
Cluster Type	E	C	D	F	G	B	A

Validation of Clusters

This step involved the use of several techniques to ensure the robustness of our seven-cluster solution. First, we employed split-sample cross-validation to test replicability of the cluster solution (Aldenderfer and Blashfield 1984; McIntyre and Blashfield 1980). We randomly divided the sample into two halves and conducted cluster analysis following the approach used earlier (McIntyre and Blashfield 1980; Yeung, Chan, and Lee 2003). A high level of similarity was found between the subsamples and the original solution as well as from cross-validating the two subsets through bootstrapping (average Rand Index = 0.87, average adjusted Rand Index = 0.46). It indicates an acceptable level of internal consistency and generalisability of our results. Second, three tests checked the internal and external validity of the original clustering (Everitt et al. 2011). The internal validity test results from one-way ANOVA based on 17 variables were all significant (P < 0.001) (Appendix C, supplementary material), which suggests that the selected variables contribute significantly to differentiate the clusters. In order to check the superstability among any possible pair of clusters and to safeguard against type I error, Scheffé multiple comparisons were conducted (Scheffé 1959; Yeung, Chan, and Lee 2003). They revealed that 190 pairs out of the 357 possible combinations are highly significant (P < 0.05) (Appendix D, supplementary material). Although the results suggest certain pairs of clusters are similar for a specific variable, they demonstrated significant differences across most individual groups, considering the large number of variables and data type of this study. Finally, we checked the external validity of the cluster solution by using external variables that were omitted from defining the clusters but mirror some expected characteristics of the classified groups (Aldenderfer and Blashfield 1984; Yeung, Chan, and Lee 2003). The significant differences (P < 0.001) in the seven groups across the 11 external variables supported the external validity of the clusters (Appendix E, supplementary material).

Results

Interpretation of Clustering Results

The interpretation stage of taxonomy development focuses on describing the main characteristics of the clusters and labelling them to reflect their distinguishing characteristics (Yeung, Chan, and Lee 2003; Hair et al, 1992). Table 5 summarises the profile of the seven identified clusters.

Table 5. Profile information of the seven clusters.

Business model	Key resources	Customer-beneficiary relationship	Transaction content & transaction type	Distribution channel	Marketplace participant	Key revenue stream	Key social outcome
Contract Manufacturing Work Integration Social Enterprises	Physical resources Human resources	Customers ≠ Beneficiaries	Offline services provision	Direct	B2B	Usage fees	Labour participation
Green Service Work Integration Social Enterprises	Physical resources Human resources	Customers < Beneficiaries	Offline services provision	Direct	B2B, (B2C)	Usage fees	Labour participation, & Environmental protection
Green Product Work Integration Social Enterprises	Physical resources Human resources	Customers < Beneficiaries	Offline & Online product sales	Direct Indirect	B2B, B2C	Asset sales	Labour participation, & Environmental protection
Green Innovation Social Enterprises	Physical resources Intellectual resources	Customers = Beneficiaries	Offline & Online product sales	Direct Indirect	B2B, B2C	Asset sales	Environmental protection
Fair-trade Social Enterprises	Physical resources Intellectual resources	Customers < Beneficiaries	Online & Offline product sales	Direct Indirect	B2B, B2C	Asset sales	Social cohesion, & Environmental protection
Platform-based Social Enterprises	Network resources	Customers < Beneficiaries	Online service provision	Direct	B2B, B2C	Usage fees Brokerage fees	Labour participation, Social cohesion, Well-being, & Environmental protection
Community-based Social Enterprises	Physical resources Human resources	Customers = Beneficiaries	Offline service provision	Direct	B2B, B2C	Usage fees	Well-being

Work Integration Social Enterprises

Work Integration Social Enterprises (WISEs) refer to social enterprises that pursue a social mission of helping the long-term unemployed back to the workforce (Battilana et al. 2015). In line with observations by the European Commission (2014), WISEs constitute the dominant form of social enterprise in the study, as 167 out of the 357 social enterprises (46.8%) fell into this category. However, the delivery of work integration activities can be achieved through a wide range of product and service provisions (European Commission 2014). Specifically, we identified three different WISE business models in our sample.

Cluster A: Contract Manufacturing Work Integration Social Enterprises

Social enterprises in this cluster generally have a clear mission of providing job opportunities to people neglected by the regular labour market. Beneficiaries of these social enterprises are people with physical or intellectual disabilities, unemployed women in developing countries, young people with limited levels of education, and migrants facing exclusion from new communities. To fulfil the mission, social enterprises in this cluster often make use of physical (e.g. workshops and factories) and human resources (e.g. experts and managers with social work backgrounds) in their operations. They generally work as contractors to provide outsourced services, such as product assembly, facility management, and construction work, directly to business customers. Usage fees from providing contractual services are the revenue source in this business model. Because of these main characteristics, we label this cluster as ‘Contract Manufacturing WISE’. Interestingly, we found that Contract Manufacturing WISEs engage in providing not only the low-skilled activities frequently mentioned in previous studies (c.f., Battilana et al. 2015; Longoni et al. 2019), but also in providing high-skilled services, such as digital/Internet-based software programing and creative design (c.f., Smith and Besharov 2019). These activities further reflect the innovativeness of social enterprises in solving structural unemployment.

Brainport Assembly is a typical Contract Manufacturing WISE. The organisation firmly believes that an inclusive society requires providing people with occupational disabilities a sustainable way of making a living. Guided by that mission, over 70% of its workforce consists of people who are marginalised from the labour market (Brainport Assembly 2021). Brainport Assembly creates job opportunities and training for workers through parts and product assembly for local small to medium-sized firms. Income from providing contractual services constitutes the main source of its revenue. To accelerate social changes, Brainport Assembly donates 25% of its profits to an independent foundation to support other employers who also wish to contribute to an inclusive society.

Cluster B: Green Service Work Integration Social Enterprises

Although sharing several characteristics with Contract Manufacturing WISEs, the main difference of social enterprises in Cluster B lies in the integration of environmental protection into their missions. Specifically, these WISEs tend to emphasise the use of environmentally friendly materials, renewable energy, and green logistics in their service operations. Moreover, a small percentage of them also include missions that promote social cohesion (12.3%) or well-being of citizens (5.5%). To reflect the environmental focus of these social enterprises, we label them as ‘Green Service WISEs’. While business customers are still their main focus, many of them also provide services to individual customers (42.9%). Due to the dual social aims, target beneficiaries of these social enterprises range from the disadvantaged to customers who would like to make a positive impact on the environment. Usage fees from providing the services are the major revenue sources of this business model.

Life2 is a good example of Green Service WISEs. Focusing on the lifespan extension of furniture, Life2 provides furniture restyling service to local businesses in a budget- and environmentally friendly way (Life22, 2021). In doing so, Life2 contributes to the circular economy and provides job opportunities to the disabled. Another illustrative case is Restaurant De Oude Keuken (The Old Kitchen) in Amsterdam (Oude Keuken 2021). Run by people with intellectual disabilities and a team of professionals, De Oude Keuken provides catering services to the local community. It is also an ideal place for companies to hold meetings and events. De Oude Keuken prefers to use organic ingredients and biodegradable packaging to reduce negative impacts on the environment and promote healthy eating.

Cluster C: Green Product Work Integration Social Enterprises

Different from the previous two types of WISEs, social enterprises in Cluster C focus on product provision, and they target both individual (96.2%) and business customers (67.3%). In this business model, disadvantaged workers make handcrafts and small items using recyclable materials. Products are then sold to customers directly or indirectly through online and/or physical shopping. Thus, the main revenue comes from product sales.

For example, EchtWaar is a workshop for ceramic and textile products that uses and reuses sustainable materials as much as possible (EchtWaar 2021). It welcomes people with occupational disabilities or those otherwise excluded from the regular job market. By working side-by-side with experts to make handcrafts, EchtWaar stimulates the growth and long-term development of employees who will eventually be guided to a regular workplace and make an independent living. Customers and organisations can buy products from EchtWaar’s shop or from other retail stores.

Cluster D: Green Innovation Social Enterprises

Different from WISE business models, social enterprises in Cluster D predominantly address environmental issues through product innovation and invention. Accordingly, we call them ‘Green Innovation Social Enterprises’ (GISEs). Direct beneficiaries are customers who are willing to pay for a greener planet. GISEs generally develop exclusive intellectual resources, such as formulas, patents, designs, or special manufacturing techniques that differ from existing unsustainable practices. Relying on factories and other physical resources, products are developed and manufactured. Sales of green products can be online or offline, using B2B and B2C models. Both direct and indirect distribution channels can be identified in GISEs, and their revenue comes from product sales.

A typical case of GISEs is BE O, which has designed and manufactured bioplastic water bottles since 2017 (BE O, 2021). To minimise hydrocarbon consumption and reduce carbon emission, BE O developed a sustainable bottle that is 100% recyclable made of sugarcane instead of plastic. Customers, who are also beneficiaries, automatically benefit from purchasing BE O’s bottles, since sugarcane absorbs more CO₂ than what is emitted during production. Individual customers can order BE O bottles directly online or indirectly from selected retailers, while companies can customise the bottles’ appearance and purchase in bulk. For every bottle sold, BE O will plant a tree to contribute to a greener future.

Cluster E: Fair-Trade Social Enterprises

Social enterprises in Cluster E pursue dual missions of social cohesion (89.8%) and environmental protection (72.9%) by producing fair-trade products using environmentally friendly manufacturing processes. Beyond customer fulfilment, these social enterprises intentionally reduce profit margins to ensure that a greater proportion of value rents can be earned by upstream workers or farmers (Nicholls and Huybrechts 2016). In some cases, they also provide capacity building to suppliers to reach those aims together. Therefore, we label them ‘Fair-Trade Social Enterprises’ (FTSEs). Agricultural products, such as tea, cocoa, and coffee, as well as electronic components made in developing countries, are commonly fair-trade items in this cluster. While physical resources (e.g. factories or inventories) are generally required (96.6%), intellectual resources are often needed (59.3%), depending on the manufacturing technology required for the final products. Similar to other product-driven social enterprises, FTSEs’ distribution channels can be direct and indirect, and are open to both B2B and B2C models online and offline. Sales of fair-trade products are the main revenue source for this social enterprise business model.

A well-known example of FTSEs is Fairphone (c.f., Akemu, Whiteman, and Kennedy 2016), a smartphone producer. Fairphone is widely recognised as a successful pioneer in combining social and environmental missions in its product design, manufacturing processes, and supply chain management. At the design stage, long-lasting and easily repairable smartphones are developed to reduce waste and facilitate later recycling. Fairphone protects and improves the incomes of miners through responsible procurement and builds long-term relationships with raw material suppliers. Workers’ well-being and job satisfaction in factories are also important concerns for Fairphone. It works closely with key manufacturers and regularly assesses working conditions in their factories. When conflicts between workers and the factories arise, Fairphone takes a mediator role to encourage communication between the two parties. In addition, Fairphone freely shares its supply chain information with the public, which further promotes transparency in the smartphone industry.

Cluster F: Platform-Based Social Enterprises

Social enterprises in Cluster F integrate elements of platform business model to facilitate transactions between independent participants, thus we call them ‘Platform-Based Social Enterprises’ (PBSEs). The idea behind them is that suppliers and customers lack direct interactions, which impedes the creation and scalability of transactions (Eisenmann, Parker, and Van Alstyne 2009; Rohn et al. 2021; Täuscher and Laudien 2018). Relying on network resources acquired from both the supply and demand side, PBSEs bring suppliers and customers together to exchange products and services with social missions. Consequently, demand-side participants can have more choices and better accessibility to sustainable products, and supply-side participants can reach a wider customer- and beneficiary-base (Eisenmann, Parker, and Van Alstyne 2009; Parker, Van Alstyne, and Choudary 2016). Multiple social outcomes can be achieved in this cluster, including environmental protection (76%), social cohesion (76%), labour participation (52%), and well-being promotion (70%). The intended beneficiaries include customers who directly benefit from a healthier lifestyle or more sustainable environment as well as the disadvantaged who obtain support. Brokerage fees, in the form of commissions, subscriptions, and membership fees, are the key revenue source (76%). Depending on the transactional content of the platform, usage fees (60%) and asset sales (24%) may also be included. Both B2B and B2C models are found, but the exchanges mainly take place online (90%).

A representative case of PBSEs is Excess Materials Exchange, which actively participates in the circular economy (Excess Materials Exchange 2021). Through its digital matching platform, high-quality reusable materials are exchanged between companies. While the financial value of the materials increases 10% on average, the ecological footprint is reduced by 60% (Excess Materials Exchange 2021). Accordingly, revenues of Excess Materials Exchange are obtained through the provision of matching services from its database and commissions from the transactions between participants. Another example is the online sustainable giftshop, Geschenk met Verhaal (Gift with Story). As a reliable gift platform for individuals and organisations, Geschenk met Verhaal not only considers the satisfaction of gift recipients, but also the sustainability and fairness of the gift-making process (Geschenk met Verhaal 2021). It only sources products that are made in an environmentally friendly manner, processed under decent working conditions, paid at a fair price, and contribute to the wellbeing of the end-users. Within its supply chain, Geschenk met Verhaal also collaborates with suppliers and invests in the future sustainable development of these workshops.

Cluster G: Community-Based Social Enterprises

Finally, social enterprises in Cluster G focus on providing public and social services, such as organising medical services for the disabled, providing special training and education to vulnerable groups, and arranging social housing in neighbourhoods. Their target beneficiaries are community members who pay for the services in return for improvement in their physical or mental well-being. Hence, we call them ‘Community-Based Social Enterprises’ (CBSEs). Experienced staff and professionals (95.2%) are usually required to coordinate related services, and physical resources (64.3%) may be needed to deliver the services. Both individuals and organisations can use the services, which are usually performed offline without the involvement of any agents. Revenues are in turn generated from usage fees for the service.

For instance, Beleef Dementie (Experience Dementia) is an experienced healthcare training centre that provides paid training to informal caretakers, healthcare professionals, and organisations involved in dementia care (Beleef Dementie 2021). Through dementia simulation workshops, Beleef Dementie aims to relieve the burden of customers by increasing their understanding of and empathy towards patients. As a result, people with dementia can experience a higher quality of life with their families for longer. Another example is Monkey Moves, a community sports training organiser that helps parents, nurseries and schools to improve young children’s physical fitness (Monkey Moves 2021). By working together with sports scientists, teachers, and therapists, Monkey Moves provides multisport programs and lessons that assist these customers in an inclusive atmosphere.

Supplemental Analyses and Interpretation

In addition to the within-cluster analysis described above, we also conducted two supplemental analyses to facilitate comparisons across clusters. First, we mapped the distribution of different social enterprise models across eight relevant industrial domains based on the 2-digit Standard Industrial Classification (SIC) code (see Figure 1). We can infer that design of a business model is influenced by its industrial justification about what constitutes value in addressing a social issue. Specifically, service-oriented business models (i.e., Contract Manufacturing WISEs, Green Service WISEs, and CBSEs) are more likely to appear in service-related industries, such as the business services and social services stream, while product-oriented business models (i.e., Green Product WISEs, GISEs, and FTSEs) frequently appear in the manufacturing industry, such as the food and related products sector and the miscellaneous manufacturing sector. In addition, the retail trade industry, particularly the sub-sectors of eating & drinking places and miscellaneous retail mostly applies WISE models (i.e., Contract Manufacturing WISEs, Green Service WISEs, and Green Product WISEs). In these models, technical training, such as cooking, catering, and handcrafting, as well as soft skills, including attendance, discipline, communication, and socialisation, are provided to the disadvantaged (Longoni et al. 2019). Moreover, the industrial distribution of PBSEs also clearly indicates their corresponding business model characteristics. Very few cases of PBSEs are identified in the manufacturing industry, since they do not produce any products (Eisenmann, Parker, and Van Alstyne 2009; Parker, Van Alstyne, and Choudary 2016; Rohn et al. 2021). Instead, they are evenly distributed over the remaining industrial domains through intermediary service provision.

Figure 1. Cluster distribution across industries.

After mapping the distribution of the different models, we also reviewed the correlation between all variables (see Appendix B, supplementary material) by matching them to corresponding business models. Our results reveal a strong correlation between resource configuration and the objective of a social enterprise business model. People-oriented social enterprises (e.g. Contract Manufacturing WISEs) are more likely to rely on human resources (Jaccard coefficient = 0.72, sig.=0.001) and physical resources (Jaccard coefficient = 0.69, sig.=0.022) to provide support, while green-oriented social enterprises (e.g. GISEs) tend to rely on physical resources (Jaccard coefficient = 0.53, sig.=0.033) and intellectual resources (Jaccard coefficient = 0.29, sig.=0.001) for sustainable manufacture and product innovation. As indicated by Fjeldstad and Snow (2018), the configuration of an organisation must support its purpose and objectives. Our results show that a social enterprise combines and deploys different resources to design its business model, depending on the mission pursued.

To conclude, the additional analyses illustrate patterns behind business models that social enterprises opt for across industries as well as possible factors that lead to such selections. Our results further indicate that the business model of a social enterprise is expected to reflect its value logic, resources configuration, and the environment within which it is embedded (Fjeldstad and Snow 2018; Russo et al. 2022; Sharma, Beveridge, and Haigh 2018).

Discussion & Implications

Theoretical Implications

This study contributes to the theoretical development of social enterprises and the business model perspective in three ways. First, we enrich the current understanding of heterogeneity among social enterprises by showing the detailed forms of entrepreneurial paths and resource configurations they adopt to create value. Compared to prior typologies that mainly categorise social enterprises according to the extent of prioritising their social mission or taxonomies based on few aspects or limited data, our study provides a fine-grained taxonomy grounded by substantial empirical data with the theoretical guidance from the business model perspective. It highlights the importance of differentiating social enterprises through multiple aspects that are embedded in their dual value creation process (Chatterjee, Cornelissen, and Wincent 2021; Seelos and Mair 2005; Short, Moss, and Lumpkin 2009). As a result, our new taxonomy enables the integration of previously unrelated classifications that are based on key resource (i.e., Mair, Battilana, and Cardenas 2012), the relationship between customers and beneficiaries (i.e., Santos, Pache, and Birkholz 2015), and outcome (i.e., Neck, Brush, and Allen 2009).

A comprehensive taxonomy is of great value, as it can be instrumental to interpret and unify findings from prior studies (Seelos 2014; Short, Moss, and Lumpkin 2009). For example, it offers an plausible explanation for the divergent resource mobilisation and legitimacy process between Fair-Trade Social Enterprises (c.f., Akemu, Whiteman, and Kennedy 2016) and Community-Based Social Enterprises (c.f., Hertel, Binder, and Fauchart 2021). In the study by Akemu, Whiteman, and Kennedy (2016), the key resources of FTSEs were acquired via professional investors, and its legitimacy process was pushed by powerful stakeholders (e.g. media, government, and large corporations). By contrast, the resource mobilisation and entrepreneurial legitimacy of CBSEs were largely supported by local community members (Hertel, Binder, and Fauchart 2021). Following our taxonomy, these seemingly discrepant findings can be attributed to the fundamentally different value creation approaches between these two business models. Specifically, the establishment of FTSEs requires more capital investment to accumulate necessary physical and intellectual resources (Akemu, Whiteman, and Kennedy 2016), while local legitimacy and related experiences are more accessible from community members in CBSEs (Hertel, Binder, and Fauchart 2021). Here, we do not claim that FTSEs will not mobilise resources from community members or CBSEs cannot obtain capital recourses from corporate investors. Instead, we highlight the importance of acknowledging the diversity of social enterprises and the value of using a common framework to facilitate comparative studies (Nicholls 2010).

Second, our taxonomy of social enterprises could provide opportunities to revisit and extend current theories of social enterprises through more nuanced insights of the core organisational process. For example, our taxonomy may provide additional explanation to the intensively discussed phenomenon of mission drift (e.g. Battilana et al. 2015; Longoni et al. 2019; Pache and Santos 2010; Smith and Besharov 2019). Mission drift describes the risk that social enterprises prioritise customers over beneficiaries in their efforts to generate revenue, and it usually occurs when an organisation pursues incompatible goals simultaneously (Battilana et al. 2015; Ebrahim, Battilana, and Mair 2014). Adding to this phenomenon, our taxonomy suggests that the relationship between customers and beneficiaries of a social enterprise could impact the likelihood of mission drift. Specifically, in business models where target customers are the same as target beneficiaries (e.g. Green Innovation Social Enterprises), positive value spill-overs – the increase in value to third parties beyond a transaction – are automatically gained by beneficiaries (Dahlman 1979; Santos, Pache, and Birkholz 2015). Because profit generation is already aligned with impact creation, the corresponding social enterprise business models are more like commercial models and the risk of mission drift becomes relatively low (Santos, Pache, and Birkholz 2015).

However, in models where customers are different from or just a subset of beneficiaries (e.g. WISEs, FTSEs, and PBSEs), additional managerial efforts, such as training, monitoring, and awareness-raising are required from social enterprises to generate value spill-overs for beneficiaries (Santos, Pache, and Birkholz 2015). As a result, social enterprises may have a higher tendency to focus on the needs of customers who create income, but neglect beneficiaries who request financing (Battilana et al. 2015; Santos, Pache, and Birkholz 2015). The risk of mission drift and eventual failure of these social enterprises become high. Our above interpretation also explains why mission drift is frequently examined in the setting of Contract Manufacturing WISEs where the customer-beneficiary relationship is typically detached (e.g. Battilana et al. 2015; Pache and Santos 2013; Smith and Besharov 2019) compared to GISEs where the customer/beneficiary relationship is aligned. Future research could empirically investigate whether certain social enterprise business models are associated with a higher risk of mission drift and suggest any potential solutions.

Finally, we also contribute to the growing body of work on business model analyses (Fjeldstad and Snow 2018; Laasch 2018; Massa, Tucci, and Afuah 2017; Teece 2018; Zott, Amit, and Massa 2011). Our study holds the empirical promise of using the business model perspective to understand nascent organisational forms in partially commercial and even non-commercial contexts. While the concept of business model has been widely used in understanding value creation of commercial organisations shaped by the homogeneous logic of profit-making, few studies adopt it to explore hybrid organisations that embrace heterogeneous logics (Laasch 2018). Throughout this taxonomy development, the business model perspective has enabled us to capture the essence of the value creation process of social enterprises and identify the key variables that delimit them. Its applicability beyond a purely commercial context has, therefore, been corroborated.

Furthermore, our taxonomy suggests two possible approaches for social enterprises to achieving business model innovation, which refers to the modifications of organisational structures and activity systems for value creation (Amit and Zott 2010). First, our findings suggest that social enterprises may learn from each other and include additional elements to adjust their existing business models. Specifically, compared to traditional single-mission social enterprises (e.g. Contract Manufacturing WISEs, GISEs, & CBSEs), Green Service WISEs, Green Product WISEs, FTSEs, and PBSEs embrace multiple missions when bringing about social change. Identifying such a variety is important, as it generates new insights into the possible evolutionary trajectories of social enterprise business models (Foss and Saebi 2018). Nonetheless, it is still unclear how social enterprises are motivated to embrace multiple missions, what kind of organisational changes they need to adopt a multiple-mission model, and how they benefit from these choices. While more complex models may have the advantage of pursuing social mission efficiently, little is known about whether they generate social value more effectively and perform better than single-mission social enterprises. Future studies may explore these questions through a longitudinal research design at the organisational level.

Second, our study suggests that social enterprises may absorb innovative elements beyond their industry boundaries in face of socio-technical changes and opportunities (Casadesus-Masanell and Ricart 2010; Seelos and Mair 2005). Specifically, Platform-Based Social Enterprises have shown how social enterprises learn from digital solutions, sharing economy, and crowdsourcing to restructure their activity system. This network-based business model not only facilitates the transactions’ speed and efficiency for suppliers, but also reduces customers’ research and increases opportunities for community members to participate in social movements (Busch and Barkema 2022; Rohn et al. 2021). Hence, like commercial organisations, business model innovation also provides social enterprises with a tool to scale their social impact and the long-term value creation (Amit and Zott 2010).

To conclude the above discussions, business model innovation can be regarded as one of the most important drivers of value creation (Amit and Zott 2010; Zott, Amit, and Massa 2011). The trend of social movements and the marketisation of social service sectors in developed countries have encouraged more organisations to embrace a stronger focus on using commercial activities as a means to achieve social impact (Doherty, Haugh, and Lyon 2014; European Commission 2014; Laasch 2018). Therefore, our taxonomy can serve as a reference point for social enterprises to create more viable business models (Mair, Battilana, and Cardenas 2012; Pullman, Longoni, and Luzzini 2018).

Managerial Implications

Our findings also provide several managerial implications for social entrepreneurs and policymakers. For social entrepreneurs, our taxonomy can help them design and execute their value creation processes. The taxonomy specifies seven business modelling approaches for practitioners to structure their own social enterprises. It indicates necessary resources, value propositions, market offerings, and revenue streams with at least seven possible configurations inspired by successful social enterprises with an average tenure of ten years. To a certain extent, our taxonomy provides social entrepreneurs with ‘turn-key’ models that can be implemented directly as well as a useful toolbox to check any potential gaps in their current organisational design. As many social enterprises have access to limited resources to realise their ideas (Hertel, Binder, and Fauchart 2021; Mair and Schoen 2007), practitioners may refer to our study for guidance when making decisions about how to exploit pre-identified opportunities and deploy their resources more appropriately, depending on the type of business model they chose.

For policymakers, a clear mapping of the heterogeneity among social enterprises could facilitate their decision-making process as well as the legitimacy of this nascent organisational form (Mair, Battilana, and Cardenas 2012; Short, Moss, and Lumpkin 2009). Notwithstanding the rapid development of social enterprises globally, policymakers’ limited understanding of social enterprises has been a key barrier to their progress (Bolzani, Marabello, and Honig 2020; European Commission 2014). In particular, several countries (e.g. Finland, Poland, and Sweden) narrowly define social enterprises as WISEs, which has negatively affected the recognition and the legitimate establishment of other types of social enterprises (European Commission 2014; Short, Moss, and Lumpkin 2009). Considering that social enterprises undertake a wide range of activities beyond the scope of WISEs to generate social benefit, our taxonomy can help policymakers to systematically define social enterprises in their policy documents to effectively support their success.

Limitations and Future Research Directions

This study has several limitations implying future research opportunities. First, although we have included a considerable number of cases, social enterprise business models with unique characteristics and small group size could have been omitted from our clustering. For example, microfinance institutions (e.g. Grameen Bank) that provide financial services to low-income individuals or organisations that have limited access to financial resources (Battilana and Dorado 2010; Defourny, Nyssens, and Brolis 2021; Doherty, Haugh, and Lyon 2014) are not identified from the results, as we removed ‘financial resources’ at the variable selection stage and most of them then fall into Cluster F. However, we believe our framework provides a valuable model that can be generalised. By inputting the characteristics of microfinance institutions, scholars can identify the corresponding business model. Thus, we encourage future research to apply our theoretical framework to larger databases and other countries to increase its external validity. In addition, it would be fruitful to explore if and how the institutional context would shape the distribution and configurations of social enterprise business models (Defourny, Nyssens, and Brolis 2021; Erpf, Tekula, and Neuenschwander 2019; Zahra et al. 2008).

Second, our study applies a static business model perspective, neglecting the dynamics of social enterprise business model over time. Effective organisations always adapt their configurations to fit the environment, as the elements of a business model are flexible (Cosenz and Noto 2018; Fjeldstad and Snow 2018; Sharma, Beveridge, and Haigh 2018). Therefore, a dynamic business model perspective reflects an organisation’s goal to redesign itself in order to have a higher chance of success (Johnson, Christensen, and Kagermann 2008). Social enterprises can also iteratively update their business models for resiliency and achieve greater social impact (Sharma, Beveridge, and Haigh 2018; Smith and Besharov 2019). Thus, useful insights into strategy formulation and business venturing would be generated by understanding how social enterprises adjust their model types and add new elements into existing models. Moreover, a dynamic view may help reconcile the discrepancy between ‘a matter of type’ and ‘a matter of degree’ discussions of hybrid organising (Battilana, Besharov, and Mitzinneck 2018, 22). By using longitudinal data and developing a taxonomy of social enterprises based on their evolutionary trajectories, future research can elaborate on the type of social enterprises while at the same time acknowledging their changing degree of institutional plurality. Such studies might also help to better understand why some social enterprises flourish, while other decline or disappear.

Moreover, while social enterprises in our study implement only a single business model in operations, multiple models can be applied in conjunction by individual social enterprises. The original Social Enterprise NL database includes cases that produce both goods and services but aim for different social missions. As the complexity of the social enterprise business model is not the focus of our study, we removed them from the cluster analysis. However, organisations from various industries have been actively adding multiple business models to exploit new opportunities or in response to competitors’ behaviours (Snihur and Tarzijan 2018). Thus, we would encourage future research to explore why social enterprises choose to manage multiple business models, what kind of organisational complexity they may face, and how they are able to effectively combine some of the seven models identified in our study.

Finally, it would also be intriguing to examine the relative effectiveness and long-term success of different social enterprise business models. While our study has shown seven common models of social enterprises, little can be inferred about which model is more effective in generating social and economic value, and under what conditions. Future studies may empirically measure the social and commercial performance of different social enterprise business models to provide insights into how social enterprises develop competitive advantages.

Conclusion

Social enterprises have introduced innovative ways to address persistent social problems (Zahra et al. 2009). With this research, we aim to better understand the heterogeneity among social enterprises’ value creation processes and the underlying mechanisms that lead to such heterogeneity. Taking the business model perspective, we identified seven types of social enterprises through a cluster analysis of 357 Dutch social enterprises. While WISE is the most visible model applied by social entrepreneurs, we also identified other models that actively contribute to social changes, including Green Innovation Social Enterprises, Fair-Trade Social Enterprises, Community-Based Social Enterprises, and Platform-Based Social Enterprises. Our results show that a social enterprise business model serves as an overall reflection of resource configuration, value logic, and industrial justification of social enterprises, while the heterogeneity among social enterprises emerges from their different value creation processes. Additionally, we found that multiple missions can be achieved by a single social enterprise, and social enterprises are sensitive to innovative elements beyond their industry boundaries when solving pressing issues. It is our hope that this study could inspire scholars to pay more attention to the heterogeneity among social enterprises in their research design, results interpretation, and findings comparison processes. We also encourage social entrepreneurs to use our work as a blueprint when starting new ventures and policymakers to recognise the heterogeneity of social enterprises when making political decisions.

Acknowledgement

The authors are grateful to Fan Zhang (University Medical Center Groningen) for the great methodological support and to Katia Mosina for the data collection. The authors would also like to thank Björn Mitzinneck (University of Groningen) for his comments and suggestions on earlier versions of this manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Note

Notes

1 We only consider beneficiaries who can directly benefit from the spillovers of transactions. Beneficiaries who indirectly benefit from the spillovers of a better environment, a more inclusive society, and less social inequality as a whole are beyond the scope of our research.