Determinants of Cloud Computing Adoption in Small and Medium-Sized Enterprises: A Mixed-Methods Study

This is a sample doctoral dissertation prepared by Assignment Help Center to demonstrate the depth, structure, and academic register expected of a PhD thesis in Information Systems. It is a condensed but complete example showing every chapter. All empirical results reported in Chapters 5 and 6 are illustrative and were generated for demonstration purposes only; they are not drawn from a real dataset and must not be cited as evidence. Students should treat this document as a structural and stylistic model, not as a source.

Abstract

Cloud computing has become a defining infrastructure of the contemporary digital economy, yet its diffusion across small and medium-sized enterprises (SMEs) remains uneven and, in many sectors, disappointingly slow. While large organisations have absorbed cloud services into their operating models with relative ease, SMEs continue to face a distinctive constellation of financial, technical, organisational and institutional constraints that shape whether, when and how they migrate. This thesis investigates the determinants of cloud computing adoption in SMEs, integrating individual-level technology acceptance theory with organisational and environmental perspectives to build a more complete explanatory account than either tradition offers alone.

The study adopts a pragmatist philosophical position and a sequential explanatory mixed-methods design. In the quantitative strand, a cross-sectional survey instrument was developed from established scales and, in this illustrative demonstration, administered to a sample of owner-managers and senior IT decision-makers in SMEs. The data were analysed using confirmatory factor analysis and structural equation modelling to test a conceptual framework synthesising the Technology Acceptance Model (Davis, 1989), the Unified Theory of Acceptance and Use of Technology (Venkatesh et al., 2003), the Technology–Organisation–Environment framework (Tornatzky and Fleischer, 1990) and Diffusion of Innovations theory (Rogers, 2003). In the qualitative strand, semi-structured interviews with owner-managers were analysed thematically to explain, contextualise and, in places, complicate the statistical patterns.

The illustrative findings suggest that adoption is driven less by perceptions of technological sophistication than by perceived relative advantage, top-management championing, external competitive pressure and the credibility of vendor and peer signals. Perceived security risk and data-sovereignty concern operate as significant brakes, while cost is revealed to be a more ambiguous factor than the deterministic ‘cost-saving’ narrative of the practitioner literature implies. The qualitative strand foregrounds the role of trust, owner-manager identity and the improvised, resource-constrained character of SME decision-making — dimensions that the variance models capture only partially.

The thesis contributes to knowledge in three ways. Theoretically, it advances an integrated, multi-level model of SME cloud adoption and demonstrates the analytical value of combining acceptance and diffusion perspectives. Methodologically, it illustrates how a mixed-methods design can reconcile explanatory breadth with interpretive depth in adoption research. Practically, it offers policymakers, vendors and SME managers a more differentiated understanding of the levers that move adoption. The concluding chapter reflects candidly on the study’s limitations and sets out an agenda for longitudinal and cross-national research.

Keywords: cloud computing; technology adoption; SMEs; TAM; UTAUT; TOE framework; diffusion of innovations; mixed-methods.

Chapter 1: Introduction

1.1 Background and context

The proposition that computing might be delivered as a metered utility — provisioned on demand, billed by consumption and abstracted from the physical hardware on which it runs — is no longer a speculative vision but the ordinary condition of enterprise information technology. Cloud computing, defined in the influential formulation of Mell and Grance (2011) as a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources, has reshaped how organisations acquire, scale and pay for their digital capabilities. Its three service layers — infrastructure, platform and software as a service — and its deployment variants — public, private, hybrid and community — together constitute a flexible menu from which organisations of very different sizes and capabilities might, in principle, compose an information systems strategy.

For large enterprises, the migration to cloud has been substantial and, in many respects, complete. The economics of elastic capacity, the appeal of shifting capital expenditure to operating expenditure, and the promise of accelerated innovation cycles have proven compelling. Yet the picture among small and medium-sized enterprises is markedly more complicated. SMEs constitute the overwhelming majority of businesses in most economies and account for a large share of employment and value added, and it is precisely for resource-constrained firms that the cloud’s promise of enterprise-grade capability without enterprise-grade capital outlay ought to be most transformative. In practice, adoption among SMEs has been slower, more hesitant and more sectorally variable than the straightforward cost logic would predict (Gangwar, Date and Ramaswamy, 2015; Senyo, Addae and Boateng, 2018).

This gap between the apparent rationality of adoption and the observed reluctance of many SMEs is the empirical puzzle that motivates this thesis. It is a puzzle with real consequences. If cloud adoption confers competitive advantages — and there is reason to believe it often does — then a laggard SME sector risks a widening productivity gap relative to larger, cloud-native competitors. Understanding what actually determines the adoption decision, as opposed to what economic models assume ought to determine it, is therefore a question of both scholarly and practical importance.

1.2 The research problem

The scholarly literature on technology adoption is rich but fragmented, and its fragmentation is part of the problem this thesis seeks to address. Individual-level acceptance theories — most prominently the Technology Acceptance Model (Davis, 1989) and its descendant, the Unified Theory of Acceptance and Use of Technology (Venkatesh et al., 2003) — offer parsimonious, well-validated accounts of why individuals form intentions to use a technology. These models excel at explaining perception-driven acceptance but were developed largely in the context of individual users adopting workplace systems, and they say relatively little about the organisational and environmental conditions under which firms adopt.

Organisational and environmental theories, by contrast — the Technology–Organisation–Environment framework of Tornatzky and Fleischer (1990) and Rogers’s (2003) Diffusion of Innovations theory — foreground firm-level and contextual determinants such as organisational readiness, top-management support, competitive pressure and the characteristics of the innovation itself. These frameworks capture the firm-as-adopter but are frequently criticised for their looseness: the TOE framework in particular specifies broad categories rather than a determinate set of constructs, and its very flexibility has produced a proliferation of studies whose findings do not readily accumulate.

The specific problem, then, is twofold. First, no single dominant framework adequately captures cloud adoption in SMEs, because the phenomenon is simultaneously perceptual (an owner-manager’s beliefs about ease and usefulness matter), organisational (a small firm’s slack resources and leadership matter) and environmental (competitive and regulatory pressures matter). Second, the SME context introduces particularities — the fusion of owner and manager, the scarcity of specialist IT expertise, the salience of trust and personal networks — that the general theories were not designed to accommodate and that quantitative variance models alone struggle to render. There is, in short, a need for an integrated framework and a research design capable of testing that framework while remaining open to the contextual texture that gives SME decisions their distinctive character.

1.3 Research aim, questions and objectives

The overarching aim of this thesis is to develop and empirically examine an integrated, multi-level explanation of cloud computing adoption in small and medium-sized enterprises, and in doing so to reconcile the explanatory ambitions of individual acceptance theory with the contextual insights of organisational and diffusion perspectives.

This aim is pursued through the following research questions:

RQ1. Which technological, organisational and environmental factors significantly influence the intention of SMEs to adopt cloud computing, and what is their relative importance?

RQ2. How do individual-level perceptions (of usefulness, ease of use, risk) and firm-level conditions (readiness, top-management support) jointly shape the adoption decision?

RQ3. How do SME owner-managers themselves understand, narrate and justify their cloud adoption decisions, and what do these accounts reveal that variance models do not?

RQ4. How can quantitative and qualitative evidence be integrated to produce a more complete explanation of SME cloud adoption than either strand affords in isolation?

The corresponding objectives are: to synthesise the relevant theoretical traditions into a coherent conceptual framework and testable hypotheses (addressing RQ1 and RQ2); to operationalise that framework in a survey instrument and test it using structural equation modelling (RQ1, RQ2); to elicit and interpret owner-managers’ own accounts through semi-structured interviews and thematic analysis (RQ3); and to integrate the two strands in a discussion that assesses convergence, complementarity and divergence between them (RQ4).

1.4 Contribution to knowledge

A doctoral thesis is distinguished not merely by the competence of its execution but by its contribution to knowledge, and it is appropriate to state at the outset what this study claims to add.

The first contribution is theoretical. The thesis synthesises four established but rarely combined frameworks into an integrated model that spans the individual, organisational and environmental levels of analysis. Crucially, it does not simply juxtapose them but specifies how constructs drawn from acceptance theory (perceived usefulness, ease of use) mediate the effect of firm-level and environmental antecedents (top-management support, competitive pressure) on adoption intention. In articulating these mediating pathways, the study offers a more mechanistically specified account than the additive models common in the TOE literature.

The second contribution is methodological. Much adoption research is mono-method, and the field has been criticised for an over-reliance on cross-sectional survey designs whose findings, however statistically robust, remain thin on the reasoning and context behind respondents’ answers. By deploying a sequential explanatory mixed-methods design, this thesis demonstrates how qualitative evidence can be used not decoratively but analytically — to explain anomalous statistical results, to surface constructs the survey under-specified, and to test the interpretive validity of the quantitative model.

The third contribution is contextual and practical. By centring the SME, and specifically the owner-manager, the study produces knowledge tailored to a category of adopter that the dominant theories treat as a residual case. Its findings speak directly to vendors designing SME propositions, to policymakers designing adoption-support programmes, and to owner-managers themselves weighing a consequential decision.

1.5 Scope and delimitations

Several boundaries define the study’s scope. It concerns adoption — the decision to take up cloud services — rather than post-adoption assimilation, routinisation or the realisation of benefits, though the concluding chapter argues that these are natural extensions. It focuses on SMEs as conventionally defined by employee headcount and turnover thresholds, and it treats the owner-manager or senior IT decision-maker as the relevant unit of decision, acknowledging that in larger SMEs the decision may be more distributed. The study is cross-sectional in its quantitative strand, a delimitation whose implications for causal inference are addressed directly in Chapters 4 and 8.

1.6 Structure of the thesis

The thesis proceeds in eight chapters. Following this introduction, Chapter 2 reviews the theoretical and empirical literature on technology adoption and cloud computing in SMEs, and identifies the research gap. Chapter 3 develops the conceptual framework and states the hypotheses. Chapter 4 sets out the research philosophy, mixed-methods design, sampling strategy, instruments, analytical procedures, ethical considerations and criteria for rigour. Chapter 5 reports the illustrative quantitative findings, and Chapter 6 the illustrative qualitative findings. Chapter 7 integrates the two strands and discusses their implications for theory and practice. Chapter 8 concludes by restating the contributions to knowledge, reflecting on limitations and proposing directions for future research.

Chapter 2: Literature Review

2.1 Introduction

This chapter establishes the theoretical and empirical foundations of the thesis. It begins by examining the major individual-level theories of technology acceptance, then turns to organisational and environmental frameworks, before reviewing the empirical literature specific to cloud adoption in SMEs. The chapter concludes by articulating the gap that the study addresses. Throughout, the aim is not encyclopaedic coverage but a critical synthesis that motivates the integrated framework developed in Chapter 3.

2.2 Individual-level acceptance theories

#### 2.2.1 The Technology Acceptance Model

The most influential theory of individual technology acceptance is the Technology Acceptance Model, introduced by Davis (1989) and elaborated by Davis, Bagozzi and Warshaw (1989). Adapting the Theory of Reasoned Action from social psychology (Fishbein and Ajzen, 1975), TAM proposes that an individual’s behavioural intention to use a technology is determined by two central beliefs: perceived usefulness, defined as the degree to which a person believes that using a system would enhance job performance, and perceived ease of use, the degree to which a person believes that using a system would be free of effort. Perceived ease of use is theorised to influence perceived usefulness, and both jointly shape attitude and intention.

TAM’s enduring appeal lies in its parsimony and its robust empirical support across an extraordinary range of technologies and settings. Meta-analytic reviews have confirmed the stability of its core relationships (King and He, 2006; Venkatesh and Davis, 2000). Yet the model’s virtues are also the source of its limitations. Its very parsimony means it excludes social and contextual influences; it was designed to explain individual acceptance of workplace systems, not organisational adoption decisions; and its treatment of the technology as a fixed object underplays the negotiated, evolving character of complex innovations such as cloud services. Venkatesh and Davis (2000) responded with TAM2, adding social influence and cognitive instrumental processes, and Venkatesh and Bala (2008) with TAM3, but the fundamental individual-level orientation remained.

For the present study, TAM contributes two constructs of clear relevance — the owner-manager’s perceptions of usefulness and ease of use are plausibly central to SME adoption — while its silence on organisational and environmental determinants signals precisely the incompleteness that motivates an integrated approach.

#### 2.2.2 The Unified Theory of Acceptance and Use of Technology

Venkatesh et al. (2003) synthesised eight competing acceptance models, including TAM, the Theory of Planned Behaviour and the Motivational Model, into the Unified Theory of Acceptance and Use of Technology. UTAUT posits four direct determinants of behavioural intention and use: performance expectancy (closely analogous to perceived usefulness), effort expectancy (analogous to perceived ease of use), social influence (the extent to which important others believe one should use the technology), and facilitating conditions (the degree to which supporting organisational and technical infrastructure exists). Four moderators — gender, age, experience and voluntariness of use — condition these relationships. A later consumer-oriented extension, UTAUT2, added hedonic motivation, price value and habit (Venkatesh, Thong and Xu, 2012).

UTAUT’s advance over TAM lies in its incorporation of social influence and facilitating conditions, both of which are pertinent to organisational adoption. Social influence resonates with the diffusion literature’s emphasis on peer effects, while facilitating conditions gesture towards organisational readiness. Nonetheless, UTAUT remains an individual-level theory. Its facilitating-conditions construct is a perception held by an individual, not an objective property of the firm, and the model does not theorise environmental pressures such as competition or regulation. For SME research, UTAUT usefully broadens the perceptual account but does not, on its own, reach the firm and its environment.

#### 2.2.3 Why individual-level theories are insufficient for organisational adoption

Before turning to firm-level frameworks, it is worth stating precisely why the acceptance tradition, for all its power, cannot carry the explanatory burden alone in the SME setting. Three limitations are pertinent. First, the acceptance models locate causation entirely within the individual’s beliefs, treating the surrounding organisation and market as, at most, sources of the social influence and facilitating conditions that later extensions added. They therefore cannot represent an environmental determinant such as competitive pressure as an antecedent in its own right. Second, they were validated predominantly in settings where individuals adopt a system chosen or mandated by their employer; the SME adoption decision, by contrast, is a discretionary organisational investment made by the very person whose perceptions the models measure, which blurs the individual and organisational levels the models keep separate. Third, and relatedly, acceptance research typically measures intention to use a system that already exists in the respondent’s environment, whereas SME cloud adoption often concerns whether to acquire the capability at all — a decision with a stronger investment and risk character than the usage decisions the models were built to explain. These limitations do not diminish the value of perceived usefulness and ease of use as constructs; they establish that those constructs must be embedded within, not substituted for, a firm-and-environment account.

2.3 Organisational and environmental frameworks

#### 2.3.1 The Technology–Organisation–Environment framework

The Technology–Organisation–Environment framework, articulated by Tornatzky and Fleischer (1990), shifts the level of analysis from the individual to the firm. It proposes that the adoption of technological innovations is shaped by three contextual dimensions. The technological context comprises the internal and external technologies relevant to the firm, including the characteristics of the innovation itself. The organisational context encompasses firm-level attributes such as size, structure, slack resources, managerial support and the quality of human resources. The environmental context includes the industry, competitors, regulatory regime and the availability of supporting infrastructure and service providers.

The TOE framework has been widely applied to enterprise IT adoption, including electronic commerce, enterprise systems and cloud computing (Oliveira, Thomas and Espadanal, 2014). Its principal strength is comprehensiveness: it directs the researcher’s attention to firm and environment as well as technology. Its principal weakness, acknowledged even by its adherents, is under-specification. TOE names categories rather than constructs; it does not stipulate which technological, organisational or environmental variables matter, nor how they combine. Consequently, its empirical application varies widely, and cumulative theory-building is hampered. This looseness is an opportunity as much as a limitation: it invites researchers to populate the three contexts with theoretically justified constructs, which is precisely the strategy this thesis pursues.

#### 2.3.2 Diffusion of Innovations theory

Rogers’s Diffusion of Innovations, most fully developed across successive editions of his canonical work (Rogers, 2003), offers a complementary account centred on the characteristics of the innovation and the social process of its spread. Rogers identifies five perceived attributes of an innovation that influence its rate of adoption: relative advantage (the degree to which an innovation is perceived as better than what it supersedes), compatibility (consistency with existing values, experiences and needs), complexity (the perceived difficulty of understanding and use), trialability (the extent to which it can be experimented with), and observability (the visibility of its results to others). He further describes adopter categories — innovators, early adopters, early majority, late majority and laggards — and the role of opinion leaders and change agents in the diffusion process.

DOI enriches the study of cloud adoption in two ways. First, several of its innovation attributes map onto, and extend, the perceptual constructs of TAM and UTAUT: relative advantage overlaps with perceived usefulness, complexity with ease of use, while compatibility, trialability and observability introduce dimensions the acceptance models omit. Second, DOI’s attention to social systems and opinion leadership speaks directly to the SME reliance on peer networks and trusted intermediaries. Its limitation, for present purposes, is that it is primarily a theory of diffusion across a population rather than of the determinants of a single firm’s decision; it supplies constructs more readily than a determinate causal model.

2.4 Cloud computing adoption in SMEs: the empirical literature

A substantial body of empirical work has examined cloud adoption specifically within SMEs, much of it drawing on one or more of the frameworks reviewed above. Several consistent themes emerge, alongside notable inconsistencies that themselves motivate further research.

First, perceived benefits and relative advantage are recurrently found to drive adoption. Gangwar, Date and Ramaswamy (2015), integrating TAM with the TOE framework, report that perceived usefulness and ease of use, together with organisational and environmental factors, significantly influence adoption intention. Studies in developing-economy contexts, such as Senyo, Addae and Boateng (2018), similarly emphasise the salience of relative advantage while noting context-specific constraints.

Second, top-management support and organisational readiness are repeatedly identified as critical enablers. In the SME context, where the owner-manager frequently is the top management, this construct takes on a distinctive character: the personal disposition, digital literacy and risk appetite of a single individual can be decisive. Low, Chen and Wu (2011), applying the TOE framework to cloud adoption, find top-management support and firm size among the significant determinants.

Third, security, privacy and trust concerns are consistently reported as inhibitors. The relinquishing of direct control over data and infrastructure that cloud entails provokes anxieties about confidentiality, availability, vendor lock-in and regulatory compliance — anxieties that may weigh especially heavily on SMEs lacking the specialist expertise to appraise them. Data sovereignty and jurisdictional concerns have grown more prominent as data-protection regimes have tightened.

Fourth, the role of cost is more equivocal than the practitioner discourse suggests. Although the shift from capital to operating expenditure is often marketed as a cost saving, empirical findings on the effect of perceived cost on adoption are mixed. Some studies find cost a significant driver; others find its effect weak or non-significant once other factors are controlled; and qualitative work suggests that owner-managers frequently reason about cost in terms of unpredictability and hidden charges rather than simple magnitude.

Fifth, external pressure — competitive, from trading partners, and from the wider institutional environment — features prominently in TOE- and DOI-informed studies, reflecting the socially embedded character of adoption. SMEs do not decide in isolation; they respond to what competitors, customers and suppliers are seen to be doing.

Alongside these convergent themes, the literature displays real inconsistencies. The direction and significance of ease-of-use effects vary; the influence of firm size is contested; and cost, as noted, resists a stable interpretation. Oliveira, Thomas and Espadanal (2014) observe that such inconsistencies partly reflect methodological heterogeneity — differences in constructs, measures, samples and contexts. They also, this thesis contends, reflect the limitations of mono-method variance research, which can establish that a relationship holds without illuminating why, or why it sometimes does not.

2.5 Integrated and hybrid models: promise and shortfall

A distinct strand of the literature has responded to the incompleteness of single frameworks by proposing hybrids, and it is instructive to examine why these have only partly succeeded. The most common hybrid pairs an acceptance model with the TOE framework, on the intuition that the former supplies constructs for the technological context while the latter supplies the organisational and environmental scaffolding. Gangwar, Date and Ramaswamy (2015) exemplify this move, and the TAM–TOE integration has become something of a default in cloud-adoption research. A second family of hybrids marries DOI with TOE, exploiting the fact that DOI’s innovation attributes populate the technological context that TOE leaves empty; Oliveira, Thomas and Espadanal (2014) draw on precisely this combination.

These hybrids are genuine advances, but they share a characteristic limitation that this thesis takes as its point of departure. In most integrations, the borrowed constructs are entered into the model as parallel, additive predictors of intention — perceived usefulness alongside top-management support alongside competitive pressure, each with its own coefficient, none theorised to act through another. The result is a longer list of determinants rather than a deeper account of mechanism. Yet the theoretical logic of the constituent frameworks suggests that the relationships are not merely additive. A firm’s competitive environment plausibly shapes an owner-manager’s perception of relative advantage; top-management commitment plausibly operates by making the technology’s usefulness salient. To model these as independent additive effects is to leave the connective tissue between the levels unspecified. The framework developed in Chapter 3 is designed expressly to supply that tissue, by theorising perceived usefulness as a mediator through which contextual antecedents partly transmit their influence — a specification that turns a list of determinants into a layered explanation.

A further shortfall of the hybrid literature is its near-uniform reliance on cross-sectional survey evidence. Even the most carefully specified hybrid, tested on a single wave of self-reported data, can establish covariation but not sequence, and cannot access the reasoning that would adjudicate between rival interpretations of a coefficient. This is not a criticism of any individual study but of a collective methodological monoculture, and it is the second motivation for the mixed-methods design adopted here.

2.6 The research gap

Three interlocking gaps emerge from this review. The first is theoretical: despite the evident relevance of individual, organisational and environmental determinants, the literature remains fragmented across frameworks that are seldom integrated in a way that specifies the mechanisms linking their constructs. Studies that combine TAM with TOE, for instance, often treat the constructs additively rather than modelling how perceptions mediate contextual effects. The second gap is methodological: the field’s reliance on cross-sectional surveys, valuable as they are, leaves the reasoning behind adoption decisions under-examined, and the inconsistencies noted above are unlikely to be resolved by more surveys alone. The third gap is contextual: the SME, and specifically the owner-manager as decision-maker, is under-theorised, the general models having been developed with larger organisations or individual employees in mind.

This thesis addresses all three. It develops an integrated framework that specifies mediating mechanisms (Chapter 3); it employs a mixed-methods design in which qualitative evidence explains and extends the quantitative model (Chapter 4); and it centres the SME owner-manager throughout. The next chapter sets out the framework and hypotheses.

Chapter 3: Conceptual Framework and Hypotheses

3.1 Rationale for an integrated framework

The literature review established that no single theory adequately explains SME cloud adoption. The individual-level models illuminate perception but neglect context; the organisational and environmental frameworks capture context but under-specify constructs and mechanisms. The conceptual framework developed here integrates the four traditions on the premise that adoption is a multi-level phenomenon in which environmental and organisational antecedents influence adoption intention both directly and indirectly through the individual perceptions of the decision-maker.

The framework is organised around the TOE framework’s three contexts, which provide the overarching architecture, while TAM and UTAUT supply the perceptual constructs located within the technological and, partly, organisational contexts, and DOI supplies the innovation-attribute constructs. The central theoretical move is to position perceived usefulness (relative advantage) and perceived ease of use (complexity, inverted) as partial mediators between the contextual antecedents and adoption intention. This specification responds directly to the criticism that TOE studies model determinants additively.

3.2 Constructs and hypotheses

#### 3.2.1 Technological context

Within the technological context, the framework draws on both acceptance theory and DOI. Perceived usefulness / relative advantage captures the extent to which the owner-manager believes cloud adoption will improve the firm’s performance, flexibility and competitiveness. Consistent with the overwhelming weight of prior evidence, it is hypothesised as the strongest proximal driver of intention.

> H1. Perceived usefulness (relative advantage) is positively associated with SMEs’ intention to adopt cloud computing.

Perceived ease of use captures the anticipated effort of adopting and using cloud services. In line with TAM, it is hypothesised to influence both usefulness and intention.

> H2a. Perceived ease of use is positively associated with intention to adopt. > > H2b. Perceived ease of use is positively associated with perceived usefulness.

Perceived security risk captures concerns about data confidentiality, availability, vendor lock-in and compliance. It is hypothesised as an inhibitor.

> H3. Perceived security risk is negatively associated with intention to adopt.

Compatibility, from DOI, captures the fit between cloud services and the firm’s existing systems, workflows and values.

> H4. Perceived compatibility is positively associated with intention to adopt.

#### 3.2.2 Organisational context

Within the organisational context, two constructs are central. Top-management support captures the commitment, advocacy and resource allocation of firm leadership — in the SME, typically the owner-manager. It is hypothesised to influence intention both directly and indirectly by shaping perceived usefulness.

> H5a. Top-management support is positively associated with intention to adopt. > > H5b. The effect of top-management support on intention is partially mediated by perceived usefulness.

Organisational readiness captures the firm’s financial and technological resources and human capital.

> H6. Organisational readiness is positively associated with intention to adopt.

#### 3.2.3 Environmental context

Within the environmental context, competitive pressure captures the perceived intensity of competition and the extent to which rivals are seen to be adopting cloud. Drawing on both TOE and DOI, it is hypothesised to influence intention directly and indirectly through perceived usefulness.

> H7a. Competitive pressure is positively associated with intention to adopt. > > H7b. The effect of competitive pressure on intention is partially mediated by perceived usefulness.

Vendor and peer trust captures the credibility the owner-manager attaches to cloud providers and to the experiences of trusted peers, integrating UTAUT’s social influence with DOI’s opinion-leadership.

> H8. Vendor and peer trust is positively associated with intention to adopt.

3.3 The conceptual model

Taken together, the eight hypotheses articulate a model in which adoption intention is the focal endogenous construct; perceived usefulness is both a determinant of intention and a mediator; and the technological, organisational and environmental antecedents exert direct and, in the cases of top-management support and competitive pressure, indirect effects. The model is deliberately over-identified relative to the simplest additive specification, because its central claim — that context works partly through perception — requires the mediating paths to be estimated. The framework thus reconciles the acceptance and contextual traditions not by juxtaposition but by specifying the mechanism through which they connect.

The following chapter describes how this model was operationalised and tested, and how the qualitative strand was designed to explain and extend it.

Chapter 4: Methodology

4.1 Introduction

This chapter sets out the philosophical foundations, design, procedures and quality safeguards of the study. It begins with the research philosophy and its implications, justifies the mixed-methods design, and then details the quantitative and qualitative strands in turn — sampling, instruments, data collection and analysis. It concludes with the ethical considerations and the criteria by which the rigour of each strand, and of their integration, may be judged.

4.2 Research philosophy: pragmatism

Every empirical study rests on assumptions about the nature of reality (ontology) and of knowledge (epistemology), whether or not these are made explicit. This thesis adopts a pragmatist philosophical position, a stance that has become the customary and well-justified foundation for mixed-methods research (Tashakkori and Teddlie, 2010; Creswell and Plano Clark, 2018).

Pragmatism declines the forced choice between positivism, with its commitment to an objective, measurable reality and hypothesis testing, and interpretivism, with its emphasis on socially constructed, context-bound meaning. Instead, it treats the research question as paramount and holds that the value of a method lies in its usefulness for answering that question. Ontologically, pragmatism accepts that some aspects of the phenomenon — the statistical regularities in perceptions and intentions across many firms — are amenable to measurement, while other aspects — the situated reasoning of an owner-manager — are best approached interpretively. Epistemologically, it endorses methodological pluralism and abductive reasoning, the iterative movement between data and theory.

This position is not a philosophical evasion but a considered fit to the research problem. The problem, as framed in Chapters 1 to 3, is simultaneously about the generalisable determinants of adoption (a question well served by measurement and modelling) and about the meaning and context of adoption decisions (a question well served by interpretation). Pragmatism licenses the integration of both, and it does so without the incoherence that would attend an attempt to combine strict positivism with strict constructivism.

4.3 Research design: a sequential explanatory mixed-methods design

The study employs a mixed-methods design, defined as research that combines quantitative and qualitative approaches within a single investigation to draw on the strengths of each (Creswell and Plano Clark, 2018). Specifically, it adopts a sequential explanatory design, in which a quantitative strand is conducted first and a subsequent qualitative strand is used to explain, elaborate and contextualise the quantitative results.

The rationale for this particular design is threefold. First, it matches the logic of the research questions: RQ1 and RQ2 are answered principally by the quantitative strand, RQ3 by the qualitative strand, and RQ4 by their integration. Second, sequencing quantitative before qualitative allows the survey results — including any surprising or ambiguous findings, such as a weak cost effect — to inform the sampling and questioning of the interview strand, so that interviews probe precisely what the numbers leave unexplained. Third, the design honours the pragmatist commitment to using each method where it is strongest: the survey for breadth, representativeness and hypothesis testing; the interviews for depth, mechanism and meaning.

The point of integration in a sequential explanatory design is deliberate and staged. Integration occurs at the design stage (quantitative results shape the qualitative instrument), at the analysis stage (qualitative themes are mapped against quantitative findings), and at the interpretation stage (Chapter 7 synthesises the two). This is distinct from a convergent design, in which the strands run concurrently and are merged only at interpretation; the sequential structure is chosen precisely so that the qualitative strand can do explanatory work on the quantitative results.

4.4 The quantitative strand

#### 4.4.1 Population and sampling

The target population comprises small and medium-sized enterprises, with the owner-manager or the senior individual responsible for IT decisions serving as the key informant. The sampling frame in an actual study would be drawn from a business register or a comparable directory, stratified by sector and firm size to ensure variation across the technological, organisational and environmental conditions the framework theorises.

A stratified random sampling approach is preferred over convenience sampling to support generalisability and to guard against the over-representation of digitally advanced firms. Sample-size requirements for structural equation modelling were considered at the design stage: conventional guidance recommends a minimum of ten observations per estimated parameter and, for models of moderate complexity, samples comfortably in excess of two hundred (Hair et al., 2019). The illustrative analysis in Chapter 5 assumes a realised sample of n = 312 usable responses, a figure consistent with these requirements and with comparable published studies.

#### 4.4.2 Instrument development

The survey instrument was developed by adapting validated multi-item scales from the source theories, following established scale-adaptation procedure. Perceived usefulness and perceived ease of use items were adapted from Davis (1989); social-influence and facilitating-condition items informing vendor and peer trust from Venkatesh et al. (2003); innovation-attribute items (compatibility) from the Rogers (2003) tradition; and top-management support, organisational readiness and competitive pressure items from established TOE operationalisations (Low, Chen and Wu, 2011; Gangwar, Date and Ramaswamy, 2015). Adoption intention was measured with a standard multi-item behavioural-intention scale.

All perceptual items were measured on a seven-point Likert scale anchored from ‘strongly disagree’ to ‘strongly agree’. Each construct was measured reflectively with at least three items to permit assessment of internal consistency and to support the measurement model. The draft instrument was subjected to expert review for content validity and to a pilot with a small number of owner-managers to check clarity, wording and completion time; minor refinements followed.

#### 4.4.3 Data collection

In an actual study, data would be collected via an online questionnaire distributed to the sampled firms, with follow-up reminders to improve the response rate. Steps to mitigate common-method bias would be built into the design: procedural remedies such as guaranteeing respondent anonymity, separating predictor and criterion items, and using varied scale anchors, followed by a statistical check (Harman’s single-factor test and, ideally, a marker-variable technique) at the analysis stage (Podsakoff et al., 2003).

#### 4.4.4 Analytical strategy

The quantitative analysis follows the standard two-stage approach to structural equation modelling recommended by Anderson and Gerbing (1988). In the first stage, a confirmatory factor analysis assesses the measurement model — establishing convergent validity (via factor loadings, composite reliability and average variance extracted), internal consistency (Cronbach’s alpha and composite reliability) and discriminant validity (via the Fornell–Larcker criterion and, as a stricter test, the heterotrait–monotrait ratio). In the second stage, the structural model is estimated to test the hypothesised paths, with model fit assessed against conventional indices (the comparative fit index, the Tucker–Lewis index, the root-mean-square error of approximation and the standardised root-mean-square residual). Mediation hypotheses (H5b, H7b) are tested using bootstrapped indirect effects. Descriptive statistics and correlations are reported first to characterise the sample.

4.5 The qualitative strand

#### 4.5.1 Sampling and participants

The qualitative strand employs purposive sampling to select owner-managers whose firms vary along theoretically salient dimensions — adopters and non-adopters, different sectors, and, informed by the survey, firms whose responses exemplified or contradicted the quantitative patterns. This maximum-variation logic is intended not to achieve statistical representativeness, which is not the purpose of qualitative sampling, but to secure analytically rich variation. The illustrative study assumes a sample of 18 interviews, a number consistent with the point at which thematic saturation is commonly reported in qualitative interview studies of this kind.

#### 4.5.2 Data collection

Data were, in an actual study, collected through semi-structured interviews of approximately forty-five to sixty minutes, guided by an interview schedule derived from the conceptual framework and refined in light of the survey findings. The semi-structured format balances comparability across interviews with the openness needed to pursue unanticipated lines of reasoning. Interviews would be audio-recorded with consent and transcribed verbatim.

#### 4.5.3 Analytical approach

Interview data were analysed using reflexive thematic analysis following the six-phase procedure of Braun and Clarke (2006): familiarisation with the data, generation of initial codes, searching for themes, reviewing themes, defining and naming themes, and producing the report. The analysis combined deductive coding, informed by the conceptual framework, with inductive coding open to constructs the framework did not anticipate. This blended orientation is deliberate: it allows the qualitative strand both to speak to the quantitative model and to exceed it.

4.6 Ethical considerations

The study was designed in accordance with established research-ethics principles and would require institutional ethics approval before any data collection. Informed consent would be obtained from all participants, who would be provided with a plain-language information sheet describing the study’s purpose, their voluntary participation, their right to withdraw, and the handling of their data. Confidentiality and anonymity would be protected through the pseudonymisation of firms and individuals; interview recordings and transcripts would be stored securely and access-restricted. Because the research involves business decision-makers rather than vulnerable populations, the ethical risks are modest, but the commercial sensitivity of some information — for instance, about a firm’s IT weaknesses — warrants careful anonymisation. Data would be managed in compliance with the applicable data-protection regime.

4.7 Rigour and trustworthiness

Rigour is assessed against criteria appropriate to each strand. For the quantitative strand, validity and reliability are addressed through the measurement-model assessment described above — convergent and discriminant validity, composite reliability — and through attention to common-method bias and to the statistical-conclusion validity of the structural estimates. External validity is supported by the stratified sampling approach, while the cross-sectional design’s limits on internal (causal) validity are acknowledged frankly.

For the qualitative strand, trustworthiness is evaluated against the criteria of Lincoln and Guba (1985): credibility (via prolonged engagement with the data, and member checking where feasible), transferability (via thick description enabling readers to judge applicability), dependability (via an audit trail documenting analytical decisions), and confirmability (via reflexivity about the researcher’s influence on interpretation). A reflexive journal would document the researcher’s assumptions and their evolution.

For the mixed-methods design as a whole, the study attends to the additional criterion of integrative rigour — the quality with which the strands are combined. This is pursued through explicit joint displays mapping quantitative findings against qualitative themes, and through candid reporting of divergence as well as convergence, in the recognition that the analytical value of mixing lies partly in the productive tensions it surfaces.

4.8 Researcher positionality and reflexivity

Consistent with the pragmatist and, in the qualitative strand, interpretively informed orientation of the study, it is necessary to make explicit the researcher’s position and its potential influence on the inquiry. A researcher is not a neutral instrument but an active participant in the construction of qualitative meaning, and the credibility of the thematic analysis depends in part on acknowledging this. In an actual study the researcher would document, in a reflexive journal maintained throughout data collection and analysis, their prior assumptions about SME technology adoption — for instance, a plausible initial expectation that cost would dominate owner-managers’ reasoning — and would track how the data confirmed, qualified or overturned those assumptions. The illustrative finding that cost is experienced as uncertainty rather than magnitude is precisely the kind of result that reflexive practice is designed to make visible rather than to suppress in favour of a prior expectation.

Reflexivity also bears on the relationship between the two strands. Because the researcher who designed and interpreted the survey also conducted and analysed the interviews, there is a risk that the qualitative interpretation is unduly shaped by prior commitment to the quantitative model — that themes are read as confirmations of hypotheses rather than encountered on their own terms. The blended deductive–inductive coding strategy, and the deliberate search for divergence as well as convergence described in Section 4.7, are the principal safeguards against this confirmation-seeking tendency. The point is not to eliminate the researcher’s influence, which is neither possible nor desirable in interpretive work, but to render it accountable.

4.9 Summary of the methodology

The methodology may be summarised as a pragmatist, sequential explanatory mixed-methods design in which a theory-testing survey, analysed by structural equation modelling, is followed by a theory-elaborating interview study, analysed by reflexive thematic analysis, with integration pursued at the design, analysis and interpretation stages and rigour assessed by criteria appropriate to each strand and to their combination. The two chapters that follow report the illustrative results of the two strands in turn.

Chapter 5: Quantitative Findings

> Illustrative results. All figures in this chapter are illustrative and were generated for demonstration purposes. They do not derive from real data collection and must not be cited as empirical evidence. They are presented to show the form, sequence and interpretation of results in a structural-equation-modelling study.

5.1 Introduction

This chapter reports the illustrative results of the quantitative strand. It proceeds from sample description, through the assessment of the measurement model, to the structural model and the hypothesis tests, before summarising the pattern of supported and unsupported hypotheses.

5.2 Sample characteristics and preliminary analysis

The illustrative analysis assumes 312 usable responses from an initial distribution, after the removal of incomplete and inattentive cases. The illustrative sample is distributed across firm sizes, with roughly three in five respondents from small firms (fewer than 50 employees) and the remainder from medium-sized firms, and across sectors including professional services, retail, manufacturing and information and communication. Respondents are predominantly owner-managers and IT-responsible senior staff, consistent with the key-informant design.

Preliminary screening (illustrative) indicated no problematic patterns of missing data and acceptable univariate normality, with skewness and kurtosis values within conventional thresholds. A Harman single-factor test extracted no dominant factor accounting for a majority of variance, providing preliminary reassurance that common-method bias is not severe, though this test is treated as indicative rather than conclusive.

Table 5.1 presents illustrative descriptive statistics for the principal constructs.

Table 5.1 — Illustrative construct descriptives (seven-point scales)

The illustrative means suggest that respondents perceive cloud computing as useful and enjoy reasonable top-management support, while readiness and trust sit somewhat lower and perceived security risk registers at a moderate level — a configuration consistent with the ambivalence noted in the literature.

5.3 Measurement model

The measurement model was assessed via confirmatory factor analysis. Illustrative standardised factor loadings all exceeded the 0.70 threshold, with the lowest at 0.71. Composite reliability values for all constructs fell between 0.84 and 0.92, comfortably above the 0.70 criterion, and Cronbach’s alpha values were similarly satisfactory. Average variance extracted (AVE) for every construct exceeded the 0.50 threshold, ranging illustratively from 0.58 to 0.74, supporting convergent validity.

Discriminant validity was supported by the Fornell–Larcker criterion, with the square root of each construct’s AVE exceeding its correlations with other constructs, and by heterotrait–monotrait ratios all below the conservative 0.85 threshold. The illustrative overall fit of the measurement model was acceptable (CFI = 0.96, TLI = 0.95, RMSEA = 0.045, SRMR = 0.041).

Table 5.2 — Illustrative reliability and convergent validity

5.4 Structural model and hypothesis tests

The structural model was estimated and displayed illustrative fit within accepted bounds (CFI = 0.95, TLI = 0.94, RMSEA = 0.052, SRMR = 0.056). The model illustratively explained a substantial share of variance in adoption intention, with R² = 0.61, and a moderate share of variance in perceived usefulness, R² = 0.44.

Table 5.3 reports the illustrative standardised path coefficients and the associated hypothesis outcomes.

Table 5.3 — Illustrative structural paths and hypothesis outcomes

The illustrative mediation tests, using bootstrapped indirect effects (5,000 resamples), indicated a significant partial mediation of top-management support through perceived usefulness (H5b: indirect β = 0.08, 95% CI [0.04, 0.13]) and of competitive pressure through perceived usefulness (H7b: indirect β = 0.07, 95% CI [0.03, 0.12]), with both direct effects remaining significant.

5.5 Summary of quantitative findings

The illustrative pattern is coherent and, in several respects, instructive. Perceived usefulness emerges as the strongest proximal driver of intention, as hypothesised, and functions as a mediator carrying part of the influence of top-management support and competitive pressure — evidence for the framework’s central mechanistic claim. Perceived security risk operates as a significant brake. Compatibility, readiness, competitive pressure and vendor and peer trust each contribute, while organisational readiness, though significant, is the weakest of the supported paths.

The single unsupported hypothesis is notable: the direct effect of perceived ease of use on intention (H2a) is non-significant, even though ease of use significantly shapes perceived usefulness (H2b). This suggests that, for these illustrative respondents, ease of use matters instrumentally — by informing judgements of usefulness — rather than as an independent determinant of the decision. This is precisely the kind of finding that the qualitative strand is positioned to explain, and to which Chapter 6 returns.

Chapter 6: Qualitative Findings

> Illustrative results. The themes, patterns and paraphrased participant perspectives in this chapter are illustrative and were composed for demonstration purposes. No real interviews were conducted; no verbatim quotations from real individuals are reproduced. The chapter shows how a thematic analysis would be presented and how it would be used to explain the quantitative results.

6.1 Introduction

This chapter presents the illustrative findings of the qualitative strand, organised around four themes developed through reflexive thematic analysis of the (illustrative) interviews with 18 SME owner-managers. The themes are: (1) usefulness as a business, not a technology, judgement; (2) trust, control and the anxiety of dependence; (3) the owner-manager as the decision; and (4) the improvised, resource-constrained character of the decision process. Each theme is discussed in relation to the quantitative findings it helps to explain.

6.2 Theme 1: Usefulness as a business judgement

The most pervasive illustrative theme was that owner-managers appraised cloud computing not as a technology but as a business proposition. When they spoke of usefulness, they spoke of accessibility of data across sites, of the freedom to scale without buying servers, of resilience during disruption, and of the ability to compete with larger rivals — rarely of technical features in the abstract. Usefulness, in these accounts, was thoroughly instrumental and thoroughly concrete.

This theme directly illuminates the quantitative finding that perceived usefulness is the strongest driver of intention and that ease of use exerts no significant direct effect. In the illustrative interviews, ease of use surfaced only as a component of the usefulness calculation: a system that was hard to use was, for that reason, less useful, but ease was never valued for its own sake. Owner-managers reasoned, in effect, that they did not adopt technology because it was easy but because it did something worth doing. This maps cleanly onto the supported H2b (ease of use shapes usefulness) and the unsupported H2a (ease of use does not independently drive intention), and offers a mechanism for a result that the numbers alone leave puzzling.

6.3 Theme 2: Trust, control and the anxiety of dependence

A second strong illustrative theme was ambivalence about relinquishing control. Participants described a felt tension between the convenience of cloud services and an unease at entrusting their data and continuity to a third party they could not directly oversee. Concerns clustered around confidentiality, the fear of being unable to operate if a provider suffered an outage, the difficulty of switching providers once committed, and — for firms handling personal or regulated data — the location and legal jurisdiction of that data.

This theme corresponds to and enriches the significant negative effect of perceived security risk (H3). But it also extends beyond the survey construct. Where the quantitative instrument measured risk as a perception of hazard, the interviews revealed risk as entangled with trust and identity: owner-managers who trusted a particular provider, often because a respected peer had vouched for it, experienced the same objective risk as tolerable. This resonates with the supported effect of vendor and peer trust (H8) and suggests that trust and risk are not independent but dialectically related — a nuance the additive variance model cannot represent.

6.4 Theme 3: The owner-manager as the decision

A third illustrative theme was the extraordinary centrality of the individual owner-manager. In these small firms, the adoption decision was not the output of a committee or a formal appraisal process but, very often, the disposition of one person whose personal digital confidence, prior experiences, risk appetite and even sense of professional identity determined the outcome. Some participants embraced cloud as an expression of a modern, forward-looking self-image; others resisted it as alien to how they had always run their business.

This theme explains the significant effect of top-management support (H5a) and its partial mediation through perceived usefulness (H5b) in a way the numbers cannot. Top-management support, in the SME, is not a distinct organisational variable acting on an individual decision-maker; it very nearly is the decision-maker. The mediation finding — that leadership works partly by shaping perceptions of usefulness — takes on a concrete meaning in the illustrative interviews: a digitally confident owner more readily perceives cloud as useful and, perceiving it so, adopts. The construct that the survey treats as organisational is revealed to be, at this scale, deeply individual.

6.5 Theme 4: The improvised, resource-constrained decision

The fourth illustrative theme was the makeshift character of the decision process. Lacking dedicated IT staff and the slack to conduct formal evaluations, owner-managers described relying on trusted intermediaries — accountants, sector peers, a friend ‘who knows computers’ — and on incremental, low-commitment experimentation. Cost figured in these accounts, but not as the simple magnitude the practitioner literature assumes; participants worried about unpredictability, creeping subscription charges and the difficulty of forecasting spend, rather than about the headline price.

This theme speaks to the modest, though significant, effect of organisational readiness (H6) and to the framework’s inclusion of vendor and peer trust (H8), and it offers an explanation for the equivocal treatment of cost in the wider literature. Cost is not unimportant to these owner-managers; it is, rather, experienced as uncertainty, a qualitative texture that a single Likert item measuring perceived cost cannot capture. The reliance on trusted intermediaries, meanwhile, gives empirical flesh to the diffusion-theoretic notion of opinion leadership operating within resource-constrained social systems.

6.6 Summary of qualitative findings

The illustrative qualitative strand does more than corroborate the survey. It explains the anomalous ease-of-use result, reframes security risk as entangled with trust and identity, reveals top-management support to be almost synonymous with the owner-manager, and re-specifies cost as experienced uncertainty rather than magnitude. In each case the qualitative evidence supplies mechanism and meaning where the variance model supplies only association. Chapter 7 draws the two strands together.

Chapter 7: Discussion

7.1 Introduction

This chapter integrates the quantitative and qualitative strands to answer the research questions and to articulate the study’s contribution to theory and practice. It first synthesises the two bodies of illustrative evidence, then discusses the theoretical implications, the practical implications, and the points at which the strands diverge as instructively as they converge.

7.2 Integrating the two strands

Answering RQ4 — how quantitative and qualitative evidence can be integrated — requires more than reporting the strands side by side; it requires reading each in the light of the other. A joint display is helpful here. Across the four qualitative themes and the eight hypotheses, the relationship between the strands takes three forms: convergence, complementarity and divergence.

Convergence is evident in the central role of perceived usefulness. The survey identifies it as the strongest driver; the interviews explain that usefulness is understood as a concrete business judgement. Both strands agree, and the agreement strengthens confidence in the finding. The significant effects of top-management support and competitive pressure similarly converge with the qualitative account of decisive owner-managers responding to what rivals are seen to do.

Complementarity is where the integration is most productive. The non-significant direct effect of ease of use (H2a), read alongside its significant effect on usefulness (H2b), is a bare statistical pattern until the qualitative theme of ‘usefulness as a business judgement’ supplies the mechanism: ease matters only in so far as it feeds usefulness. Likewise, the significant risk and trust effects are complemented by the qualitative revelation that risk and trust are dialectically entangled, and the top-management-support finding is complemented by the discovery that, in the SME, this construct is very nearly individual. In each instance the qualitative strand does explanatory work that the quantitative strand set up but could not perform alone. This is the design working as intended.

Divergence, finally, is present and analytically valuable. The survey treats cost, in so far as it is captured, and readiness as discrete organisational variables; the interviews suggest that cost is experienced as uncertainty and that readiness, in a firm without IT staff, collapses into the owner-manager’s personal competence and network. These divergences do not undermine the study; they mark the boundaries of the variance model and point towards a re-specification of constructs for the SME context — a contribution in its own right.

The relationship between the strands can be set out more systematically in a joint display, the standard device of mixed-methods integration. Table 7.1 arrays the principal quantitative findings against the qualitative themes that bear on them and classifies the resulting inference.

Table 7.1 — Illustrative joint display: integrating the two strands

The joint display makes visible what a narrative alone can obscure: that the study’s most valuable inferences arise not where the strands merely agree but where the qualitative evidence supplies the mechanism behind an otherwise bare or puzzling coefficient. It is the complementarity cells, not the convergence cells, that carry the distinctive payload of the mixed-methods design.

7.3 Implications for theory

The study’s theoretical contribution, previewed in Chapter 1, can now be stated more precisely. First, the integrated framework is vindicated in its central mechanistic claim: contextual antecedents — top-management support and competitive pressure — influence adoption intention partly through the individual perception of usefulness. This finding advances the TOE tradition beyond additive determinant lists by demonstrating a mediating pathway, and it advances the acceptance tradition by locating its perceptual constructs within an organisational and environmental architecture. The two traditions are shown to be not rivals but levels of a single explanation.

Second, the study contributes to the theorisation of the SME as a distinctive adoption context. The illustrative finding that top-management support is nearly synonymous with the owner-manager, and that organisational readiness collapses into individual competence, suggests that constructs developed for larger organisations require re-specification when applied to the smallest firms. The owner-manager is not merely a decision-maker within the firm; at this scale the individual and organisational levels of analysis partially fuse, a theoretical observation with implications for how adoption models should be specified for micro and small enterprises.

Third, the study contributes a refined understanding of the risk–trust relationship. Rather than treating perceived risk and trust as independent additive terms, the integrated evidence suggests they are dialectically related: trust does not merely offset risk but transforms how risk is experienced. This points towards a moderated or reciprocal specification in future models, and it connects the adoption literature to broader theories of interorganisational trust.

7.4 Implications for practice

For cloud vendors, the illustrative findings counsel a reorientation of SME propositions away from technical sophistication and headline price towards demonstrable business relative advantage, predictability of cost, and the cultivation of trust through peer references and transparent security assurances. Because ease of use matters chiefly as an input to usefulness, marketing that foregrounds simplicity in the service of concrete outcomes is likely to outperform marketing that foregrounds simplicity as such. Because trust travels through peer networks, vendor strategies that mobilise satisfied SME customers as references may be disproportionately effective.

For policymakers and business-support bodies, the centrality of the owner-manager and the reliance on trusted intermediaries suggest that adoption-support programmes should target owner-managers’ digital confidence and should work through the intermediaries — accountants, sector associations, local networks — on whom SMEs actually rely, rather than through generic information campaigns. Addressing the experienced uncertainty of cost, perhaps through clearer cost-modelling tools, may also lower a barrier the deterministic cost-saving narrative overlooks.

For SME owner-managers themselves, the study offers a mirror: it suggests that the adoption decision is more shaped by personal disposition and trusted opinion than a purely rational appraisal would imply, and that deliberately widening the circle of consulted expertise, and appraising relative advantage explicitly, may improve the quality of a consequential decision.

7.5 Reflection on the mixed-methods contribution

It is worth pausing on what the mixing itself achieved. Had this study been quantitative alone, it would have reported a robust but partly puzzling model — a non-significant ease-of-use path, a significant but under-explained risk effect, an organisational construct whose meaning at SME scale would have been assumed rather than examined. Had it been qualitative alone, it would have offered a rich account of owner-manager reasoning without the capacity to establish which factors matter across the population or to estimate their relative weight. The sequential explanatory design delivered what neither could: a model whose parameters are estimated and whose mechanisms are understood. This methodological demonstration is itself among the thesis’s contributions.

Chapter 8: Conclusion

8.1 Introduction

This concluding chapter restates the study’s contributions to knowledge, reflects candidly on its limitations, and sets out an agenda for future research. It closes with a brief final reflection on the significance of the work.

8.2 Summary of the study and its answers to the research questions

The thesis set out to develop and examine an integrated, multi-level explanation of cloud computing adoption in SMEs. In answer to RQ1, the illustrative quantitative findings identified perceived usefulness as the dominant technological driver, alongside significant effects of compatibility, top-management support, organisational readiness, competitive pressure and vendor and peer trust, and a significant inhibiting effect of perceived security risk. In answer to RQ2, the study showed that individual perceptions and firm-level conditions jointly shape adoption, with perceived usefulness partially mediating the influence of top-management support and competitive pressure — evidence that context works partly through perception. In answer to RQ3, the illustrative qualitative findings revealed that owner-managers understand usefulness as a concrete business judgement, experience risk as entangled with trust and identity, embody the top-management-support construct almost personally, and decide through improvised, resource-constrained processes. In answer to RQ4, the integration of the strands produced an explanation richer than either afforded alone, combining estimated parameters with understood mechanisms.

8.3 Contributions to knowledge

The study makes three contributions, now restated in their fully developed form. Theoretically, it advances an integrated framework spanning the individual, organisational and environmental levels, demonstrates its central mediating mechanism, and re-specifies key constructs — particularly top-management support and readiness — for the SME context in which the individual and organisational levels partially fuse. It further contributes a dialectical, rather than additive, conception of the risk–trust relationship.

Methodologically, it demonstrates, in an exemplary sequential explanatory design, how qualitative evidence can be deployed analytically to explain anomalous quantitative results and to re-specify under-determined constructs, offering a template for adoption research seeking to move beyond mono-method limitations.

Practically, it furnishes vendors, policymakers and owner-managers with a differentiated, mechanism-aware understanding of the levers that move SME cloud adoption — relative advantage over technical sophistication, cost predictability over headline price, and trust mobilised through the peer networks and intermediaries on which SMEs genuinely rely.

8.4 Limitations

Intellectual honesty requires a frank account of the study’s limitations, several of which qualify its claims. First, and most fundamentally, the empirical results reported here are illustrative; a real test of the framework awaits genuine data collection, and the confidence expressed in the findings is contingent on that. Second, even in an actual study, the quantitative strand’s cross-sectional design would preclude strong causal inference: the paths estimated are associations consistent with the theorised causal structure, not demonstrations of it, and reverse or reciprocal causation cannot be excluded. Third, reliance on a single key informant per firm, while appropriate to the SME context, introduces the risk of informant bias and common-method variance, mitigated but not eliminated by the procedural and statistical remedies described. Fourth, the focus on adoption intention rather than realised adoption leaves the intention–behaviour gap unexamined. Fifth, the sampling and setting bound the transferability of the findings; adoption dynamics may differ across national, sectoral and regulatory contexts not represented here. Finally, the qualitative strand, however rich, reflects the interpretations of a single analyst, and the reflexive safeguards described reduce but do not remove the influence of the researcher’s standpoint.

8.5 Directions for future research

These limitations chart an agenda for future work, and the agenda follows directly from the boundaries the study has drawn around itself. A first and enabling step, of course, is to conduct the empirical test that this illustrative document only simulates: to administer the instrument to a genuine stratified sample, estimate the model on real data, and interview real owner-managers, so that the framework’s plausibility is converted into evidence. Only then do the more ambitious extensions become meaningful. The most pressing of these is longitudinal: tracking SMEs across the adoption process would permit stronger causal inference and would open the intention–behaviour gap, and the subsequent stages of assimilation, routinisation and benefit realisation, to examination. A second direction is cross-national and cross-sectoral comparison, testing whether the integrated framework, and particularly the fusion of individual and organisational levels, holds across regulatory and cultural contexts — data-sovereignty concerns, for instance, may weigh differently under different data-protection regimes. A third direction follows from the risk–trust finding: future models should test moderated or reciprocal specifications of the risk–trust relationship rather than additive ones. A fourth direction is to examine the intermediaries — accountants, peers, sector bodies — as active agents in diffusion, perhaps through network-analytic methods that the present firm-level design cannot accommodate. Finally, the re-specification of adoption constructs for micro and small enterprises invites dedicated instrument-development work, producing measures native to the SME rather than adapted from large-organisation scales.

8.6 Concluding reflection

Cloud computing offers SMEs a rare prospect: enterprise-grade capability without enterprise-grade capital. That the promise remains unevenly realised is not, this thesis has argued, primarily a matter of technology or even of cost, but of perception, trust, leadership and the improvised reasoning of owner-managers deciding under constraint. By integrating the theories that have too often been kept apart, and by combining the methods that have too often been used in isolation, the study has sought to render that decision more fully intelligible. The integrated, multi-level, mixed-methods account it advances is offered not as a final word but as a firmer foundation for the longitudinal, comparative and mechanism-sensitive research the phenomenon deserves.

If there is a single lesson to carry from the thesis, it is a methodological one with substantive consequences. The determinants of adoption are neither purely psychological nor purely structural; they are the psychological refracted through the structural and the structural made real in the psychological. A research design that can hold both — that can measure the regularities and interpret the reasoning — is not a luxury but a necessity for phenomena of this kind. The contribution of this work, at its most general, is to have shown what such a design can see that a narrower one cannot, and to have left, for those who come after, a framework and a method worth putting to a genuine empirical test.

References

Anderson, J.C. and Gerbing, D.W. (1988) ‘Structural equation modeling in practice: a review and recommended two-step approach’, Psychological Bulletin, 103(3), pp. 411–423.

Braun, V. and Clarke, V. (2006) ‘Using thematic analysis in psychology’, Qualitative Research in Psychology, 3(2), pp. 77–101.

Creswell, J.W. and Plano Clark, V.L. (2018) Designing and conducting mixed methods research. 3rd edn. Thousand Oaks, CA: Sage.

Davis, F.D. (1989) ‘Perceived usefulness, perceived ease of use, and user acceptance of information technology’, MIS Quarterly, 13(3), pp. 319–340.

Davis, F.D., Bagozzi, R.P. and Warshaw, P.R. (1989) ‘User acceptance of computer technology: a comparison of two theoretical models’, Management Science, 35(8), pp. 982–1003.

Fishbein, M. and Ajzen, I. (1975) Belief, attitude, intention and behavior: an introduction to theory and research. Reading, MA: Addison-Wesley.

Fornell, C. and Larcker, D.F. (1981) ‘Evaluating structural equation models with unobservable variables and measurement error’, Journal of Marketing Research, 18(1), pp. 39–50.

Gangwar, H., Date, H. and Ramaswamy, R. (2015) ‘Understanding determinants of cloud computing adoption using an integrated TAM–TOE model’, Journal of Enterprise Information Management, 28(1), pp. 107–130.

Hair, J.F., Black, W.C., Babin, B.J. and Anderson, R.E. (2019) Multivariate data analysis. 8th edn. Andover: Cengage Learning.

Henseler, J., Ringle, C.M. and Sarstedt, M. (2015) ‘A new criterion for assessing discriminant validity in variance-based structural equation modeling’, Journal of the Academy of Marketing Science, 43(1), pp. 115–135.

King, W.R. and He, J. (2006) ‘A meta-analysis of the technology acceptance model’, Information & Management, 43(6), pp. 740–755.

Lincoln, Y.S. and Guba, E.G. (1985) Naturalistic inquiry. Beverly Hills, CA: Sage.

Low, C., Chen, Y. and Wu, M. (2011) ‘Understanding the determinants of cloud computing adoption’, Industrial Management & Data Systems, 111(7), pp. 1006–1023.

Mell, P. and Grance, T. (2011) The NIST definition of cloud computing. Special Publication 800-145. Gaithersburg, MD: National Institute of Standards and Technology.

Oliveira, T., Thomas, M. and Espadanal, M. (2014) ‘Assessing the determinants of cloud computing adoption: an analysis of the manufacturing and services sectors’, Information & Management, 51(5), pp. 497–510.

Podsakoff, P.M., MacKenzie, S.B., Lee, J.-Y. and Podsakoff, N.P. (2003) ‘Common method biases in behavioral research: a critical review of the literature and recommended remedies’, Journal of Applied Psychology, 88(5), pp. 879–903.

Rogers, E.M. (2003) Diffusion of innovations. 5th edn. New York: Free Press.

Senyo, P.K., Addae, E. and Boateng, R. (2018) ‘Cloud computing research: a review of research themes, frameworks, methods and future research directions’, International Journal of Information Management, 38(1), pp. 128–139.

Tashakkori, A. and Teddlie, C. (eds.) (2010) SAGE handbook of mixed methods in social and behavioral research. 2nd edn. Thousand Oaks, CA: Sage.

Tornatzky, L.G. and Fleischer, M. (1990) The processes of technological innovation. Lexington, MA: Lexington Books.

Venkatesh, V. and Bala, H. (2008) ‘Technology acceptance model 3 and a research agenda on interventions’, Decision Sciences, 39(2), pp. 273–315.

Venkatesh, V. and Davis, F.D. (2000) ‘A theoretical extension of the technology acceptance model: four longitudinal field studies’, Management Science, 46(2), pp. 186–204.

Venkatesh, V., Morris, M.G., Davis, G.B. and Davis, F.D. (2003) ‘User acceptance of information technology: toward a unified view’, MIS Quarterly, 27(3), pp. 425–478.

Venkatesh, V., Thong, J.Y.L. and Xu, X. (2012) ‘Consumer acceptance and use of information technology: extending the unified theory of acceptance and use of technology’, MIS Quarterly, 36(1), pp. 157–178.