Chapter 8 Discussion Introduction This chapter reflects on the main findings of the research in terms of its contributions to: the key issues of the case study, questions raised in the e-commerce literature, the methodology of agent-based modelling applied to these issues, and the stakeholder / participatory approach. The latter two are very important aspects of the contribution to knowledge because they are an original combination in methodological approach, and so they shall be the focus of attention in this chapter. Section one returns to the discussion of stakeholder participation which was presented in chapter six, and extends it to consider in further detail the problems with lack of feedback and access to the organisation. Section two considers the extent to which the objectives of the participatory aspect of the case study were attained, and reflects upon the experience of collaborating with industry by considering the usability of the approach. It also develops some personal views on how the research progressed and what was learnt from the experience. Section three compares three alternative approaches to participatory enquiry, and draws some additional lessons from this literature. Section four reflects on the main findings of the research, by revisiting the original research hypotheses, and considering how the case study has developed understanding in these areas. The findings are placed in the context of the earlier literature about the impact of e-commerce, and are used to make some observations on the key themes of intermediation, value chain relations, and the integration of ecommerce systems. In particular the section looks at the issue of how to attain a widespread use from the perspective of the multinational. The contribution to these debates is elucidated, in view of the extent to which the case study findings can be generalised. 222 Section five presents a critique of the thesis, focusing upon the related issues of lack of data and the rather singular perspective of the research. Finally, section six suggests some avenues for further research. 223 8.1 Reflection on the Stakeholder Participation This section reflects on the outcomes of the stakeholder involvement, and positions the contribution to knowledge of this research in terms of participatory methodology. As stated previously in section 6.1, the stakeholders were very important to undertaking the project as they were involved in several key phases. To summarise the research experiences described in that section, it was concluded that: • Viability depends upon the interest of small number of people • Obtaining required access to these people can be problematic • Changing circumstances can curtail the research The first of these points signifies that it may be difficult to get an initial ‘foot in the door’ of the organisation. It illustrates the principle, described by Hammersley and Atkinson (1983, pg. 60), of informal ‘sponsorship’ that serves to validate the presence of the researcher and pave the way for access. Furthermore, as in this case study, the sponsoring individual(s) are likely to be willing to fulfil the stakeholder role. The second point signifies that there may be limited access to those persons and also to the organisation. The nature of the relationship with the sponsor / stakeholder will shape the development of the research in that it can facilitate some kinds of access (for example, to the company’s sales data) and obstruct others (for example, to knowledge about the customers of the company). From the third point it should be recognised that access negotiations are ongoing and changeable, as explained by Hammersley and Atkinson on their chapter on access in ethnographic research: “The problem of access is not resolved once one has gained access to a setting, since this by no means guarantees access to all the data available within it … negotiation of access is therefore likely to be a recurrent preoccupation”. (Hammersley and Atkinson 1983, pg. 76) 224 One aspect of organisational access that was particularly problematic was that of access to feedback from the stakeholders. There are three reasons that can be put forward to explain the lack of feedback in this case study. Firstly, there was a tendency to get sidetracked during these meetings. In particular, I felt that we spent too much time discussing the breadth of what could be included in the investigation rather than focusing on getting results from a small study, which may have been a more productive use of the available time. Though very interesting, the meetings failed to converge upon the practical considerations that likely would have led to a more complete research investigation. This has implications for the conclusions we can draw from the study (see section 8.5). Without a doubt, it is easy to find oneself swapping ideas and stories and trying to make sense of a multifaceted, emerging technology like e-commerce, and I must accept responsibility for allowing conversation to drift from topic to topic. It was interesting for me to hear about the practical problems and perspectives of those involved in setting up these systems, and to see how this sat with the literature produced. The stakeholders were very fond of exchanging anecdotes and stories about their experiences, but were also interested to hear the latest academic views, which were often outside the scope of their learning material and sources of background information. Setting aside these fertile discussions, in order to make good progress with the research these experiences suggest that setting goals for each meeting would be beneficial. Also the researcher must if possible take the initiative and set the agenda for the meeting, in order to make sure those goals are attained. The second reason to explain the difficulty of obtaining feedback is that there were often long lengths of time between meetings. The stakeholders, and myself, would sometimes forget what had been previously discussed. It is very important in participatory projects that everyone knows what the others are doing, and what they have agreed to do themselves. In contrast to many empirical methods such as interviews or surveys, in participatory projects the ‘subjects’ of the research are themselves part of the research team. 225 It was perhaps unavoidable that there are large gaps of time. Lack of man-hours to the project meant that progress was quite slow and a long time had passed between the data collection and the point where the first results were available (a period of 9 months) for example. During this time I did provide a report on the fieldwork analysis – a partial first draft of chapter five of this thesis. However, the stakeholders perhaps felt that they did not know what I was doing during these times. This was a comment dropped by one of them before our final meeting, and I suspect it to be partly a result of these relatively large time gaps. If there were a regular communication between participants, then this problem (of lack of face-to-face contact) would be alleviated, to some extent. This is important because large gaps in time are likely to increase the danger that the participants interests in the project will diminish. Thirdly, and finally, the largest reason for poor feedback was the increasing lack of involvement of the stakeholders because of their time constraints. My research logbook shows several cancelled or rescheduled meetings and lower e-mail and telephone response. In my opinion, this was clearly the most important reason for the lack of feedback. On more than one occasion, the stakeholders themselves said that their own time constraints were the main factor in their lack of involvement, adding that they were regretful of this situation. They did evaluate the model, and provide feedback in a limited sense, which was less than had been expected. We tested the model by discussing the list of drivers and inhibitors of Internet e-commerce that were based on the fieldwork findings, and by discussing the endorsements model that was being developed. We also looked at some of the preliminary results, once the model was working, but this was quite brief. Most of the time was spent discussing the assumptions of the model, which was in fact more appropriate at that stage of development (see section 6.6). Given that one feature of participatory enquiry is the likelihood of access problems mentioned earlier, and given the discussion of the problems of achieving feedback issues, one final principle would recognise the imperative to take all opportunities available to proceed with the work to a satisfactory conclusion. 226 The next section considers what was achieved by the research approach, in terms of informing the stakeholders and bringing insight to their understanding of some aspects of the business. 8.2 Reflection upon the Usability of the Approach One of the objectives was for the research to produce some findings that would be useful to the stakeholders. This section reflects upon the question of how well the approach did what it was hoped in chapter four - what exactly it achieved in terms of helping the stakeholders. Bearing in mind here what was discussed in 8.1, i.e. the lack of involvement on the part of the stakeholders, it is difficult to answer this question. As stated earlier, I did set out to inform stakeholders on the benefits and opportunities associated with the development of e-commerce. However, this was found to be difficult in practice, as result of the lack of access. This became increasingly a problem after late 2002 whilst the model was still under development and the experiments had not yet been carried out. The results from the prototype experiments that were taken back to the company inspired the stakeholders to ask questions and pursue the introduction of further studies (i.e. surveys), which is an indication of their interest and the fact that they value the potential of this kind of work to highlight customer relationship and internal efficiency shortcomings. In addition, they could see how it could lead to MAS applications, for example they were interested in the development of software systems for managing and analysing their e-commerce data. This kind of research and its applications was seen by the stakeholders as something which could give them an advantage in terms of expertise in e-commerce, which was something considered very important within the strategic planning of the company. To clarify, the situation was that the research approach was seen as showing strong potential to be useful to the stakeholders, and indeed it did produce some interesting findings (see later section 8.4 for a summary), though ultimately these findings were 227 not used by the stakeholders or the organisation to inform policy making in the sense of the perspective of action research. Action research is based on the premise that the participants will act upon the findings of the research, that the research and the real-world problem situation are directly entwined, and therefore any intervention capabilities are planned for and made a part of the research design. Viewed from the perspective of action research, the participatory method that was used in this project could have had a much more problem-solving and interventionist focus. Instead there was more of an emphasis placed upon observation, description and reflection than upon choosing particular courses of action. In this sense, the approach has more in common with a soft systems methodology (SSM) approach (Checkland 1981; Checkland and Scholes 1990) than with action research. Hindle, Checkland et al. (1995) discuss differences of research objective focus by comparing social science, action research and soft systems methodology. We return to the comparison with SSM and other participatory approaches in the next section (8.3). The idea of taking action based on the findings of this case study was discussed at the beginning of the project. Furthermore, although it was put forward as part of an ‘evaluation’ in an internally circulated e-mail, the fieldwork was clearly considered by some of the interview respondents as an investigation which may inform some course of action. This can be linked to the literature on ‘models of the researcher’ as ‘expert’ or ‘critic’, that may be an inaccurate expectation on the part of the manager (Hammersley and Atkinson 1983, pg. 75) However, I do not know that any decision-making in the organisation relating to ecommerce has been influenced by the findings. Moreover, it is unlikely to contribute to policy change at the company or within the value chain. This is a result of several factors. Firstly, the collaborators did not evaluate the results in detail, partly due to changing circumstances at the company. Secondly, company policy in e-commerce is still emerging and taking new directions. For example, the company is now reconsidering the possibility of direct sales via the Internet to all customers, an outcome that had been ruled out at the time of the study. 228 Company policy on intermediation might similarly change in line with overall ecommerce strategy. At the time of the study, policy suggested allowing disintermediation of a limited number of less competent / technologically reluctant suppliers. Whilst intermediaries would not be deliberately cut out of the chain, there might be a more subtle process where certain suppliers are favoured over others, with the main objective of improving customer service provision. Such fast-changing policy implies that the model will be increasingly difficult to keep up-to-date and relevant. This is an unfortunate but typical state of affairs where ecommerce policy is concerned – it should be remembered that industry is still in the very early stages of developing Internet EC systems. Furthermore, other research (Caswell 2000) has shown that manufacturers are not very successful at stating their EC policy, though at least most are aware of the likely importance of this technology in the future. These factors underline the difficulty of doing modelling to inform policy making in this area. These difficulties notwithstanding, I do believe that the approach has potential to be used in a more interventionist manner. There are a number of conclusions that can be made from the evidence of this project, which would guide future participatory projects to make them more useful to stakeholders. Firstly, there needs to be developed ways to speed up the development process of models to complete them in a more useful time frame. Secondly, it was apparent that developing better visualisation tools would illustrate the dynamics of the model and make the simulation outcomes more understandable. Thirdly, developing closer links by planning for regular meetings would give the stakeholders more control over the direction of the research, and allow them to fine-tune it as their needs change over the course of time. It should be stated that, as a modeller entering an organisational setting for the first time, I was very much on a learning curve in forming my strategy for doing participatory research. Without taking much heed of other participatory approaches, I conducted the research on e-commerce impacts with the input of the two stakeholders who were acting as my ‘sponsors’. As demonstrated in section 8.1, this rather naïve approach lead to the unfolding of research issues typical of those relating to what are considered the ‘classic’ problems of ethnography, i.e. access and field relations. 229 In view of this fact, consideration of other participatory approaches in the literature has been deferred to this chapter rather than in an earlier section where the case study methodology was described (section 4.2). In the next section, these participatory approaches are explored with a view to suggesting what lessons might be learnt from them that were not considered in the case study, and how these lessons could be absorbed into future agent-based modelling research. 8.3 Other Participatory Approaches The purpose of this section is to make comparisons between the stakeholder approach used in this case study with other established participatory approaches: cognitive mapping, business process modelling (BPM), and soft systems methodology (SSM). Cognitive mapping is a technique used by management and organisational researchers which has its roots in the field of cognitive psychology (Korzybski 1933; Tolman 1948). It aims to offer cognitive explanations for organisational performance, and is used, chiefly, to study the understandings and thinking of managers and how this is linked with knowledge and learning in the organisation. The cognitive map itself describes individual managers’ mental maps (i.e. their internal representations of particular domains of business knowledge) by identifying the sets of concepts which the managers perceive, and their beliefs about how those different concepts are related. Although the relationship between two concepts A and B could be any specified phenomenological relationship (e.g. A is a member of B, or, e.g. if A then not B) it is most common to study the causal or efficacy relationships, i.e. those of the form A influences B or A is an outcome of B. The resulting maps are known as cause maps. Cause maps are of great interest to strategy researchers because they are believed to best reflect decision and policy-making in the organisation. 230 The traditional matrix (Bougon, Weick et al. 1977) technique for eliciting this information involves presenting the manager with a pairwise sequence of elements, and asking for clarification of the efficacy relationship. Constructing the matrix and calculating the indegree values could lead to the discovery of interesting patterns in the data. An alternative technique used by qualitative researchers involves semistructured interviews with management (Eden 1988; Laukkanen 1994). The objective is to probe the manager’s cognitive structures by similarly focusing on questions about the causal relationships amongst a number of key concepts. Both techniques involve creating lists of standard terms to group together sets of key terms or concepts having a virtually identical meaning. The main difference is that with the former technique, the standard terms must be defined well in advance of the fieldwork, whereas the second technique allows a more flexible approach in terms of the ability to derive the standard terms from the data and to analyse the data using more than one set of terms. On the one hand the data analysis technique used in cognitive mapping is rigorous and less dependent upon the interpretative skills of the researcher. However, on the other hand the potential drawback of cognitive mapping is that the technique tends to isolate features of managerial cognition from their social and organisational context. There is plenty of discussion about comparison of different maps across and between organisations and management groups (Laukkanen 1994; Swan and Newell 1998). It is also standard practice to involve stakeholders in the role of evaluating the graphs. There is starting to be more of a focus upon how to mediate between the individual maps generated by the technique, and the normative patterns of the organisational culture. Researchers want to understand the performance of the organisation and how it can be improved through learning and effective use of the knowledge that has accumulated over time. However, because the mapping technique tends to reveal quite well established patterns of managerial thinking, it is difficult to capture learning processes with this approach unless the researcher can be involved during a period of organisational change. 231 Another approach which uses participatory techniques and analyses qualitative data is Business Process Modelling (BPM). In this sense it has similar objectives to the modelling approach described in this thesis, that is, to understand the impact of technological systems. In particular, the objective is to assess the technical feasibility of developing a computer system by analysing the business processes to identify the improvements that could be made and the steps to be taken to achieve them. In other words, to perform a requirements analysis to support organisational change. Firstly, it should be noted that there is a similarity between business process and value chain activity – they are seen in a similar light by both communities because they both stress the value adding nature of these processes. Also both consider activities undertaken jointly across organisations as within their scope of interest. More specific concepts used in BPM are as follows: • The model consists of a number of processes, each of which have defined goals • There are many agents (customers, employees, machines) that carry out different roles within the system In the sense that this analysis is process centred, concerned with dynamic behaviour, and involves different types of agent, there are evident similarities in the BPM perspective as that of ABM. The analysis will involve different groups of stakeholders participating in the research and will incorporate cycles of reformulating and re-evaluating the problem. In this approach, the research design more deliberately makes use of the model to provide a common language for discussion among the participants. It aims to provoke a discussion amongst the various stakeholders that will identify the issues of contention and conflict as well as the points of agreement. This can be the first step to reaching consensus through negotiation. Loucopoulos (2003) describes the qualitative testing of models with the input of many stakeholders as follows: 232 “The deployment of testing parameters at each key process element encourages group brainstorming through which participants could focus on alternative solutions and envisage potential behaviour of the system prior to its implementation.” (Loucopoulos 2003) Since one of the main objectives of BPM is to support the process of reaching agreements among the groups in order to support the collective making of decisions impacting on the use of computer technology in organisations, it shares similarity with the action research methodology discussed in the previous section. The research is aimed at aiding managers to take action within the context of implementing new computer-based systems. SSM (Checkland 1981; Checkland and Scholes 1990) arose out of the realisation that it was very problematic to apply systems engineering principles ("hard" systems theory) to managerial decision-making. It was developed by operational research scientists to resolve situations of high ambiguity due to the subjective views of different stakeholders on what constitutes the research problem (Hindle, Checkland et al. 1995). SSM aims to elicit managers’ “worldviews” and use this understanding to base the discussion and negotiate agreements among stakeholders. This involves constructing models of ‘purposeful human activity systems’ which are proposed solutions for going from a starting state to some desired end state. The system diagrams produced are called “holons” and normally consist of contingent relations represented by arrows linking the activities. The diagrams therefore resemble those used in cause mapping research. However the differences are that the holon is an organisational worldview rather than an individual one, which integrates the views of all stakeholders in a single system diagram, and that it is more prescriptive in terms of describing a desired path, solution, or opportunity for intervention, rather than describing the understandings of managers in any business domain. For these reasons it produces a less descriptively rich (i.e. qualitative) formalisation than that encapsulated in either BPM or cognitive mapping diagrams. In other words, SSM provides tools for making a problem more specific. It relies heavily upon the experienced researcher acting as a ‘process facilitator’ to encourage 233 the stakeholders to engage in negotiation and development of a process model. It is suggested that better results can be obtained when the facilitator has a relatively vague problem definition in the early stages in order not to constrain the project. However, it is also proposed that a formal methodology should be stated in advance of undertaking the research, which describes the aims of the intervention as well as the framework of ideas or theories that will be applied. The central issue of SSM of the contrast between the imprecision of the problem domain and the formal nature of the available techniques is one that is also key to the field of social simulation. In both approaches, the design of models is seen as a disambiguation process that abstracts from the problem definition and research issues (which may be contested) to a limited set of formal statements (a process model or an agent-based model). In the abstraction process much of the complexity of the problem is lost, and this leaves us with the difficult issue of identifying to what extent the model is representative of and relevant to the problem area. Concluding from this discussion of other participatory approaches, ABSS can draw upon these insights to inform future interactions with domain experts or field ‘insiders’. In a similar way to SSM and BPM, interactions that involve identifying (and generating debate about) participants’ conflicting worldviews could become a good focus for eliciting knowledge. Any consensus-building activity that takes place within this context would be useful in developing the conceptual model of business activities, and the engagement would be a bridge to them becoming more strongly engaged in the subsequent stages of evaluation. A further principle that can be taken from this discussion is that stakeholders may require the researcher’s guidance in coming to conclusions and reaching consensus, in far more of an advisory capacity. This could be carried out by ‘chairing’ the discussion. The modeller must possess skills in acting as an interface between the stakeholders and the model as well as facilitating debate and consensus-building amongst the stakeholders in order to fully realise the potential of the approach. This role requires a lot of additional work on the part of the researcher, and skills that need to be attained through experience of working in different research and organisational 234 settings. The difficulties of engaging the stakeholders could be eased by the use of more intuitive model visualisation tools, and by using less jargon, for instance. 8.4 Revisiting the Research Hypotheses In this section the findings of chapters five, six and seven are discussed and evaluated against the research hypotheses presented in chapter four. It outlines the contribution of this research to knowledge about value chain, e-commerce and intermediation issues. In section 4.2.4 it was hypothesised that market intermediation will remain a very important function in the distribution of the manufacturer’s products, based on the nature of the products, which are generic, and the nature of the market, which is diffuse. In chapter five, however, it was seen that there are factors which complicate this analysis, namely that products are increasingly seen as complex in terms of requiring specialist knowledge in developing and marketing them. Management want to promote these ‘specialised’ characteristics of their products and services over the traditional image of ‘off-the-shelf’ products sold by ‘box-shifting’ distributors. E-commerce is part of this developing perspective because it is also seen as enabling greater collaboration in product development as well as facilitating stronger customer relations management. These aims (of management) are not served by market-like interactions but instead require the development of high knowledge and expertise in working with customers. Whilst rejecting market-like arrangements as a solution for these requirements, it should also be recognised that there are problems with vertically-integrated (i.e. hierarchical) operations since, as discussed earlier, it is difficult for the company to achieve an expertise over many different business areas. Drawing upon the fieldwork analysis of the case study, the favoured strategy was found to be one of encouraging more customers to deal through distributors, and encouraging those distributors to provide the required consultancy / technology expertise or advanced e-commerce facilities as their ‘co-specialised asset’. This can 235 be described neither as a strict markets nor a hierarchies solution to the distribution arrangement, but a form of tightly-coupled network whose function is best described by the value chain perspective. The effect of the introduction of e-commerce technology is to make production and distribution processes more complex and knowledge intensive. This increase in complexity is consistent with the finding that a limited amount of disintermediation was anticipated in the wake of the e-business initiative. However, new opportunities were shown to exist in developing expertise in value-adding activities such as improved inventory management systems or specialist technical knowledge of developing market sectors. Whilst the overall market share of intermediation was expected to go up, there would be increasing competition amongst distributors that would make it an increasingly difficult climate to do business without some kind of ‘advantage’. Furthermore, from the manufacturer’s perspective it was thought imperative that the improvements in the transparency of business information should be carried through to the smaller customers. These two factors (the increasing competition and the information imperative) would likely threaten the intermediaries and lead to a reduction in their number. The case study findings suggested that some of the distributors will struggle to reintermediate themselves under a fully enabled e-commerce system, one where electronic orders might potentially be mandated by the manufacturer. On the future of those distributors, there was detected a very pessimistic view on the part of some respondents. Furthermore, in the literature review there were few actual examples of the disintermediation – reintermediation (IDR) cycle in e-commerce where a particular company has successfully managed to adapt. However, one factor which may have been neglected is the difficulty of setting up the longitudinal case studies required to track such adaptations. This could be an area for future research. Experiments with the simulation model looked at the importance of price and technical competence characteristics of the distributors. Comparing across different scenarios, analysis of intermediary performance (market share, profitability and survival rate) suggested that intermediaries profitability is strongly correlated to intermediaries’ technical competence, but it is not correlated to selling price. In 236 addition, agents that do not survive in the simulation always have the lowest technical competence level. Market share is closely related to price, but due to variability in intermediary mark-up, agents with large sales are not necessarily the most profitable. Simulation results seemed to support the analysis of the fieldwork. Some intermediaries having low technical knowledge might not survive in a value chain augmented by Internet technology; increasing competition would make it difficult for them to remain profitable. Bearing in mind the earlier literature on adoption of EDI technology, research questions centring upon diffusion of e-commerce systems were investigated. The fieldwork showed that, most importantly, the onus was on the manufacturer to describe the possibilities of EC to the customers, and to make it attractive for them to adopt. A heavy reliance would be laid upon the different support strategies that were available to the company. The study revealed these incentives or ‘drivers’ of ecommerce, along with a number of inhibitors, largely result from technical and organisational under-preparedness of the customers and distributors. Clearly if the manufacturer does take into account such obstacles to diffusion, the customers will not be able to engage effectively in EC and hence the value chain will under perform. These factors were then later incorporated prominently in the simulation model as customer perceptions and beliefs. The model was intended to be representative of the case study situation, and was reliant on the input of the stakeholders for establishing the validity of the mapping between model and target system. Simulation experiments were carried out to investigate the issues surrounding the e-commerce initiative and to develop understandings of those scenarios. The model was also based on the assumption that social influence plays an important role in the attitude formation of potential users of the e-commerce system. Scenario analysis carried out with the model suggested that a simple customer-customer referrals process can produce a strong effect in influencing agents’ choice of supplier that can ultimately drive high levels of inequality in market share for intermediaries, as well as changes in levels of adoption and use of e-commerce. 237 The results on intermediaries’ market shares were plotted to test for the existence of the Pareto Power Law distribution, a ‘statistical signature’ of many observed market systems, and there was found to be a logarithmic relationship between the cumulative number of units sold and the ordering of intermediaries. Experimental outputs also exhibited s-shaped e-commerce adoption and usage curves, resembling patterns identified in empirical studies of technology diffusion, as well as the actual uptake plotted from statistical logs collected by the stakeholders for the first fourteen months of operation of the electronic mall. In summary, a combination of interpretative fieldwork analysis and agent-based modelling research was used to address the research hypotheses and produce some findings and conclusions to add to knowledge on value chains, intermediation and the impact of e-commerce. Whilst it is not proposed that modelling techniques developed here are sufficiently penetrating to use as replacement for multiple, qualitative case studies, it can be argued that much can be gained from using agent-based modelling in parallel. In essence, this is the argument for multimethodology – a debate which is attracting much attention in the area of management of information systems. In this thesis, agent-based modelling is put forward as one such alternative. Moving on from the discussion of what new knowledge or insight has been gained from this case study, the next section considers the issue of limitations to the approach. 8.5 Limitations of the Research In addition to the main shortcoming, the uncertainty over the extent to which the research would be useful to the stakeholders, discussed in detail in section 8.3, there are other areas where the limitations of the findings should be clarified. 238 A critique of the thesis should focus upon the lack of data to inform the model, such as the basic information about the nature or characteristics of intermediaries and customers. There is also a shortage of information about the behaviour of intermediated customers (i.e. customers who do not have contact with the manufacturer) and the perceptions that drive their behaviour. Since it was not possible to speak to any intermediaries to get their views and those of their customers, this must necessarily lead to a one-sided perspective of e-commerce development. It also resulted in an exploration of possible strategic options for the manufacturer only. There was very little consideration of customer and distributor alternatives, i.e. their responses to manufacturer actions. In general, the lack of information was problematic because it left a number of model assumptions that were less well validated than they might have been. A quantitative survey would have added to the study, and made it easier to quantify some of the parameters, in particular those involved in the sensitivity analysis (i.e. initial perceptions and the cohesiveness / density of the network). Collecting information articulating the perspectives of the business partners would also allow development of the critical theory approach, where conflicts and the power relationships that structure interactions are more fully considered. The study had an essentially singular perspective of employees at the company and so couldn’t capture these aspects. Since only a single case study was considered, that was centred upon one manufacturer, the findings are not safely generalised to development of a pluralistic perspective on the impact of e-commerce. The study has concentrated upon a specific case involving empirical research that adds to knowledge about markets, the value chain, and technology diffusion. To the question of whether the findings could be generalised to other situations, the position taken here is that although similarities were found with other empirical work (see chapter two) across different industrial sectors, the case study itself is very specific. The model developed as part of this case study, is relatively fine-grain and therefore unlikely to be relevant to other cases. The flexibility of the agent-based approach, however, allows the possibility for particular components of the model to be replaced. 239 Though this would require a significant amount of work, in the long run suites of models could be built that would allow a gradual establishment of more general findings relevant to more than one situation. We return to the generalisation issue in the following section (i.e. suggested further research). However, this leaves us with another question: how successfully does the modelling address the research hypotheses? This is a more general issue than that concerning the substantive findings of the research, which was reviewed in the previous section. Here, the interest is in how useful are the answers, and to what extent they are limited? It should be noted that what can be inferred from the model results about the target system, i.e. the policy implications of the model, is largely determined by the level of confidence in the model. On the one hand, it was difficult to give policy recommendations because lack of data with which to calibrate the model. On the other hand, policy prescriptions such as encouraging adoption through various types of intervention strategies and distributed communication amongst customers and distributors, which were supported by the fieldwork analysis as well as by simulation modelling, are reasonable ones to recommend to management. The modelling is limited in the sense that it is not feasible to explore every possible parameter setting but uncertainty must be dealt with through rigorous testing of the results, repeated simulation, statistical analysis, and sensitivity analysis. However, the model may be limited in other ways, not only upon how much trust can be placed in the results. It also was limited in that there were no conflicts apparent in the views of stakeholders which could significantly complicate the research. In one sense it could be considered fortunate that the stakeholders very closely shared views because it made easier the task of formulating objectives suitable for a collaborative project. On the other hand, conflicts amongst the stakeholders are interesting precisely because they bring up issues which can be drawn into the model, by probing them, making them explicit and understanding them through the model development process. Negotiating upon conflicting views of the participants could therefore be a possible use and benefit of the modelling as discussed earlier in section 8.3. An explanation 240 for the lack of conflicts is that it was a result of having a very small group of stakeholders, who possessed essentially unified views. However, with larger groups of stakeholders, the likelihood is that different members hold different views, and therefore there is a greater possibility of conflicts. The ABSS methodology is not sufficiently developed to give guidance on how to manage conflict issues, should they emerge, during the course of a participatory project. It is an important question because it would, at least in part, indicate how the method would scale up to projects involving lots of stakeholders. More attention would need to be paid to conflicts and power relations amongst stakeholders, and the relations between those individuals and the organisation. It would also be likely to cause more confusion as to the project objectives (i.e. what are we going to study and for what purpose – i.e. who is supposed to benefit?) and problem specification. 8.6 Reflection on the Areas Deserving Further Work This chapter will now conclude with some suggestions for further research along these lines of enquiry. Firstly, it has been argued that the research approach worked reasonably well for a single case study and has proved worthy of trial for a bigger study incorporating several cases and models of the impact of e-commerce upon the value chain. Similarities exist across industry in the way business-to-business ecommerce is organised. It would be worthwhile to explore how case alignment might be carried out when several (similar) studies exist, and therefore the opportunity exists for making comparisons and generalisations across these findings. Secondly, it would be valuable to undertake a study incorporating ABSS modelling of multiple perspectives of the value chain. There are several market-types that a company may choose to enter: intermediary controlled and buyer controlled as well as those initiated by themselves. Investigating the strategies of other members of the value chain would allow us to develop further insights into the issues associated with the introduction of EC and its impact upon the value chain. 241 Thirdly, further research should aim to demonstrate the utility of the ABSS approach in developing multi-agent software systems using sociological principles. These principles encapsulate the flexible, robust, scalable, distributed and adaptive qualities that we consider important for systems that are required to operate in complex business environments. Finally, it is clear that new tools for improving visualisation of simulation data would greatly enhance the ability of industrial collaborators to interpret the meaning of simulation results and recognise behavioural patterns. These tools would also undoubtedly have a positive impact on the feedback process between industrial participants and research team. 242 Conclusion The original aspect of the work presented here was to combine qualitative with quantitative insights in the development of an ABSS model. The qualitative approach is encapsulated by the data collected through interviews, and the quantitative by sales and other data. The key problem is to relate the quantitative to the qualitative, and vice versa, i.e. to translate the qualitative data into a form suitable to apply the formal methods of computer modelling, and likewise to interpret numerical outputs of simulation experiments in terms of qualitative descriptions of modelled phenomena. This is an important point, because stakeholder participation requires developing explanations of model behaviour to which the stakeholders can understand and discuss. As discussed in chapter six, the participatory approach was central to the research although it produced problems in obtaining data that stakeholders had initially agreed to provide, in completing the model evaluation process, and in an eventual abrupt ending to the project. All of these problems were linked with the literature on ethnographic research settings, namely the issues of ‘sponsorship’, ‘ongoing access relations’, and ‘models of the researcher’. The lack of information was particularly problematic for model development because the assumptions were less well validated than they might have been. Similarly to SSM and BPM approaches to participatory inquiry, ABSS modelling is a process which forces disambiguation amongst stakeholders in specifying a ‘conceptual model’ of the system under investigation, and the (often hidden) assumptions that constitute it. In addition to simulating the consequences of managers’ decision-making and other scenarios involving dynamic processes, a main strength of the approach would be this disambiguation of the specification of a problem area. In the course of doing this, necessarily the researcher will elicit stakeholders’ subjective positions. It must be recognised then, that the model is not an objective one but, on the contrary, contains a number of subjective interpretations. Nevertheless, the view of the author is that it is possible to claim objectivity on the grounds that there are common ‘threads’ running 243 through the dialogue, and that this type of objectivity may be a desirable ontological position to aim for in the face of stakeholders conflicting worldviews and subjectivity. 244