Foundations of Operational Research and Business Analysis 1

Assignment 2013/14
Author: Thibaut Achard de Leluardière

Abstract:

Looking through the infinite number of theories and models developed in organisations, this assignment aims at finding out the founding principles of a good OR/MS model and general issues encountered in the setting-up of OR interventions. To try out and compare the insights presented, this assignment proposes to study a specific case about OR modelling in Fishery management. Fishery Management is related to the preservation of fish resources and optimisation of catch and profit of this industry, in a context of high-yield practices and increasingly more complex environmental issues present. This case applies to a large and complex system linked to today’s topics issues of sustainable development. In addition, a personal experience of analytical project related to an internship position as assistant project manager in a leading oil company is proposed to illustrate this essay.
This essay concludes by giving recommendations about what could be the characteristics of an ideal portrait of OR/MS model.

Introduction:

In a letter addressed to English universities after the Second World War, general Pile, a popular British officer who commended the Anti-Aircraft Command, claims for men of sciences ‘able to quickly understand complex issues and to find them simple’. Thus, supporting the fast economic growth in Europe after war, operational research (OR) and management sciences (MS) answered needs for better understanding and simplification of problems and arrive at optimal solutions to complex decision-making problems. During the last decades fields of applications of these disciplines have been multiplied, and techniques of operational research were boosted by IT progresses. It brings operational researchers to constantly wonder what requires a good model and a successful MS/OR intervention?

Presentation of case study and personal experience:

The case studied here is an article issued from the European Journal of Operation Research, called Modelling complex human systems: a fisheries example. The article provides two different equation models aiming at simulating fishing behaviours in order to optimize fishing policies. The study is based on data from Canadian fisheries of Nova Scotia, which is one of Canada’s Maritime Provinces. Fishery management modelling is based on the assumption that there is a maximum economic yield to maintain through a system of catches limitation or restricted fishing licences.
The first model is a simple, non-spatial dynamic model for a single specie of fish. It studies the evolution over time of catch rate and fishing efforts in response to revenue and profits obtainable, under the effects of natural fluctuations of fish resources. Different rates of answer are tested. Haddock is chosen over the other species because it is the highest price fish. Thus it ensures causality between fishing efforts and specie density. An equation can modelize the close relationship between predator, fishermen represented by the growth of fishing effort, and the prey population, fishes. The equation is randomly tested with different economic factors related to demand, profitability, and boats in competition, and biologic factors such as the age and weight of fishes, and also the fluctuations of birth rate of the specie, which vary from year to year. The model showed in this article point out the fact that human responses can amplify relatively small annual environmental situations, leading to large, quasi cyclic changes in catch and profit.
The second model adds a spatial factor and includes all kind of species in the analysis provided. This second model attaches more importance on the behavioural aspect of fishermen and the role of information and knowledge to analyse their fish efficiency. It puts the emphasis on the level of rationality they adopt in the way they make decision on where and what to fish. It distinguishes to types of skippers: those adopting a ‘stochasts’ behavourial strategy and ‘cartesians’. ‘Stochasts’ are risk-takers and try to discover new attractive fish zones. ‘Cartesians’ do not take the risk of moving outside of their zone and muster their efforts until they have exploited it completely. The model divides the fishing area in 19 equal spatial zones with different distributions of fish in dollars value. Several random tests are made depending on the presence and competition between ‘stochast’ and ‘cartesians’, and the impact of technologies on the exchange of information between them. This model tends to demonstrate that stochastic behaviours are necessary for the survival of the fishery, and that the efficiency and size of the industry depends on the level of information flows concerning catch.

My personal analytical project is related to an internship I fulfilled at the French Headquarter of ExxonMobil, in the commercial downstream department. I got an experience in an analytical project for commercial purposes, by conducing a portfolio segmentation of a particular type of customers called net buyers. Net buyers were stations owner under contract to be furnished and branded in exclusivity by Esso. The business unit (BU) was in charge of reviewing its commercial efforts more efficiently on these customers in order to optimize the results of this activity. As intern, I was asked to implement a segmentation methodology designed to model the degree of attractiveness for each customer. The information collected for this task involved economical characteristics of customers’ stations such as profitability, risk exposure, delivery cost, trend of volumes and margin, and a qualitative assessment. These information were issued from working sheets of the business team and internal IT databases of the company. The model, based on Excel, was basically structured around two types of weighted criteria: margin of progression and existing revenues. This tool was designed to categorize customers in four categories: harvest, hold, growth and handle. This segmentation was regulated through a logarithmic equation to get a majority of stations, around 60%, in the handle category in order to focus intensive efforts on the three most attractive categories: Hold, Harvest and Growth. On the opposite, handle stations supposed basic services, and minimum efforts to be at term removed from the portfolio.

Characteristics of a good OR/MS model:

A good OR model is linked to the clarity of objectives allocated: what insights the model must bring to the decision-maker about the comprehension of a complex scheme, and does the model succeed in bringing a real added value to decision makers?

To respond to that question, clearly definite purposes have to be dressed up. According to the research works of J. Bryant, A. Taket, C. Richie (Community works, 1994) the results given by a model must be ‘a representation of reality intended to be use to someone in understanding, changing, managing and controlling reality’. These four elements are the founding objectives of a model. In the case analysed, the modelling methods presented give a dynamic picture of how fisheries evolve in a specific region. Thus it confers researchers with a better understanding of the system in order to improve the effects different policies options are likely to have on fishing efforts. According to an article about Management of fisheries and Aquaculture (Journal of the European Operational Research, 2003), management of renewable natural resources is seen as one of the first fields of successful applications of different OR models: ‘The emergence of global, competitive markets is increasing the need to derive efficient production processes to increase productivity. On the other hand, the increasingly more complex environmental issues present additional challenges for OR models’. In this context, the significant role of OR models as a great added value for the purposes of understanding and management of fish resources seems evident.

The second important aspect, which may be the biggest challenge when developing an OR model, is to make it representative of the reality. The difficulty is to take into account the perception that each actor of a system has of a problem, what the British management scientist Paul Checkland calls the ‘Weltanshauung’ (Systems thinking, Systems practice, 1981). In other terms, the model needs to comprehend the whole range of stakeholders, their interest and power and the systemic structure of power interest relationship. In one of its publishing, Holt gives the example of a house that has to be built and involve different specialised workers with different views, requiring different information (A pragmatic guide to Business Process Modelling, 2009). Everyone being part of a system has different interest and priority, which are likely to interact each other.
In the case of the models of Fishery management described above, some fishermen may prefer a long-term strategy of fishing trying to protect the sustainability of their fishing reserves intelligently but they are confronted to other types of rational reasoning due to the competition of other boats. Thus short-term and long-term visions are confronted. The demand of consumer and their reaction to pricing evolution is also part of this system, as well as the level of profit expected by the fishermen.
The difficulty is to take into account all these interaction to give the nearer representation possible of the real world. Giving a description of what is an OR model, Professor M. Pidd (Tools for Thinking: Modelling in Management Sciences, 2010) explains that people have to accept that a model can never be complete and that different model can interpret differently a same real system. For him the modelling tool is a problem finisher, in opposition to a problem solver: it helps the decision process but does not bring one single absolute solution. A good model is not complete as it is a simplification of the reality and concentrate on one delimited system, but it must be representative of part of the reality, clarifying one or several aspects relative to it. Thus, in the case study about Fishery management the operation researches presented are not based on one model, but try simulations through different models. They focus on different dimension of the problem. One model focuses on the fishing effort behaviour according to fish resources fluctuations when the other model focuses on the spatial representation of fishermen activity to guarantee their profits. Both models are right in their representation of the reality and complete each other. None of these models gives a complete view of the system, as it would not be possible to simplify the reality in a way that gather all the information. The first model emphasizes the biological vision and the scientist interests for the long-term evolution of fish resources. The second model put the emphasis on the micro-economic dimension and the fishermen interests, rationally acting in order to ensure their own sustainability.
This brings us to another important feature. A good model must stay as simple as possible in its representation of real-life system. In this regard, if two models both offer a relevant analyse for the purposes of the research, the best one will be the most explicit, using just the right amount of details.
Thus, the first model of the case studied above focus on only one fish specie. It could include in its analyse other species but then the relation between fishing efforts of fishermen and fish resources would be much more complex to be modelled. As the haddock is the most attractive fish, selecting only this specie is sufficient to fulfil the model purposes. The model remains simple and the results bring insights about the whole fishery system.

In A pragmatic guide to Business Process Modelling, Holt lists few other determinant principles in building a good model. Thus the necessity of the empirical aspect is pointed out. All data supplying a model must be taken from reality and objectives, and the elements of the model coherently fitted. At the end the insights provided by a model can also be confronted to the reality to verify its coherence with the reality. To perfect a model and improve trust on it, OR can also integrate a precision tool to consider and estimate uncertainty.
Besides, he explains that the participation of decision makers in building the model is essential. It ensures the model will be designed for their understanding, in other term to be user-friendly.

Thus, for example, in the case of my personal experience in project analysis, I was closely in touch with the area managers forming the net buyers BU. I had to discuss with them about the importance of different elements of the model, and hence, make the model the most accurate and user-friendly possible. I also needed to study their qualitative assessment of customers and I could do that only by interviewing each of them to rank my customers. After working on the model I was also able to explain how to update the segmentation in the future and customize some features if necessary.

Issues in problem structuring, data collection and analysis:
Having in mind the good features of a model, different steps compose the elaboration of an OR/MS model. Problem structuring is the first stage of this process. It is issued from the interaction between organisation and OR analysts point of views to understand the problem. Operational researchers have to work closely with decision makers and stakeholders to understand the problem in order to create an adequate model. Through this work they have to embrace the whole behavioural complexity and interaction between every stakeholders. This issue is intrinsic to the problem structuring. In parallel they must try to simplify the problem as much as possible in order to start the modelling process on clear and strong foundations. Thus, in its book Tools for Thinking: Modelling in Management Sciences, Pidd warn operational researchers against the danger of over complexity and over elaboration in the structuring of the problem. OR analyst are not expert of the company activity, they are just consultant in one particular field. In order to get an accurate overview of the issue to resolve, they need to rely on the experience and skills of stakeholders.
In the case of Fishery management for example, operational researcher study closely the working life of fishermen to understand how they interact each other and how they are influenced by different variables. They also try to define objectives of improvement with environmental researchers to figure out an optimal equilibrium between fishing efforts and resources.

Problem structuring closely depends on the work of data collection. Forrester (Industrial dynamics, 1961) resume that situation, saying that ‘a mathematical model should be based on the best information available but the design of a model should not be postponed until all pertinent parameters have been accurately measured. That day will never come’. Indeed the problem will always need to be restructured looking at the accuracy and comprehensiveness level of data collected. Williams (Management Sciences in Practice, 2008) explains that primary data are more accurate but also expensive and time consuming, as they require studies specially designed for the case of operational researchers. Secondary data raise the issue of consistency and it is much more subjective, as they have been initially collected for a different purpose. The good use of these data is clearly linked to the capacity of analyse of researchers. Moreover reliability issues must be considered. Thus, the age of the data release and the expertise of the authors have to be probed. Wilson (NASA, 1971) point also out the accuracy of the scale measurement used, which can vary a lot from one study to an other. For these reasons the representativity of data must be automatically checked, through analysis of samples for example.
Actually in any cases, data must always been scrutinized cautiously. Landsberger, a professor in industrial management, warn about the issue of the objectivity of data collection, under what he calls the influence of the Hawthorne effect (Hawthorne revisited, 1958). He explains that people surveyed are influenced by the fact that they know they are observed and tend to slightly change their behaviour.

In the analysis of the data, analysts are also dealing with several issues. In the analysis of model results, it is advised to not wait for the last moment but continuously analyse the current progress of the OR process. The quality of the model analysis is the result of regularly assessed upstream work. Eventually an assessment by comparison of results can be led if other models already have been developed in the same field. Last but no least is the conclusion of the intervention. Williams refutes the idea that the analyst comes with a report at the end. Operational researchers have to work with the organization all along the modelling construction work to help its client to better understand what have been done. It is the only way to create a profitable trust relation and the client will provide its experience of the organization to orient the analysis findings in the best way.

Key difficulties with implementation:
Apart from the quality of the model, the implementation stage is the most critical part in the success of a modelling application. It implies the modification of the decision-making process and, hence, important efforts of adaptation from the members of the organisation concerned. This stage is closely related to the concept of management of change (MOC) that can be defined as ‘The way to Minimize resistance to organizational change and maximising organization's benefits through involvement of key players and stakeholders’ (Businessdictionary.com, 2004).

The first difficulty is to convince decision makers of the added value of the model. For this reason communication efforts are necessary to publicize the project and can be very time consuming. The involvement of the executive board, embracing senior managers, to convey the vision of the project to stakeholders is essential to support a model implementation. Forder, a responsible for one of the biggest British OR group stress the fact that ‘without the confidence in OR at senior levels that direct visibility should bring, its contribution will be hampered and the OR function may, indeed, become organisationally vulnerable’. Like for any new product or service, a model must be marketed and sold to the final users and it is even essential for the success of the implementation. Thus, in the Journal of Operation Research Society, Ranyard, Fildes and Crymble denounce ‘a lack of OR champions’ (1997) and ‘lack of visibility: lack of clear image and failure to publicize project successes, particularly at executive board level’ (1998) to justify the failure of OR applications.
In the other side, analysts are dealing with the challenge of fitting to the organisation. They must literally try to integrate the management team during that period of common project if they want to build efficiently a profitable cooperation. Behind this, it is about understanding the culture of operating of the system. It is not an easy thing, considering these people are multi-taking, working on many projects. Abdel-Malek et al. summarizes this idea in The Nature of Managerial Work (1999): ‘The key to communication is understanding the impatience of the client’.

The implementation can also face reticence from stakeholders involved in the system, especially when the insights given by the model could favour decision against their interests. Thus, Pidd talks about employees of a supermarket who risks loosing their job through the closing of company’s plants. Thus, it becomes even more difficult to get the support of stakeholders when their interests are in contradiction with decision made.

In addition a new tool raises training issues. Getting concerned people without technical background familiar with the operational process is part of the management consulting discipline. For all these reasons the partnership relations between OR consultants and their client are vital to ensure the good implementation of a modelling project. According to Rassam (Management Consultancy: A Handbook for Best Practice, 1998), one of the key roles of management consulting is to ‘identify and investigate problems and/or opportunities, recommending appropriate action and helping to implement those recommendations’.

Besides, Pidd points out the danger of ‘changes in the rest of the world’ that can impact the implementation. Considering that reality is dynamic and always moving, it is crucial to monitor the implementation in order to adapt the model if necessary. Hence, people involved in the implementation must regularly ensure the model is staying in adequacy with the possible modification of the system.

A publication of the Project Management Institute (2013) enhances the program evaluation and review technique (PERT) as a simple statistical tool to analyze the tasks involved in completing a project. It evaluates the time needed for each of these tasks in order to stick to an optimal implementation scheduling. This kind of method can be applied in order to structure the necessary initiatives and monitor the objectives of implementation of the model.

During my experience at ExxonMobil I was closely working with the net buyers area managers to implement successfully the model of segmentation. Different meetings were scheduled in order to ensure the success of the project implementation within a period of 6 months. The main points of these meetings were to understand the approach of the BU toward their customer portfolio and agree on the features of the model. In a second hand, the aim was to promote the interest of this model among my colleagues. Indeed they already had an overview of their customer and lot of other more important preoccupations in their work than spending time on an umpteenth project. I needed to catalyse their motivation and efforts to implement this model. I had to convey the idea that this model could bring them more accurate insights to improve the management of their business. Besides, the area managers were attached to their net buyers customers, in spite of the poor results of some of them. For this team, working on this model already meant to accept loosing part of their business. Even if they officially understood and agreed with the project they were not especially keen on having less customers for the profitability objectives of decision makers. My best tools of communication were PowerPoint, regular meeting reminders and conveying my personal motivation to succeed in that task.

In the case studied about Fishery management, we have a tripartite cooperation between biologists, fishermen, and operational researchers. Moreover the system studied here is particularly subject to modification. The system is sensible to variations coming from everywhere as it is related to renewable resources, which imply a large number of environmental and economical factors. For instance, it needs a regular monitoring about the fish resources and price fluctuations. Besides, the difficulty is to convince the Fishery management that a better understanding of their behaviours will allow improving their situation in the long term. The model can not be operated without a free effort from Fishery management to get from the stakeholders to provide the necessary information accurately. Promoting the good effect of operational research in their field is essential to get them cooperating into the project and this is obviously taking time. They will be even harder to convince as the environmentalist researchers are often considered as working against them, preferring environmental issue against fishermen interests. The acceptation and good understanding of this project by all stakeholders is necessary to get results in term of discovering and innovation. Meanwhile, biologist will need to be trained to the understanding of the model and cooperate with operational researcher to clarify their needs for a better comprehension of Fishery management. The lack of results in a first time may be harmful to that cooperation until first palpable insight appear. Communicating about the success of the model in bringing insights will allow the process of change to get implemented.

Conclusion:

An OR/MS model answers to precise objectives of finding, it must be representative of the reality, simple and provide coherent insights of a delimited system. It is the result of a cooperative work with decision makers and it involves a meticulous study of all stakeholders and tangible variables. A continuous assessment at each stage of the modelling building process is the essential to obtain relevant results. The implementation of an OR/MS model requires time and promotional efforts. As for any types of project it has to be realistic in terms of means involved to be decently conducted and the role of management is critical.

References: * - Allen, P.M. And McGlade JM. (1985). Modelling complex human systems: A fisheries example. European Journal of Operational Researches. Available: www.wok.mimas.ac.uk
- Bjørndal T. (2004) OR models and the management of fisheries and aquaculture: a Review. European Journal of Operational Researches. Available: www.sciencedirect.com
- WebFinance team. Definition of Management of change. (2005). Available: www.Businessdictionary.com
- Holt, J. (2009). A pragmatic guide to Business Process Modelling. BCS.

- Bryant, J., Taket, A., Richie, C. (1994). Community Works. Sheffield Hallam University Press. - Checkland P. (1981). Systems thinking, Systems Practices. Wiley. - Pidd, M. (2010). Tools for Thinking: Modelling in Management Science, 3rd Edition. Wiley. - Forrester, J.W. (1999). Industrial Dynamics. WHS Smith. - Williams, T. (2008). Management Sciences in Practice. Wiley. - Landsberger, H. (1958). Hawthorne revisited. Ithaca. - Ranyard, W.R. Fildes R.A., Crymble J.C. (1997 – 1998). Death of an OR group: a Case Study. Journal of the Operation Research Society. - Rassam, C. (1998). Management Consultancy: A Handbook for Best Practice. Philip Saddler.