Content area
The Royal Navy's manpower planning system represents a highly complex queue which is affected by many variables, and therefore it is essential to understand the influence of these variables and the interactions among them. As real experimentation was impractical and infeasible, a computer-based simulation was developed to model the system which needs to be studied. This paper illustrates how computer simulation and experimental design was applied to identify the key risk variables within the manpower planning system at the UK's Royal Navy.
Marcus Blosch: Marcus Blosch is IT Consultant (Research Fellow) at the Technology and Organisations Research Group, Portsmouth Business School, University of Portsmouth, Portsmouth, UK.
Jiju Antony: Jiju Antony is Senior Teaching Fellow, Warwick Manufacturing Group, International Manufacturing Centre, University of Warwick, Coventry, UK.
Introduction
Experimental design (ED) is a powerful technique which enables the effects of several process or system variables to be studied or investigated simultaneously and efficiently leading to an increased understanding of the system (Montgomery, 1991). It is a direct replacement of traditional "one variable at a time" approach to experimentation, where we vary only one variable at a time, keeping all other variables in the experiment constant. The fundamental difference between the classical approach of experimental design and the traditional approach is that the latter would not allow one to study the interactions among the key variables. In this paper, the authors demonstrate an application of ED in identifying the key variables which cause risk in the Royal Navy's manpower planning system. Though many people are familiar with the word "risk", the word itself can be a bit ambiguous. According to Rowe, risk is defined as "the potential for unwanted consequences of an event or activity" (Rowe, 1977). As businesses become more intricate and change more rapid, the so-called "gut feel" approach to risk management may no longer be adequate. It is important to identify the key sources of risk before it is managed. However, a thorough identification of key sources of risk depends upon a greater understanding of the system. The application of ED provides a systematic and positive approach to identify the key sources of risk in the Royal Navy's manpower planning system.
Experimental design and its applications in the service industry
ED has been widely accepted in both European and US manufacturing companies over the last 15 years for improving product and process quality. There appear to be very few case studies which have been carried out in ED for improving a service-based process. Some noticeable reasons are:
- service process performance is difficult to measure with accuracy;
- service process performance depends a great deal on the behaviour of the human beings involved in delivering it; and
- difficulties exist in the selection of appropriate quality characteristics for the service process.
The following are the potential applications of experimental design in the service sector (Antony and Antony, 1998):
- comparing competitive strategies on the development of new products;
- identifying the key process or system variables which influence the system or process performance;
- minimising the time to respond to customer complaints;
- minimising errors on service orders;
- reducing the service delivery time to customers (e.g. restaurants, banks, etc.); and
- reducing the length of stay in an emergency room in hospitals and health-care institutions.
Royal Navy manpower planning
The purpose of this section is firstly to provide background information to build a fuller understanding of the issues involved within naval manpower planning. For the sake of clarity many detailed technical abbreviations have been omitted and explanations of the involved issues have been made more accessible.
Background
A key component of recent government policies has been the changes initiated to the ways in which public administration was conducted - a particular aim has been to foster economy, efficiency and effectiveness. This has been conducted in a series of initiatives, each with its own title, but all geared towards the common aim of improving government businesses that were seen as lacking the commercial imperative to meet tight financial and product quality standards.
A central component in this initiative was the formation of agencies. By early 1997, some 99 agencies had been formed from areas as diverse as the prison service to the Royal mint. On 1 July 1996, the director general's naval manning organisation became a "next steps defence agency" known as the Naval Manning Agency (NMA). Figure 1 illustrates the structure of the Naval Manning Agency. The NMA employs some 290 personnel and has an operating budget of around Pounds 10 million. It provides a centralised personnel administration function for all elements of the naval service; that is the Royal Navy, Royal Marines, Queen Alexandra's Royal Naval Nursing Service, Naval Personnel Families' Service and all reserves. Amongst its responsibilities, the NMA plans the future manpower size and structure required to meet the operational needs of the fleet, advises its owner on targets for additions to the trained strength, plans the career progression of naval personnel, administers redundancy programmes when required, selects suitable personnel for promotion and career courses and, finally, and of most immediate importance, appoints/drafts personnel to meet the needs of the fleet.
The aim of the NMA is to "ensure that sufficient manpower is available on the trained strength and effectively deployed in peace, crisis, major crisis or war". To meet this aim, the NMA have the following strategic objectives:
- accurate manpower planning;
- effective deployment of manpower;
- career management; and
- advice to ministers, Parliament and public.
The first two of these objectives could be described as paramount in an operational sense - to ensure that the Royal Navy comprises the right mix of specialisations and grades to perform the myriad of operational and support tasks required and then ensure that these people are where they are needed and not elsewhere. Accurate planning considers the evolving manpower requirements of the Royal Navy, matches these to known and projected wastage rates and provides targets for promotion, career courses and additions to the trained strength. Deployment aims to provide suitably skilled and experienced manpower to individual billets, jobs, reducing unfilled billets and surpluses to a minimum and being reactive to changes in customer demand for manpower. The NMA aims to optimise the career development of the individual commensurate with his or her aspirations and abilities. Advice forms the component of accountability required of the agency.
Key issues in naval manpower
Within the naval manpower system, the following key issues are generally considered (Blosch and Antony, 1998):
- Naval personnel are input into the system at the lowest level and "grown" to meet demand in terms of specialisation and rank. This places an emphasis on accurate long-term forecasting, as for senior ranks the time taken for them to be trained and gain the right amount of experience can be considerable. It takes, for example, at least 15 years of experience for an individual to reach the rank of captain.
- The system involves a cyclic sea-to-shore job structure; time spent at sea is followed by time ashore.
- High degree of specialisation: within the Royal Navy there are approximately 90 seaman branches, (a "branch" contains specific job and skill areas, for example weapons engineering) and 25 officer branches. The possibility for interchangeability between branches is low. This is further complicated by the rank structure which further subdivides branches.
- The system must also account for "regeneration", that is, if there is a requirement for extra manpower caused by a crisis, the system must be able to allow, for example, mothballed ships to be brought into service and provided with personnel. The level to which extra personnel are included within the system is a policy decision.
- Manpower wastage from the system, particularly in the form of premature voluntary release (PVR), has a significant effect on manpower. PVR is effectively people resigning from their contracts.
Figure 2 illustrates the typical manpower cycle in the Royal Navy.
The cyclic nature of the normal career creates a number of problems mainly associated with queuing effects. As will be appreciated "seamless" relief of an individual as he or she moves on to their next posting is not always possible, and as a result there are at times "gaps". A "gap" is a job not being filled by a competent and qualified person. This effect is made worse by factors such as variation in event times, shortages of particular categories of manpower or system inefficiency. In order to attempt to alleviate these problems, an allowance is made for the manpower numbers required for actual jobs at sea or shore. This allowance takes two forms; firstly for training, and secondly for "margin". The training allowance takes into account both career training, such as given on promotion, and job training to prepare an individual for a new job. The margin is an allowance for predictable events such as leave.
The complement billets, jobs to be done and the allowances are added together to form the manpower "requirement", to which manpower numbers - the "strength" - are, as far as possible, matched. However, due to the nature of the system, there are a number of problems experienced by the Navy, these being (see also Figure 3):
- gapping: no suitably trained and qualified personnel are available for a job;
- margin over use: the actual margin use is over what is allowed for; and
- training quota under use: there is often an underuse of this allowance - it is unclear at this point if this is due to it being incorrectly calculated or, due to gapping, manpower being withheld and not allowed to go for training.
The problems highlighted above have a significant impact. Many modern ships are "lean-manned", meaning run with reduced levels of manpower. On that basis there is an increased sensitivity in terms of operational capability if jobs are gapped. Margin over use is damaging in two ways. Firstly, much needed manpower is unavailable to the system. Secondly, this incurs a significant cost - money which cannot be spent elsewhere.
The NMA recognised that there was a need to minimise the gaps at sea, ashore (if possible) and the margin usage to ensure that sufficient manpower was available for upward flow in the system. This is a highly complicated system with a large number of interacting variables. It was not possible to undertake a statistical analysis on historical data as this was not available. The Navy's central computer, the naval manpower management information system (NMMIS), was transactional. On this basis the Navy decided to proceed by building a computer-based simulation of the manpower planning system.
Computer-based simulation
It is possible, in the analysis of complex systems, either to use statistical methods on historical data or to utilise computer-based simulation (Neuse, 1998). As a complex system, such as the Navy's manpower planning, the size of the data set required in order to achieve any degree of statistical validity would require monthly data to be gathered over many years. There are a number of limitations with this approach, the first and most obvious being that, in this case, the data were not available. Secondly, even if such a data set were available, it would be of questionable value as the organisational structure and operating procedures had changed many times during the period.
For example, initiatives such as "front line first", which placed emphasis on front-line forces at the expense of support operations, was a radical change in structure. Other initiatives, of which there are many examples, include "lean manning" of ships where the number of personnel required on board was reduced to the bare essentials. These effectively created a "new" Navy each time, and so historical data would be flawed. Due to these considerations, and its unavailability, the Navy decided that a simulation approach was required.
The advent of low-cost, high-power computers and simulation software has been an important advance in making simulation more accessible (Oakshott, 1997). It is no longer necessary to write simulation applications in programming languages such as FORTRAN as many are available in easy-to-use "drag and drop" formats, taking away much of the complexity in setting up simulations (for example see http://www.std.com/vensim/brochure.html which is one of many such programs).
Simulation models have been successfully used in a wide range of applications: risk analysis (Vose, 1997); project management (Raferty, 1994); finance (Getts, 1989); business process re-engineering (Kim and Kim, 1997); and in many other areas.
A simulation model is most often a mathematical model of a "system" operating within an "environment". The "system" represents an area of interest which acts and interacts with factors both internal and in the wider "environment". Computer simulation does not necessarily invoke a purely systems-based perspective on a problem area, though this is often used in approaches such as "systems dynamics" (Larkey et al., 1997). Other approaches, such as "game theory", may equally well be simulated.
As will be appreciated, the quality of the simulation relies on the simplifying assumptions made of the system and its environment, which are often required to reduce the complexity to manageable levels. A general rule in developing simulations is to build a simplified model which contains the major mechanisms under study, and then test these. Complexity may be added gradually until the required level of accuracy on an agreed benchmark is reached. The model may then be used for analysis.
It is necessary to have a theory for understanding how the area of interest operates, and this is perhaps the most important aspect of building simulations. The mechanism and "rules" by which the model operates needs to be clearly stated and agreed by the experts in the area.
Once this model mechanism has been identified, the variables involved will often become apparent. In most cases it will not be possible to include all of these, and decisions will have to be made to omit some of the variables. Often many of these variables are highly correlated and therefore can be identified on the basis of statistical analysis. However, where this is not possible careful consideration needs to be given to variable selection and a logical basis for inclusion or omission set up. It is important that variables included in the model are representative of their real-world counterparts. For most variables this will be, for example, range, mode and median values, probability distribution and so on (Saunders 1995).
A rigorous testing regime for the model is also required and its accuracy and predictive ability must be assessed. In most cases the best test of this is using an historical data set to ensure that the model is a representative of the real world. This includes ensuring that the variable values and ranges are within acceptable tolerances. Some models may only be stable - that is provide accurate results - with input variables within certain ranges. It is important to test the model with input variables at the extremes of their likely ranges to ensure that the results are within accepted tolerances.
Once a simulation has been developed and tested there remains the problem of analysing the results produced by the simulation. The model provides a relationship between the input and output variables. There will often be interactions between input variables which may not be linear; as a result, it is difficult to evaluate the results. Some method of studying these interactions between the variables is required. Under these circumstances, experimental design is of great value (Antony, 1998).
Development of the rating supply chain model (RSCM)
As the Navy's manpower situation was unique in many ways, it was decided to develop a computer-based simulation. This was done using the "Microsoft Access" database to hold the data and a "Microsoft Visual Basic" application written to manipulate the data and provide the graphical user interface.
The rating supply chain model (RSCM) was developed as a representation of the ratings cycle from sea to shore over a period of time for a particular branch. The model is essentially a queuing model which represents the flow of ratings around the normal career cycle of sea and shore postings, with intermediate events such as recruitment, promotion and wastage.
The Navy already operates a number of highly sophisticated forecasting and simulation models which were developed and run for them by the Defence Analytical Statistics Agency (DASA). These models, however, did not contain any representation of the sea-to-shore cycle at the heart of the Navy's manpower system, and worked at the aggregate rather than the individual level.
In building the model it was decided to focus on ratings, as their drafting and promotion tends to be more to requirement than for officers, and thus would be easier to model. The model itself would be limited to one branch and account for the sea-to-shore cycle, recruitment, promotion and wastage; it would work at the individual level, and would run over a number of years.
The model takes as input the following data:
- (1)Requirement: the manpower requirement over the period broken down into each individual rank and sub-branch. This details the jobs at sea and ashore for each rank.
- (2)Strength: the manpower strength at the start of the run.
- (3)Recruitment: the model allows three types of recruitment:
- free, where the model recruits as required;
- forced, where a certain number of recruits are forced into the system; and
- capped, which is essentially free recruitment with an upper limit set.
- (4)Wastage: profiles are entered for three types of wastage:
- premature voluntary release, where individuals resign from the Navy;
- natural wastage, better known as death, dishonour or disease; and
- time expiry, where an individual completes their career.
- (5)Length of draft: this is the length of time spent in a particular location such as at sea, ashore or training, in months.
- (6)Margin times: this is the time spent in the margin in months for each of the margin elements.
- (7)Minimum time in rank: this is the minimum time that an individual must spend in a rank before promotion.
- (8)Promotion bonus: this is a bonus which is given to an individual to account for the distribution in times for promotion. Approximately 12 per cent of individuals are given a "bonus", either positive or negative, to their time in current rank.
- (9)Minimum time ashore: each individual is guaranteed a certain period ashore after completing a tour of duty at sea. The amount of time depends upon their particular branch and rank, but in general the more senior the individual, the less time at sea they will spend.
The model is stochastic in that all the duration variables have a distribution applied. Early versions of the model used a triangular distribution with an exponential tail, but this was found to give a significant number of individuals with durations much longer than one could reasonably expect. As a result, this was replaced with a triangular distribution where it was possible to define the minimum, mean and maximum points.
Essentially, the model moved each individual from sea to shore as part of a normal career, with periods spent in the margin or training, and applied promotion and wastage. The model attempts to capture all the key identifiable variables within the manpower system and model the way they work. There is, naturally, noise in the system which cannot be modelled, for example, ship movements which prevent smooth drafting, or the different approaches of the various drafting desks and so on. Figure 4 illustrates the ratings supply chain model.
The model was tested primarily by running it against a known branch and seeing if its results were reflected. The input variables were, as far as possible, taken from historical data held by the Defence Analytical Statistics Agency (DASA), though, as this was not at the individual level, a number of assumptions were made. A number of runs were made against various branches until the model was accepted by the Navy.
Risk analysis of naval manpower planning
Though most people are familiar with the word risk, the word itself can, however, be ambiguous. It is important therefore to establish a common definition of the word risk at the outset. The concept of risk essentially contains three components:
((1)) an unwanted event;
((2)) the events impact; and
((3)) probability of the event occurring.
The purpose of risk analysis is therefore to identify the potential sources of risk, estimate the impact of these risks and finally to manage these risks, i.e. risk identification, risk analysis and risk management. This is in accord with Recher's (1983) principles of maximising expected value, avoiding catastrophe and ignoring remote possibilities. The extent to which risks are identified must accord with the "unwanted event" to which they are related.
There are three stages to risk analysis:
((1)) identification, which aims to discern the risks;
((2)) ranking, which aims to quantify the risks; and
((3)) management, which aims to eliminate, reduce or provide mechanisms for managing risks.
The Naval Manning Agency was concerned to accurately provide manpower, and felt that this could be more effectively done by managing the risks involved.
The risk analysis is documented elsewhere (Blosch and Antony, 1998). The central risk factor which the Navy wished to control was gaps at sea, which directly affected its operational capability. The risk analysis also identified the following key variables which were felt to influence the availability of manpower to be drafted to sea:
- sea-shore ratio (represented by "S");
- length of time spent at sea (represented by "L"); and
- the ratio between the number of jobs at sea and the number of jobs ashore (represented by "B"). This is also called billet ratio.
Other variables, such as recruitment and wastage, are also of importance, but it was felt that these were of secondary importance and therefore could be considered at a later stage of this research project. This decision was made on the basis that in many instances branches were seen to have sufficient or even surplus manpower and yet still exhibited gaps at sea. The next task then was to design an experiment by which the influence and interactions of these variables could be studied.
Role of experimental design in risk analysis
Virtually no case studies were found on the application of experimental design in risk analysis or risk assessment. The experimental design approach was chosen to understand the system and the most dominant variables which cause gapping at sea. In order to evaluate the potential of experimental design, it was initially decided to select only three variables: sea-shore ratio (S); length of sea draft (L); and billet ratio (B). It was also decided to study the nature of interactions among these variables (Antony and Kaye, 1988). It was decided to choose three levels for each factor as the authors believed that some non-linear effects of the variables were being exhibited. In order to minimise the mathematical complexity and also the calculations involved in the approach, Taguchi's Orthogonal Array (OA) design (Taguchi and Konishi, 1987) was selected for the experiment. Table I illustrates the list of variables and the interactions of interest for study.
As it was possible to run the model at no extra cost, it is proposed to perform 27 experimental trials (based on 3[sup]3 = 27). The uncoded design matrix showing all the possible combinations of variables at their respective levels is shown in the Appendix. The response of interest was the gapping at sea. Having obtained the data based on the simulation study, the first step was to compute the average gapping at each level of the factors chosen for the study. The results are shown in Table II.
For studying the interaction effects, the average margin at each combination of factor level was determined. An interaction occurs between two factors when the effect of one factor on the response or output depends on the level of the other factor (Antony, 1999). For example, Figure 5 illustrates the interaction plot between billet ratio and length of sea draft.
Figure 5 shows that the effect of length of sea draft at various levels of billet ratio are almost the same. In other words, the lines are almost parallel to each other. This clearly indicates that there is no interaction between the billet ratio and length of sea draft. In order to discover which of the effects are statistically significant, it was decided to perform the analysis of variance (ANOVA) technique. ANOVA is a powerful technique which sub-divides the total variation into useful and meaningful components of variation. It also provides a relative significance of each main and/or interaction effect to the total variation. Table III provides the results of the ANOVA.
From F-tables, the tabled values of F-statistic are determined as follows:(see equation 1)Similarly,(see equation 2)
Table III shows that all the main effects are statistically significant and therefore will have significant impact on the gapping. Moreover, these three effects have accounted for more than 90 per cent of the total variation. Having identified the significant effects, it was quite important to predict the average gapping at the optimal condition of the system variables. It was therefore important to determine the optimal level of each factor prior to predicting the average gapping. Table IV illustrates the optimal settings of the system variables that will minimise gapping at sea.
The predicted average gapping at the optimal condition was estimated to be approximately seven, which corresponds to trial condition number 12 in the uncoded design matrix (see Appendix). This has shown an improvement of more than 40 per cent over the existing gapping at sea.
Conclusions
Process and system modelling are becoming an almost standard approach within many service organisations as a method of innovation. Risk analysis offers a method of both identifying opportunities for innovation and, more importantly, allowing organisations to evaluate and reduce their overall exposure to risk. This paper illustrates the identification of risk factors or variables involved in the Royal Navy's manpower planning system. The application of experimental design (ED) based on the Taguchi approach has provided a greater understanding of the system, especially in terms of reducing gapping at sea. As this is an initial investigation of the project, only three variables were considered for the simulation experiment. However, more variables will be incorporated into the simulation model as part of future research. It was found that the approach explained in the paper was very interesting and easily understood by all involved in the study, and thereby formed the basis of a new way of thinking about risk management issues at the Royal Navy's manpower planning system. See Table AI
References
1. Antony, J. (1998), "Some key things industrial engineers should know about experimental design", Logistics Information Management Journal, Vol. 11 No. 6, pp. 386-92.
2. Antony, J. (1999), "Improving the wire bonding process quality using statistically designed experiments", Microelectronics Journal, Vol. 30, pp. 161-8.
3. Antony, J. and Antony, F.J. (1998), "Teaching and educating Taguchi methods to engineers and managers", unpublished work.
4. Antony, J. and Kaye, M. (1988), "Key interactions", Journal of Manufacturing Engineering, Vol. 77, No.3, pp. 136-8.
5. Blosch, M. and Antony, J. (1998), "A process based approach to business risk analysis: the Royal Navy experience", Business and Process Management Journal, accepted for publication.
6. Getts, G.R. (1989), Portfolio Risk Management: A Computer Simulation for Stocks and Options, Addison Wesley, Reading, MA.
7. Kim, H. and Kim, Y. (1997), "Dynamic process modelling for BPR: a computerised simulation approach", Information and Management, Vol. 32, No.1, pp. 1-13.
8. Larkey, P. et al. (1997), "Skill in games", Management Science, Vol. 43 No.543 pp. 596-609.
9. Montgomery, D.C. (1991), Design and Analysis of Experiments, John Wiley and Sons, NY.
10. Neuse, D.M. (1998), "Why simulate", Capacity Management Review, Vol. 26 No.2, pp. 1-7.
11. Oakshott, L.A. (1997), Business Modelling and Simulation, Pitman Press, London, UK.
12. Raferty, J. (1994), Risk Analysis in Project Management, Chapman and Hall, London.
13. Recher (1983), Risk: A Philosophical Introduction to the Theory of Risk Evaluation and Management, University Press of America, NY.
14. Rowe, W.D. (1977), An Anatomy of Risk, John Wiley and Sons, NY
15. Saunders, D. (1995), The Simulation and Gaming Yearbook: Games for Simulation and Business, Kogan Page, London.
16. Taguchi, G. and Konishi, S. (1987), Orthogonal Arrays and Linear Graphs, ASI Press, Michigan.
17. Vose, D. (1997), Quantitative Risk Analysis: A Guide to Monte Carlo Simulation and Modelling, John Wiley Chichester.
Appendix
See Table AI.
Caption: Table I; List of variables and the interactions of interest; Table II; Main effects; Table III; ANOVA results; Table IV; Optimal settings of the variables; Table AI; Uncoded design matrix used for the simulation; Figure 1; Structure of the Naval Manning Agency; Figure 2; Typical naval manpower cycle; Figure 3; Requirement and strength; Figure 4; The ratings supply chain model; Figure 5; Interaction plot between billet ratio and length of sea draft; (see equation 1); (see equation 2)
Copyright MCB UP Limited (MCB) 1999