RLC as a multi-attribute decision problem

In the social and behavioural science, many of the concepts are not directly measurable by a generally accepted measuring instrument, which are called

“latent variables” (Blunch, 2008). The concept of “regional logistics capability” is one of such non-measurable variables and therefore must be measured by so-called “manifest variables”. For example, the RLC concept has to be measured by five direct measurement (five dimensions in the section 2.3.8) and several sub-measurement (twenty four indicators in the section 2.3.8). This fits the characteristics of a multi-attribute decision analysis problem. This section illustrates how the MCDA tools could solve the proposed research questions with examples of previous studies.

Multi-criteria decision analysis (MCDA), sometimes called multi-criteria decision making (MCDM), is a discipline aiming at supporting decision makers faced with making numerous and sometimes conflicting evaluations. MCDA aims at highlighting these conflicts and deriving a way to come to a compromise in a transparent process (Triantaphyllou, 2000). Multi-attribute utility theory (MAUT) is a popular MCDA tool.

MAUT provides a comprehensive set of quantitative and qualitative approaches to evaluate alternatives for complex problems involving multiple objectives (Collins et al., 2006; Dyer et al., 1998).

MAUT first produces an “attributes by options” matrix for identifying single attributes and then evaluating alternatives on them (see Table 4-2).

Attributes A1 A2 A3 … An Total

Table 4-2. The MAUT attributes by options matrix.

Source: Roberts and Goodwin (2002) utility of the Oi, which is composed of two values: the scores of the options with respect to each attribute and the weights of the attributes. Generally, MAUT methods use a simple additive function to aggregate U(Oi) (Keeney and Raiffa, 1976):

Obviously, 0≤U(Oi)≤100.

MAUT models have been used in a variety of settings to solve real problems, from sitting an electricity generation facility (Keeney, 1980), choosing among vendors for the commercial generation of electricity by nuclear fusion (Dyer and

)

Winterfeldt, 1994). One of the primary tasks in the application of MAUT is to identify the overall best-in-class performer and justify a decision between alternatives (Collins et al, 2006). The MAUT approach enables the decision maker to incorporate preference and value trade-offs for each metric and measure the relative importance of each (Keeney and Raiffa, 1993).

Butler et al. (1997) illustrated the application of MAUT with a simple example site selection problem - coal power plant site selection. The selection is based on three notions: cost, environmental concerns and other technology specific features, which are captured by the measures of cost, air quality and site biology. They first establish the scaling constants (weight of each measure) and then derive the utility functions (component score of each measure) for each candidate site. Thereafter the best site choice is the one provides the highest value of a simple additive functions (See equation 4.1).

Dyer, et al. (1998) adopted the MAUT method to compare alternatives for the disposal of surplus weapons-grade plutonium. They evaluated 13 disposal alternatives on a hierarchy of objectives (Non-proliferation, Operational effectiveness and Environment, Safety and Health), sub-objectives and measures. The members of the Safeguards and Security time of the Department of Energy acted as advisers of the relative weights among different objectives and sub-objectives. And a team of experts from the Office of Fissile Materials Disposition of the Department of Energy assessed the single-attribute utility functions to be used for each measure to calculate the component scores.

Similarly, an additive multiattribute utility model can be used to aggregate the results to identify the best disposition location among the 13 alternatives.

Collins et al. (2006) used MAUT method to benchmark warehouse performances. They first identified most interested areas for warehouse operating such as the tolerance levels for inventory accuracy; how well errors are handled during operation; the highest accuracy rate could be expected, etc.

Then they select a set of proper performance metrics for the benchmarking study, including: Picking accuracy, Inventory accuracy, Storage time and Order cycle time. The next step is assigning utility values - they did so by working closely with experts associated with the study and assigning scores according to the warehouses‟ performance in each metrics above. Thus, each warehouse is given a component score for each metric. To bring these component score together to identify the best-in-class performer, Collins et al. (2006) also calculated the relative weights for the four metrics. They discussed with the sponsor organisation about the priorities of the current warehouse policies and determined initial relative weights should be assigned to the data, and then performed a sensitivity analysis to justify the relative weights. After the sensitivity analysis, the participant with the highest combined utility value is identified using the same additive function (see equation 4.1).

From the above examples, one can see that the MAUT methods provide a logical and tractable means to make trade-offs among conflicting objectives.

Although the measurement of a region‟s logistics capability is not a decision as such, it does need to consider and balance various factors that affect the logistics performance of a region (as discussed in section 2.3) and eventually presents a quantitative model that reflects the overall logistics capability of the

Due to its multi-attribute nature, regional logistics capability could be measured by the MAUT approach in this study, if we see the 11 regions in GB as

“alternative options” and different RLC indicators as “attributes”. In this way, it is possible to benchmark how each regions performs on each of the RLC indicator and more importantly, get an overall RLC measure for each region that aggregates all the RLC indicators.