When a technology product is evaluated according to technologicalperformance, the hierarchical decision model is frequently used. The criteriaand sub criteria are decided by customers with regard to the utility that theproduct provides. At the lowest level a product model is composed of a seriesof technological features, which are listed, and then compared to each other,because these features embody the utility of each product model to customers. [13]

After the technological features are identified, a model fortechnology performance evaluation is set up, as shown in Figure 1. In thismodel, we used n technologicalfeatures.

Some features are numerical and the other features are descriptive.For the numeric features the utility might be a function of the numeric value,for example, the higher the maximum bandwidth, the better the performance - theutility of the feature increases as the maximum bandwidth rises. However, itdoesn’t increase in proportion to the maximum bandwidth value; it increases asa function of natural log function. So we can transform the natural log of themaximum bandwidth into the utility function of bandwidth.

For the descriptive features, such as “easeof use”, it is not as simple to get the utility directly from customers. It canonly be obtained by doing direct survey research. So, for this reason, thisstudy avoided using descriptive features to evaluate the models.

The data values should reflect the utility value since this projectused an output oriented DEA model.  Inthe case of a feature such as weight, where the utility increases with adecrease in value, the reciprocal can be used to transform the variable.  Lower values for weight are valuable forportable applications; while this feature is less important than space forbench models.  However, since space canbe positively correlated with weight (e.g. thinner units weigh less) then wecan use this transformation regardless of segment.

                                                                        

Figure 1. CommonModel for Technology Performance Evaluation

To calculate the efficiency of each model intechnological performance, the formulation to calculate the technologicalefficiency (TE) can be shown as Equation 1.

  

Where,

o-the sequencenumber of the Decision Making Units (DMUs) which is targeted to calculate itsefficiency;

-the technological performance efficiency of the targeted DMU;

i-the sequencenumber of output features;

j-the sequencenumber of the DMUs, j=1, 2, 3,…,J;

yij-the featurevalue of the ith output of the jth DMU.

Since the input is a constant 1, we choose to useVariable Return to Scale (VRS) and Output Orientation (OO). The DMUs with TEvalue equal to 1 are efficient in term of technological performance. The DMUswith TE value greater than 1 are inefficient in term of technologicalperformance. [4], [15]

To perform DEA calculation, the Benchmarking package in R is used. The newest version of R was downloaded from www.r-project.org.Manuals and other materials are alsoavailable on the same website. The “Benchmarking” package was theninstalled and the command >library(Benchmarking) is used to access the package.  Research data was then imported from a text filewith comma separated values (.csv file - using Microsoft Excel choose “Save as”and select csv from the pull-down).  Inthis case, the input x is a matrix with asingle column and a row for each DMU – the value for each DMU in this column isset to a constant of 1.  The output y is a matrix with the feature values with each rowcorresponding to a DMU and each column representing a technological feature.

The following is an example of the Rcode used.

>x<-constant1

>y<-6technology.features

>e<-dea(x,y,RTS="vrs",ORIENTATION="out")

>e