Professional Documents
Culture Documents
A Study of The MRP Implementation Process
A Study of The MRP Implementation Process
A Study of The MRP Implementation Process
TABLE 2
MAJOR REASON FOR MRP IMPLEMENTATION AND MAJOR PROBLEM ENCOUNTERED
100%
“computerization.” The company size factors was decided to hold back one-fourth of the data
was heavily weighted by facility sales, total com- from the analysis. A stepwise discriminate anal-
pany sales, total number of employees and num- ysis was performed on the remaining three-fourths
ber of employees in P&C. The computerization of the data to discover the most promising vari-
factor was most heavily weighted by the extent to ables from the developed framework. The analysis
which various modules of the MRP system were was performed on 116 complete cases, 52 suc-
computerized and the integration of the separate cessful implementors and 64 unsuccessful imple-
modules. These two factor scores were included mentors.
in the discriminant analysis. Prior condition of Table 7A shows the results of the final discrim-
company refers to a question as to what type of inant analysis. Of the 116 cases used to fit the
system was used previous to the MRP system. model, 90 percent of the successful companies and
The other variables in this classification have been 86 percent of the unsuccessful implementors were
discussed previously or are self-explanatory. correctly classified. Overall, classification was 88
System features of the particular MRP system percent correct.
have been classified as independent variables. The jackknifed classification is also shown in
Although decisions must be made about selection Table 7B. In a jackknifed classification, each case
of features, this selection is viewed here as sepa- is classified as either successful or unsuccessful
rate from the actual implementation process. using a discriminate function computed from all
The intervening variables are those factors other cases. The jackknifed classification both
which relate directly to the process of implemen- tests the predictive power of the model and gives
tation. In addition to the problems encountered a more robust measure than the simple classifi-
(discussed in previous sections) length of time cation [5, 91. As can be seen, the jackknifed clas-
since the implementation of MRP and information sification is close to the simple classification, with
about the process used are considered as poten- 85 percent of all cases correctly classified. The
tially important. discriminant analysis, therefore, is very reliable
The first step in the actual analysis was the in identifying successful from unsuccessful imple-
construction of a hold-out sample. As in all anal- mentors.
ysis, there is the possibility of “over-fitting” the The hold-out sample contained 44 complete
data: the analysis may describe the data being cases: 21 successful and 23 unsuccessful compa-
analyzed, but be inappropriate for other data. To nies. To further test the predictive power of the
test the predictive power of the fitted model, it model, the discriminant function was used to clas-