Identification of Key Drivers of Net Promoter Score Using a Statistical Classification Model - Efficient Decision Support Systems: Practice and Challenges from Current to Future - page 156

In the case of the constrained MLEs, the slopes of the last levels of the covariates are implied
zeros resulting from the full-rank dummy variable coding used when fitting the model.
Table 5 shows that incorporating the constraints do not lead to a substantial change in the
estimated slopes. In an indirect way, this provides a sanity check of the proposed model. We
will use the constrained estimates for the remainder of the case study.

3.4 Model utility
3.4.1 NPS for individual customers
The estimated coefficients of the model can be used to predict the distribution of Y for a
customer with a given set of covariates as follows. Suppose a customer is from the second
Country Code within the first Region Code, has the third Job Title and is associated with the
second Business Code. These demographic covariates are coded as follows:

RC=(1,0,0,0,0,0,0)
CC=(0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0)
JT=(0,0,1,0,0,0,0,0,0,0)
BC=(0,1)

Suppose further that the customer gives sub-element scores for q79, q82a, q82b, q82d and
q82f of 6, 6, 7, 5 and 6, respectively. These sub-element covariates are coded as follows:

q79=(0,0,0,0,0,1,0)
q82a=(0,0,0,0,0,1,0)
q82b=(0,0,0,0,0,0,1)
q82d=(0,0,0,0,1,0,0)
q82f=(0,0,0,0,0,1,0)

4

Combining these covariates into x  , using the estimates of given in Table 2,

and evaluating the equations in (2) gives the probability distribution for Y for this customer
profile as:

{ i   and 



1

1 2 3 4 5

 

Y ~

 

.00052 .00377 .10437 .43860 .45274

 

from which it follows that customers with this profile x  will have an (expected) NPS of
75.2%.

3.4.2 NPS for a population of customers
Consider now the sub-population of customers in the first Region Code. For this sub-
population, the relative frequencies of the three Country Codes are 81.18%, 14.12% and
4.7%, respectively. The relative frequencies of the ten job titles are 11.76%, 17.65%, 12.94%,
18.82%, 27.06%, 3.53%, 0%, 4.71%, 2.35% and 1.18%. The relative frequencies of the two
Business Codes are 82.35% and 17.65%, respectively. Thus, the demographic covariate
vectors for this sub-population are: