4.4.4. Customer value modeling
Based on the customer usage groupings found in the last
part, we applied RFM to evaluate customers’ potential
monetary contribution. The concept of purchasing behavior
from the original RFM is somewhat different from that of
network usage behavior, because the former represents
transaction of actual purchasing and payments, and the latter
may not be regarded as transaction of network service
until certain traffic threshold is reached. Thus, we refer to
the survey conducted by TWNIC (2004) to define this
threshold. The survey indicates that, on an average, the longest
continuous network usage interval in Taiwan is between
2 and 3 h. The company’s experienced ISP personnel suggest
that this network flow threshold be defined as 1% of stipulated
bandwidth over three continuous hours. As a result,
we modified the original definition of three factors R, F,
and M of the RFM model in the following to suit the
ISP environment; while the division of 5 intervals and
the weight assignments of RFM implementation remain
the same.
R: time of most recent traffic that lasts for 3 h and its
network flow exceeds the threshold.
F: usage counts over 7 weeks; with each usage lasting for
3 h and its network flow exceeding the threshold.
M: cost per network traffic unit, which is calculated as
network-monthly-rental/monthly-network-traffic.
For this study, our analysts had decided that R will not
be taken into consideration, because our purpose is to
study customers’ behaviors that should last over a longer
period and has little to do with the most recent usage.
With the definition of F and M, one can see that a
higher F means a higher demand on the network usage,
and a higher M means a higher usage cost per traffic unit.
Thus different combinations of F and M value would imply
different business intelligence. For example, customers with
high F and low M value tend to be frequent users with low value schemes, and those with low F and high M value are infrequent users with high value schemes. The worst case
is customers with low F and low M; it means they subscribed
to low value schemes and also rarely use network
facilities.
4.4.5. Facility utilization modeling
In developing the facility utilization model, we first
examine locations of districts and routers to define network
flow and utility rate for comparison. However, different
router or district may serve different number of customers;
hence we develop formulas (6) and (7) as the basis for
proper comparison. Formula (6) measures average network
flow per unit time of a router or a district, while (7)routers serve both Kaoshiung City, consisting of nine districts,
and Kaoshiung County, consisting of fifteen districts,
as shown in Table 6.
4.5. Analysis and evaluation
With the results of the modeling phase known, this
phase proceeds to analyze the imbedded business knowledge,
which could facilitate the development of relevant
service strategies. One can examine the patterns and brightness
of Fig. 4 to determine the VIP status of a customer.
The cross examination of usage behaviors of groups in
Table 4 could lead to the discovery of where the issues of
personalized services lie and how to approach these issues.
For example, the Overall Heavy Usage group is a group of
great immediate value to management. This group provides
management a much focused target with heavy usage
for nearly all time, and management could develop and
market high value-added products that fit their needs. Similarly,
for the three office hour groups with light, medium,
and heavy usage, they present management very focused
business targets with different applications needs. On the
other hand, the Overall Light Usage group happens to be
the largest grouping with 47.01% of users, most of them
are probably individual users, and it would certainly presents
the challenge to management to conduct further analysis,
so that some of them may be converted to more
regular users. For resource utilization, Fig. 5 displays the
utilizations of various routers and districts in colors, which
allows one to quickly realize that the router #20 has the
highest utilization, #12 comes in second, and is followed
by #13 & #18 and the rest. In terms of districts, Daliao
Township has the highest network usage in Kaohsiung
county and is followed by Fongshan city, Ciaotou,
Gangshan, Hunei, and Dashu townships. These high usage
districts present management targets for understanding their social-demographics factors that underlie the needs of
network services, and project the future needs of capital
investments.
The results of application of the modified RFM are giving
in Table 7. This table presents the cross analysis
between customers’ network usage patterns and customers’
potential contribution. The five intervals of F (frequency)
and M (monetary) are given in the second column; with
F5 being the frequency percentage of weight 5 (first
20%), M1 being the monetary percentage of weight 1 (last 20%). . ., etc. The R factor was not shown in the table,
because ISP specialists considered it irrelevant in this particular
case. The third column of the table presents suggested
marketing strategies for each group by marketing
personnel, who has examined both F and M values. For
each of the nine groups, their advices of strategy consist
of ‘‘Active marketing’’, ‘‘Customer care’’, or ‘‘Other’’.
‘‘Active marketing’’ means the development of value-added
products and aggressive promotions, ‘‘Customer care’’
means providing appropriate schemes to match customers’
needs, and ‘‘Other’’ literally indicates to management that
relevant strategies remain to be discovered. The percentage
attaching to each strategy represents the emphasis for that
particular group. Some examples of important findings are
presented here. The third group, the overall heavy usage
group, is characterized by 100% F5 and 100% M1, and it
signifies that customers in this group are heavy users;
however their monetary contributions are relatively low.
This is a big surprise for the management. Hence, 100%
‘‘Active marketing’’ is suggested to market to them higher
value services. On the contrary, customers in the overall
light usage group, Group 7, is characterized by values of less frequent categories F1, F2, F3, and high M categories
F5 and F4, which indicates that, although, they are good
monetary contributors, they are nonetheless very infrequent
users. For customers of this group, ‘‘active marketing’’
is not an issue, instead, a small portion of them
requires ‘‘customer care’’, and the remaining 90% is in
‘‘Other’’ category, which indicates that further investigation
is needed to find new strategy that can convert them
to frequent users. Another example that is different from
the two is the midnight medium usage group, Group 1,
which is characterized by high F categories 12.5% F5,
87.5% F4 and low M categories 50% M1, 25% M2 and
25% M3, thus, the suggestions are 75% active marketing
and 25% taking care of customers’ needs.