The Mining Data Mart - Data Mining Techniques in CRM: Inside Customer Segmentation - page 154

Table 4.14 The derived fields concerning channel utilization.

Customer_ ··· % ATM % BRANCH % INTERNET % PHONE Number of Flag of

ID transactions transactions transactions transactions different alternative

channels channels

1001 ··· 30 70 0 0 2 F

1002 ··· 0 21 62 17 3 T

(b) A numeric field measuring the number of different channels used by each
customer. This field is considered as a measure of the diversity of channel
utilization.

The additional information derived is presented in Table 4.14.
3. Merge the channel utilization information with the rest of the prepared
data to form the MCIF. The information for the channel utilization joins the
rest of the prepared data (customer demographics, product balances, type of
transactions made, etc.) to form the MCIF. The MCIF outlines the ''signature''
for each customer.

Derived Measures Used to Provide an ''Enriched'' Customer View

The MCIF should be designed to cover the key analytical and marketing require-
ments of the organization; therefore its contents will vary according to the specific
objectives set by the marketers. Apart from customer demographics, a MCIF
incorporates simple and composite derived measures which summarize customer
behavior. Typically, those derived measures include:

• Sums/averages: Behavioral fields are summed or averaged to summarize usage
patterns over a specific time period. For instance, a set of fields denoting the
monthly average number of transactions by transaction channel (ATM, Internet,
phone, etc.) or by transaction type (deposit, withdrawal, etc.) can reveal the
transactional profile of each customer. Sums and particularly (monthly) averages
take into account the whole time period examined and not just the most recent
past, ensuring the capture of stable, non-volatile, behavioral patterns.
• Ratios: Ratios (proportions) can be used to denote the relative prefer-
ence/importance of usage behaviors. As an example let us consider the relative
volume of transactions (percentage) per transaction channel. These percent-
ages reveal the channel preferences while also adjusting for the total volume
of transactions of each customer. Other examples of this type of measures
include the relative usage of each call service type (voice, SMS, MMS, etc.) in