Mining Your Own Business in Telecoms Using DB2 Intelligent Miner for Data

64 Mining Your Own Business in Telecoms Using DB2 Intelligent Miner for Data

However, there are some indicators to evaluate the quality of the segmentation

result. As a result of the segmentation, you can get the

confidence with which the

individual customers have been placed in the particular cluster and the

alternative cluster (second choice) the customer would be allocated.

You should check whether the confidence is low (close to 0.5). If this is the case,

additional iterations of the segmentation run may be needed. Examining the

variable that indicates the second choice of cluster will indicate which clusters

are relatively similar to the given cluster. If a cluster has low confidence and high

similarity with other clusters, then this also needs more interactions of the mining

run.

4.6.2 Interpreting the results from business perspective

The most significant part of whole mining process is to interpret the result in a

business sense and extract valuable information or knowledge for the people to

who it matters; in this case the marketing department.

Figure 4-2 shows one example of behavioral segmentation and you’ll get an idea

of how to interpret the result from a business perspective when you read through

this section.

The overall segmentation visualization initially describes how differently

customers behave based on their behavior. You can figure out:

򐂰 The customers who have high call usage and low service utilization

(Segment 8)

򐂰 The customers who have low call usage but high service utilization

(Segment 5)

򐂰 The customers who have high call usage at certain time frame

(Segment 4)

For example, when looking at segment 0 which is the third row from the top in

Figure 4-2 you find customers who

call a lot across all the time frame and

generate high revenue

. For the marketing perspective, this group of customers is

identified as a high revenue generating segment and may need special treatment

to keep them in the company.

From now on, we will look at the details for two interesting segments and explain

what the characteristics are of that segment, and what kind of business meaning

is implied in their characteristics.

Chapter 4. How to discover the characteristics of your customers 65

Night friends

A closer look at Segment 4 is shown in Figure 4-4. Here, 7.83% of the whole

population is in this segment. This segment has customers who are very young

students and call at night and use discount services.

Figure 4-4 Segment 4 - Night friends

Segment 4 has customers who have characteristics, such as:

򐂰 Mostly call at night (NIGHT_DUR)

򐂰 Not interested in most services (SVC_CALL,SVC_FREE) except friends and

family type of discount service (SVC_DISCOUNT)

66 Mining Your Own Business in Telecoms Using DB2 Intelligent Miner for Data

򐂰 Relatively many different telephone numbers used for outbound calls

(OUTSPHERE): many of them have 30-40 different numbers to call monthly

򐂰 More duration (minutes) in discount time frame (DISCOUNT_DUR)

򐂰 More minutes in inbound call (INBD_DUR)

򐂰 Teenage students (AGE, JOB)

These customers tend to have many different people to call, however their

outbound calls are not long compared to their inbound minutes. Even if a majority

of this customer group doesn’t subscribe to any services, their characteristics are

more likely to have discount type services. This is also indicated by their

spending more minutes in the discount time frame compared to a regular time

frame. After describing the characteristics inside the segment, we give the best

name to the segment, for example “night friends” for segment 4.

This segment doesn’t look like a profitable customer group in the first place.

However, they have potential to be more profitable customers if you encourage

them to use more outbound calls using the knowledge from this interpretation.

From the marketing perspective, based on this interpretation of segment 4, you

can design a special night discount price plan for the students to encourage them

to make outbound calls or develop “bring your friends to get a special offer” type

of campaign for this segment, for example.

Service-oriented

Let’s look at another interesting segment, illustrated in Figure 4-5. Here, 9.65% of

the whole population is in this segment. This segment has customers who are 40

years old and are interested in various services but their call usage is fairly low.