Segmentation Process and Strategy

1
 Contents

Segmentation Process…3
Dart’s “Custom Segmentation” Approach ….4
Applications for Segmentation…5
Techniques & Data Used …6
Overview of the Process with Timeline …8
Keeping Segmentation Relevant…10
Further Analysis …11

Segmentation Example…12
Test Case ….13

2
 Segmentation Process
Dart’s “Custom Segmentation” Approach
Applications for Segmentation
Techniques & Data Used
Overview of the Process with Timeline
Keeping Segmentation Relevant
Further Analysis

3
 Dart’s “Custom Segmentation”
Approach

Dart builds sophisticated “custom” segmentation models.

Purpose:
To achieve highly differentiated customer segments that make marketing more efficient and
effective.

Method:
Experienced modelers use a combination of science and intuition to create a custom segmentation
scheme. A good solution requires that the segments be distinct, predictive of behavior,
implementable, and reflective of the business needs for which they were created.

We also perform data quality checks and report any problems or questions before we arrive at a
final solution.

Results:
An elegant cluster solution that is practical, makes sense and can be implemented.

4
 Applications for Segmentation

There are many uses for segmentation. These are some examples.
Purposes:
Needs Based Segmentation - Auto makers for example design vehicles to match the needs of
buyers, ranging from economy cars to luxury cars and minivans to pickup trucks.

Product Segmentation – Manufacturers diversify products within each needs base to appeal to
buyers with different tastes and wealth.

Customer Segmentation – Customers are segmented based on their needs and product
preferences. Segments grow or shrink over time as products improve, become obsolete or tastes
change.

Niche Segmentation – Niche segments are characterized by strength in one needs base and product
within it. Restaurants are good examples, ranging from delis to Chinese food with décor appealing
to McDonald’s patrons to those preferring a three star experience.

Global Segmentation - Insurance firms and medical and legal practices also use product
segmentation, and sometimes attempt to cover all the product space.

In-store Display Segmentation – Drug stores, grocery stores, book stores, and other retail outlets
use segmentation in order to keep like products close to each other within the store, making
shopping convenient and cross selling more profitable.

5
 Techniques

Techniques to Developing Clusters:

Statistical clustering techniques include neural networks, discriminant analysis, factor analysis,
hierarchical clustering, and perhaps most commonly, "nearest neighbor" or "k means" algorithms.
All of these approaches determine what variables are similar and dissimilar in statistical terms,
forming segments.
The analyst picks the number of clusters through an iterative process, looking for uniqueness
between the segments and a number of segments that are practical and manageable from a
marketing perspective.

Data Definition:
How variables are defined makes a substantial difference in the outcome. Age, for example, can be
characterized as a set of age-ranges or as a continuous variable. These characterizations lead to
different segmentation solutions. So, selection of the best way to characterize the variables used
for segmentation involves considerable judgment, from both a statistical and a business
perspective.

6
 Data Used

Within practical limits, the more data the better, in the initial stages. The data relevant
to the segmentation scheme is revealed through the statistical process. But, the solution
must make sense and the variables used must make a contribution.

Customer Data:
Transaction Details – Frequency, amount and timing of purchases, items bought, prices paid, use of
cash or credit, and use of coupons.
Acquisitions Details – Marketing channel, promotion type, and address/city.

Appended Database Data:

Life Style – Profession/occupation, vehicle ownership, Internet use, travel, pets, and hobbies.
Financial – Investments, credit card usage and type, living expenses, and credit worthiness.
Demographic – Age, income, education, gender, marital status, and number of kids.
Geographic – Own/rent, urban/rural, size of city, region, and size of dwelling.

Market Research Data:

Behavioral – Purchase patterns, why they bought, what they use the product for, responsiveness to
different marketing channels.
Attitudinal – Product preferences, willingness to try other brands, price sensitivity, shop for
convenience, opinion of the company and the competition.

7
 Overview of the
Segmentation Process
Project Timeline: 15 days to several months
depending on the size of the project
Data Prep/Hygiene:
Data is read into an analytic file. Data records and
variable values are examined for accuracy. Records Days to
2 3 2 2 4 2
with duplicate match code ids are compared prior to Complete
de-duping those records. Variable values are examined Task
to make sure they are within acceptable ranges. 0 3 6 9 12 15

Initial Exploratory Analysis: Data Prep/Hygiene: 2+ Days

The heart of the work - Data description and looking Initial Exploratory Analysis: 3+ Days
for explanatory patterns in the data, which lead to a Segmentation Analysis: 2+ Days
picture of your business, customers, products, Implementation: 2+ Days
environment, and financials. Descriptive Profiles: 4+ Days
Financial Analysis: 2+ Days
Segmentation Analysis:
Selection of the clustering technique and the variables that will be used.
Implementation:
First, the sample file is scored with the segmentation scheme. Then, all other records that contain the
data used to make the segments are scored. The remaining records that do not contain the necessary
data (such as those not included in a survey that was used) must be assigned to the segments using
other means. There are several methods to accomplish this, including regression and neural networks.

8
 Completing the Segmentation Process

While the segments have been defined by this stage, a face still needs to be
put on them for them to make sense.

Name Assignments:
Typically, descriptive names are given to segments, instead of referring to them as Segments A,
B, and C. These names generally reflect the key components that describe them.

Descriptive Profiles:
Profiles describe the attributes of each segment. For example, Customers in “Segment A” are
36% more likely to buy frequently than customers in “Segment B.”

Some variables not used in the clustering process are retained for describing the segments. For
example, while segments may be based primarily on their behavioral characteristics, it is still
worthwhile to note their demographics.

Financial Analysis:
Determine the expected financial performance of each segment. Response indexes and residual
income from likelihood of repeat business is often part of the analysis.

9
 Keeping Segmentation Relevant

It's important to monitor the performance of a segmentation scheme over

time and recalibrate as necessary.

Shifts in Market Conditions:

Work with client to track performance measures for each segment. A monthly performance
scorecard is a good mechanism for tracking changes in performance and the company’s position in
the market place.

Fixed Intervals:
A simple alternative to this tracking process is to recalibrate the segmentation scheme at fixed
intervals, such as once a year.

10
 Further Analysis

Get the most out of your segmentation strategy.

Optimize Profitability through Financial Modeling:
Expand the initial financial analysis into an interactive model. This allows “what-if” scenario testing
to maximize the segmentation mix, marketing mix, mail strategy and product pricing.

Increase Prices without Losing Sales:

Scientific price/incentive test to quantify the price elasticity of demand. This analysis drives the
price component of the financial model.

Improving Segmentation through Appended Database Data:

Database enhancement research with cost/benefit analysis reveals which additional data provides
the most predictive power for the investment.

Using Market Research in Combination with Segmentation:

1) Validate segments in the real world,
2) Collect data to fine tune the segments,
3) Better understand purchase motivation, behavior, and desirable product attributes, leading to
more effective offers, and
4) Better target creative, resulting in better response to solicitations.

11
 Segmentation Example
Test Case

12
 Descriptive Profiles
Automotive Credit Card Segmentation
The chart to the right shows the distribution of automotive
credit card accounts by segment. “Low Spenders”, “Game
Players” and “Credit Needy” were the biggest segments.
Low
Spenders Game Players
The charts below describe the “Game Players” segment: 19% 16%
Credit Needy
Segment Highlights: Credit 14%
Challenged Loyal Credit
- They are high spenders, accumulating as much rebate as 11% Worthy
possible through the program. Mixed Bag 7%
13% Used
- They have the highest likelihood to redeem their points Aspirational
Conquest
- They are more likely to own a new car made by the mfg Credit 13%
7%
sponsoring the program
- They have normal age and income distributions Income Distribution

16.3 17.1
Age Distribution 18.0
16.0 14.4
13.9 14.0
14.0 12.2
30.0 12.1
Vehicle Ownership by MFG 12.0
23.2 25.0 10.0
25.0 8.0
European
19.6 6.0
Imports 20.0
4% 4.0
15.2 2.0
15.0 13.7
Asian 0.0
Imports
10.0

+
39

49

79
19

9
18% GM

$2

$5

0
-$

-$
-$
-$

00
3.3

0-

0-
49% 5.0

0,
0

0
$0

$2

$5
$3

$4

$6

$8
Other
Domestic 0.0
29% <25 25-34 35-44 45-54 55-64 65+

13
 Financial Analysis

Description: This example is based on a credit card with an automotive rewards program, where people
accumulate a percentage of their purchases towards a new automobile. Revenue is based on credit
income and profits from auto sales. Expenses come from redemptions and marketing/operating costs.

Key Findings: Game Players were very costly to the program. Credit Challenged were expensive due to
bad debt. Low Spenders were profitable as were Conquest Credit (due to high incremental sales rate).
Game Credit Loyal Credit Used Conquest Mixed Credit Low
Players Needy Worthy Aspirational Credit Bag Challenged Spenders Total
Cardholder Accounts
Number of Average Accounts 551,500 494,500 260,500 443,500 233,000 437,500 375,000 704,500 3,500,000
Average Active Accounts 496,350 296,700 197,980 252,795 137,470 266,875 176,250 274,755 2,100,000
% Active 90% 60% 76% 57% 59% 61% 47% 39% 60.0%

Financial Summary ($1000's)

Interest Income 106,361 63,579 42,424 54,170 29,458 57,188 37,768 59,053 450,000
Interchange Income 35,454 21,193 14,141 18,057 9,819 19,063 12,589 19,684 150,000
Other Income 23,636 14,129 9,428 12,038 6,546 12,708 8,393 13,123 100,000
Total Credit Revenue 165,450 98,900 65,993 84,265 45,823 88,958 58,750 91,860 700,000
Automotive Profit 39,708 16,107 3,960 14,445 15,318 15,000 8,057 7,405 120,000
Total Revenue 205,158 115,007 69,953 98,710 61,141 104,208 66,807 99,265 820,000
Redemption Expense 132,951 35,321 37,815 27,160 15,825 32,300 12,875 14,081 308,329
Other Expenses 106,361 63,579 30,500 54,170 14,729 57,188 96,308 27,166 450,000
Total Expenses 239,312 98,900 68,315 81,331 30,553 89,488 109,183 41,247 758,329
Gross Cash Flow (34,154) 16,107 1,638 17,380 30,588 14,720 (42,376) 58,018 61,671

Program Redemption Summary

Avg $ per Redemption 2,813 1,250 2,006 1,128 1,209 1,292 767 528 1,250
Redemptions 47,271 28,257 18,855 24,076 13,092 25,000 16,786 26,662 200,000
Incremental Sales % 28% 19% 7% 20% 39% 20% 16% 9% 20.0%
% Accounts Redeeming 8.6% 5.7% 7.2% 5.4% 5.6% 5.7% 4.5% 3.8% 5.7%

14
 Contact Info

Craig Tomarkin
DART Marketing, LLC
2333 Congress St.
Fairfield, CT 06824
[email protected]
203-259-0676
Fax 419-858-8545

15