An Overview of Segmentation:

Why You Should Consider It

And a Thumbnail of Its Dynamics

Prepared by:
Edward J. Hass, Ph.D.
Vice President – Advanced Research Methods
 The Motivation to Segment

The ultimate outcome of We all understand that consumers are not all alike. This provides a
segmentation is superior challenge for the development and marketing of profitable products and
satisfaction with what services. Not every offering will be right for every customer, nor will
you provide. every customer be equally responsive to your efforts to bring your
offering to their awareness. Success in these regards requires a
targeted approach to ensure bottom-line efficient use of product
development and promotional resources.

Segmentation is an informed means to organize customers into groups

that allow such targeting. The ultimate goal of segmentation for you is
the pragmatism of superior deployment and utilization of corporate
performance capabilities in meeting the needs and expectations of the
customer population. The ultimate outcome of segmentation for the
customer is superior satisfaction with what you provide.

Segmentation touches on matters that directly impact, and derive from,

your business strategy. On what type of customers should you focus
your efforts? How broad or narrow should your focus be? In which
competitive space is it most profitable for you to operate? Getting the
answers to these types of questions by means of a segmentation
analysis increases the chance for you to establish, maintain, and grow
a loyal customer base.

What Segmentation Does

Segmentation partitions a general population with the goal of tailoring

actions toward specific subgroups. It is a way of organizing customers
into groups with similar traits, performance characteristics or
expectations. These similarities can be obvious and simple at one
extreme or subtle and complex at the other. Information requirements
and the “organizing tools” used to create the segments vary greatly as
the nature and purpose of segmentation changes.

Consider a very simple segmentation exercise, one in which segments

were predetermined. Such an a priori scheme might begin with dividing
a customer base by zip code, with the eventual goal of regionally-based
promotions. Such data could then be crossed with sales data to see
which zip codes are heavier or lighter users of the product or service.
The output from that analysis would then be weighed against a
particular business strategy, such as working on maintaining loyalty
versus winning new customers.

While this approach may serve well as an exploratory attempt at

understanding how customers differ, it is a rather blunt instrument in
several regards. First, the partitioning takes place at a fairly gross
level, employing only the two variables of zip code and level of
product use. And then, it describes a situation without prescribing
the nature of an action that might be taken. Furthermore,
segmentation by its very nature assumes that all customers within a
segment are more alike than unalike – and that they are more
different from customers in other segments than similar to them.
Hence, using this very simple segmentation scheme, all customers
in a zip code would be considered alike. So any action taken
toward that zip code would be assumed to have the same response
from all customers in that zip code, clearly a very chancy
assumption.

Now consider a more complex example. Instead of starting with

completely predetermined segments, we actually search for a
segmenting rationale to infer to a population. And we do that with
variables that will inform us of customer motivations, trying to
understand the Why? behind purchase behavior. We might include
a bank of attitude statements surrounding purchase decisions in our
survey, or ask the importance of select product/service features,
lifestyle questions and demographic questions. With this we would
not only have the potential for a much more finely-tuned partitioning
of customers, but also a much more actionable partitioning.

An Example

We have a client who is planning to enter the Internet

telecommunications marketplace and needs to understand what
customers view as important from their phone service. Having
determined whether there are differences between customers in
that regard, and the nature of those differences, our client will be
able to devise a strategy to target customers most in line with their
currently planned service offerings. For future development, such a
segmentation can also point directions for new services.

We ask respondents to rate each of the following in importance

when considering the type of Internet service they would use:

 Conveniently available
 Inexpensive
 Is portable
 High speed band width
 Free connection
 Excellent security
 Reliable service
 Phone interface
 International access
In addition, we ask our sample to rate how well they perceive each
of several providers to perform in each of these areas.

Unless each of the 1000 respondents we survey gives identical

ratings, our data will contain variations that we can use to cluster or
group respondents together, and such clusters are the segments.
The mathematical algorithms we employ produce segments such
that:

a) variation within the segments is minimized

b) variation across the segments is maximized

This yields groupings of customers that are most similar to each

other if they are part of the same segment and most different from
each other if they are part of different segments. By inference,
then, actions taken toward customers in the same segment should
lead to similar responses, and actions taken toward customers in
different segments should lead to different responses.

Another way of saying this, in our telecommunications example, is

that the aspects of Internet telecommunications that are important
to any given customer in one segment will also be important to
other customers in that same segment. Furthermore, those aspects
that are important to that customer will be different from what is
important to a customer in a different segment. Here is what the
analysis in this example showed:
 Segmentation

Convenient AT&T SPRINT

Free Connections
Segment- Segment-
A Low Cost
B
(52%) (24%)
High Speed
Security

Portable choice
MCI
Reliable

VERIZON Segment-
Ego Segment- COMCAST
International
D C
PacTel
Access (10%) (14%)
Phone Interface

Our analysis shows four distinct segments. The majority of

customers
(Segment A, 52%) want mainly convenience from their service;
AT&T is perceived as convenient. Another quarter of customers
(Segment B, 24%) seek low cost and free connections, which
seems to be a Sprint characteristic. Two smaller segments seek,
respectively, a phone interface from their Internet
telecommunications (Segment C, 14%) or international access
(Segment D, 10%); Comcast and PacTel apparently meet the first
need, and Verizon the second.

In this case, there are a number of strategic options for the client to
consider. There seem to be open competitive spaces for customers
who might want portability, or reliability. Can such a need be
created, drawing customers from existing segments, or drawing
entirely new customers who might desire such qualities? None of
the current suppliers tested seem to satisfy jointly the need for a
phone interface with international access, nor to be both convenient
and low cost. Could our client make a claim to either combination,
or to develop a service that would make such joint claims? There
might be opportunities there.
How The Analysis Is Accomplished

Segmentations are generally performed via cluster analysis. There

are numerous specific techniques used for cluster analysis. Some
operate by treating all respondents initially as part of the same
cluster, then splitting the sample along “natural” fracture lines based
on survey responses, much like cutting a diamond (divisive
techniques). Others start by assuming all respondents belong in
different clusters, but then build larger and larger groupings based
on comparing survey responses (agglomerative techniques).

All methods share in common the computation of a measure of

similarity among respondents. Some techniques define similarity
on the basis of feature comparison, others employ the physical
metaphor of distance. In any case, the analysis takes the data
we’ve collected (behaviors, geography, attitudes, beliefs, benefits
sought, etc), scores respondents in a multivariate fashion, and
using a criterion we’ve selected, clusters respondents based on
whether they meet the similarity criterion level for inclusion with
other respondents.

There are many tools at the disposal of the segmentation analyst.

You may hear such diverse techniques as single linkage, complete
linkage, Ward’s method, polythetic, k-means interative partitioning,
CHAID, latent class cluster analysis, etc, etc. In fact the choice of
technique can determine the nature of the solution, so it is
legitimate, standard practice to use multiple methods and search for
convergence. Such convergence lends assurance that a valid
solution has been found.

But even with that, the utility of a clustering solution requires an

interpretation of substance within the context of “reality” based on
the applied discipline of the research. How does the
telecommunications industry work? Does this answer make sense
in the context of my knowledge of the pharmaceutical industry?
What are the constraints of financial services that impart credibility
to this result? The statistical solution must always be filtered
through intelligence about the industry environment in order for the
result to be accepted and acted upon.

Conclusion

Marketers accept the construct that what consumers seek out in

products and services is not uniform across all potential users – we
are no longer in the age of Henry Ford, who said, “The customer
can have any color car he wants, so long as it is black.”
Segmentation analysis shares this same construct – to conduct a
segmentation is to seek out differences in customers to capitalize
upon for ensuring maximum product uptake and brand loyalty. The
sophisticated multivariate techniques we employ for segmentation
are simply the tools to reduce the massive amount of data on
customer wants and needs to a manageable form so that we can
make educated targeting decisions. Hence, strategic customer
segmentation exists in service of the goal of striving to place the
right product or service with the right customer, to the benefit of all
involved.

Next Steps

Think about the business challenge confronting you. If the

questions you ask include such words as “targeting”, “customer
differences”, or “where to focus”, the chances are very strong that
segmentation analysis will be crucial for informing your decision
making. Discuss the problem with your research consultant in light
of the background you’ve gained from this white paper.