The primary objectives of segmentation modeling would be to answer the following:
1. Do distinct and natural segments of customers exist?
2. How significant are the sizes of these segments?
3. What are the similarities and dissimilarities that differentiate between these segments?
4. Are the factors of differentiation actionable?
5. Can differential product solutions, offers, and communications be leveraged in targeting these segment clients?

Secondary objectives may be to answer the following:
6. What is the average current value of these customer segments? Which is the highest value segment?
7. What is the potential value of further and future investments in these segments?
8. With which product solutions do these opportunities lie?
9. Where should marketing resources be focused to optimize the relationships from each segment and maximize purchasing volumes (i.e., targeting strategies and tactics)?

Segmentation Approaches

There are two general approaches in segmenting a customer base. The first approach can be referred to as an a priori method, where the number of segments is determined by the business users in advance and individual members are simply placed into one of the segments through realization of some criteria.

The second approach can be referred to as post hoc, where data analysis determines the number and type of segments that naturally occur within the customer base.

Typically, there are several techniques for the same data modeling project. For segmentation, the two primary methods are hierarchical and partitioning (i.e., k-means), the latter of which was used to develop Vancity’s behaviour segments. For the purposes of this investor segmentation project, however, it is recommended that a third technique, called two-step clustering, be utilized.

The two-step clustering process is a hybrid of hierarchical and partitioning methods and offers several clear advantages over traditional clustering procedures:
 Handles mixed types of attributes (continuous and categorical variables).
 Handles outliers, which may particularly be a concern when dealing with investment and banking behaviours and portfolios.
 Provides insights as to the optimal number of clusters.
 Provides statistics on the influence of independent variables in defining a cluster group.
 

Segment Validation

The success of Phase 1 of the project will be judged on the following outcomes:
 The “goodness” of the segmentation model (i.e., maximum homogeneity within clusters; maximum heterogeneity between clusters).
 The usability of the segmentation model (i.e., are the segments small enough to gain usable information, but large enough to be actionable for marketing initiatives).
 The timeliness of the deliverables.
 The successful implementation of the segmentation model to support marketing decisions and campaigns going forward.
 Overall self-reported client satisfaction ratings with respect to the project process and outcomes.

The resultant segmentation model will be reviewed to ensure the proper attributes were used and are available for future analysis and model refresh activities. While there are no formal procedures for evaluating segmentation models, technically, the following issues need to be addressed:
1. Do distinct segments really exist? If only a few variables show meaningful differences between individuals, it is possible that no really distinct segments exist in the investment market.
2. Is the number of clusters retained conducive to meeting the managerial objectives set forth?
3. Do the means of the descriptive variables in each cluster make intuitive sense (face validity)?
4. Are the segment names intuitively appealing?
5. Are clusters sizes for all segments substantial enough to justify separate targeted marketing and communications efforts?