The Art and Science of Static and Dynamic Cluster Analysis
Today, strategies of marketing professionals in the performing arts are as much about data management and budgets as they are about marketing plans themselves. Trying to invest scarce resources in acquiring and retaining customers who are most likely to buy, for example, requires in depth understanding and exploitation of an organization’s customer database.
Fortunately, a broad range of techniques—and the experts who know how to help companies use them to best advantage—are available to assist in this effort.
Among the several types of data-driven analytical tools is a descriptive modeling technique known as “cluster analysis” (with sub-groups called static cluster analysis and dynamic cluster analysis) used to determine and qualify attributes of distinct groups (segments) in a customer and prospect database. Once identified, these attributes and groupings allow organizations to create targeted offers based on the needs and characteristics of the customer segments and improve the effectiveness and efficiency of their marketing programs.
Cluster Analysis is a sophisticated tool that enables an organization to
- Move from broad to highly targeted advertising
- Tailor product offers to unique customer/prospect segments
- Determine the most efficient and effective media and distribution channels to reach a particular segment
- Design the most appropriate “look and feel” in the creative component of a campaign
How are Static and Dynamic Cluster Analyses different?
While both static and dynamic clustering tools can be used to determine how customers are alike and how they are different, there are some important differences between these two techniques.
Static Cluster Analysis - Static clustering is a “one-size-fits-all” model. Customers and prospects are assigned to pre-defined neighborhood clusters, based on the geography that “fits” them best—even if the fit is not ideal. For example, the Prizm™ cluster methodology defines 62 static clusters, each with a roughly descriptive name like “Upper Crust,” “Old Comrades,” “College Student,” or “Hard Scrabble.” This methodology then assigns everyone living in a zip+4 neighborhood to one (and only one) of these clusters.
While some neighborhoods are full of people who are very much alike, many other neighborhoods are not. Residents of a unique zip+4 area may be quite different from each other in terms of attitudes, interests, and purchasing preferences. Static cluster descriptions tend to gloss over important characteristics and differences among the people assigned to the cluster.
Dynamic Cluster Analysis - Dynamic clustering is a “custom” model. Patron and prospect segments are not defined in advance and customers are not assigned to a pre-defined segment. Instead, the unique demographic and psychographic profile of each customer in a database or list is used to define the segments.
As a result of this specificity, dynamic clustering is the preferred method to determine:
- The demographic diversity of the organization’s current buyer and prospect groups
- The socio-economic profile of each buyer or prospect group at the household level
- The hobbies, activities, and interests that are most appealing to each group
- The family and life-cycle stage of each group
- The best ways to describe each buyer or prospect group
How is Dynamic Clustering Analysis performed?
Names and addresses of customers are used to append information about each customer, adding demographic data such as age, income, marital status, etc. In some cases, psychographic data such as hobbies, lifestyle, and preferences for products, services, and brands are added to develop a complete portrait of the customer or prospect.
Data added for dynamic clustering are appended at the household level; therefore, inferences about age, income, and even hobbies or lifestyles are likely to be far more accurate than using neighborhood averages. As a result, young families living near older neighbors, or singles living among married couples would likely belong to different clusters.
An advanced statistical technique called K-Means Cluster Analysis is used to define which factors best describe each group of customers. It is often the case that customer differences result in only five or six segments, but these segments are usually quite distinct and different from one group to another.
Conclusion
Both static and dynamic cluster analyses have a place in helping to design and implement an effective marketing plan.* Dynamic clustering is generally superior to static cluster analysis for assigning the most accurate cluster attributes to unique customers. But as is the case with all database analyses, cluster analysis is both a science and an art. Experienced database and modeling experts are the best resource for conducting such analyses and translating them into actionable marketing strategies.
To learn more about cluster analysis and how your organization can benefit from it, please contact Will Lester, Vice President, Target Resource Group at 719.314.5835 or email wlester@trgarts.com.
*Note: clustering methodology is usually less effective than “predictive modeling” for selecting lists or contact/no-contact targeting. Predictive modeling, as the name indicates, is designed to predict who is most likely to buy or respond, while profiling and cluster analysis simply describe different groups of customers. This is an important difference. (For details on different methods of data-driven marketing analyses, please see TRG’s White Paper: “Data-Driven Marketing Analysis”)
