Once the golf shop chain defines its best customer, it needs to be able to estimate revenue potential by store, select optimal sites for new stores, tailor the merchandise assortment by store, and use direct marketing to reach households that are likely to become profitable customers within each store’s trade area.

In their never-ending task of defining and locating customers, retailers such as the ABC Golf Shop use GIS products for spatial analysis, queries, and mapping. If your business uses a GIS product as ABC Golf Shop does, then you are familiar with the need for importing variables from different databases to create thematic maps. Demographic variables constitute the most frequently imported databases, but there are hundreds of other specialty databases available today. Specialty databases cover such categories as consumer expenditure, crime statistics, business and employment, and many others. One of the more powerful specialty databases is called a segmentation system. A segmentation system is useful because it allows users to add another dimension by displaying lifestyle clusters as a theme on a map.

A lifestyle custer is a classification of a neighborhood that incorporates many different variables such as family status, income, consumer spending behaviors, media and advertising influences, and even leisure and recreational activities. These variables create a “portrait” of the households in that neighborhood.

You may be familiar with the various lifestyle marketing segmentation systems available on the market today. MicroVision®, PRIZM, ClusterPlus, ACORN, and Mosaic are brand names for the leading clustering models, otherwise known as geodemographic segmentation systems. But how are these segmentation systems created, what differentiates them, and more importantly, how can they be used in GIS applications?

Traditionally, marketing and advertising departments were the primary users of marketing segmentation techniques. Segmentation is most often used for direct marketing, database enhancement, and customer profiling. As more companies emphasize relationship or one-to-one marketing strategies, they are either building or ramping up their existing customer databases. This customer information has limited power when maintained solely within the database marketing department, but it can become a tremendous asset when effectively leveraged throughout a company.

“Segmentation has provided Macy*s West with a new dimension to consider when merchandising our stores,” said Chris Lestage, Marketing Manager for Macy*s West, “By comparing a store’s customer segments with the trade area’s segment composition, we are able to identify merchandise opportunities and customize each store’s merchandise assortment.”

So, exactly what is “geodemographic segmentation?” Based on the premise that “birds of a feather flock together,” geodemographic segmentation assumes that households within a “neighborhood” are fairly similar in their demographics, lifestyles, and purchasing behaviors.

How Segmentation Systems Are Built

Defining a “neighborhood” in geographic terms is the first step in building a segmentation system. All five segmentation systems mentioned above are built at one of the two most common “neighborhood” levels: a block group (about 350 households) or ZIP+4 (about 5-10 households). The first factor that differentiates the various systems is the size of their neighborhood definitions.

The second step is to obtain or create a large database of variables that describe all households in terms of their lifestyles, consumer behaviors, and demographics. The data sources for these variables are the building blocks of the model, and can vary widely in terms of data quality, integrity and robustness. The better the data, the stronger the model. This data must be collected at the lowest level (smallest geographic area) possible. If the data variables are collected at the block group level, the “neighborhood” will be defined as a block group. It is much more challenging to collect data at a ZIP+4 level, which usually requires access to proprietary, high-end data sources. However, the benefits of using the smaller neighborhood definition ZIP+4 level are recognized to be well worth the investment because the resulting segmentation system allows for targeting of much more finite areas with much greater accuracy. This capability is particularly important in dense urban areas with great diversity among neighborhoods.

ZIP+4 Neighborhood Diversity within a Block Group

Finally, the averages are calculated across all households in the “neighborhood,” and the data is modeled using a statistical technique called cluster analysis. This clustering model will group each neighborhood with other neighborhoods that are most similar to it. Typically, there are between 50 and 100 groups with every neighborhood in the United States “put into one of the buckets.” A strong clustering model will allow for a cluster of “anomalies,” those households that are difficult to classify because they are unlike any other neighborhood cluster. If “forced” into the closest cluster, anomalous data values would flatten out or weaken the cluster into which they are assigned. Therefore, it is important to select a segmentation system that provides a cluster for anomalies.

The company that creates each segmentation system will need to periodically update the clustering model using current year information in order to maintain its accuracy. In addition, every neighborhood in the United States will need to be reevaluated and reassigned, using the model, on an annual basis. Most neighborhoods will only change slightly from year-to-year; a 15% change rate is typical. Neighborhoods that change will typically move to a very similar segment, usually to one of the nearest clusters in either direction (i.e., a segment 14 might become a 13 or 15 after time), because neighborhoods evolve gradually. The end result of this modeling process is a geodemographic segmentation system.

How to Apply Segmentation

In order to apply this segmentation, a customer “profile” must be built. A profile is simply a frequency count of each geodemographic segment in your database compared with the incidence of those same segments in your base area. A base area is the larger group from which your customers were drawn, such as the geographic regions around your stores, your direct-mail list universe, or even the entire United States. For example, if 200 households on your customer file are assigned to segment 22, out of a total customer file of 2,000 records, then 10% of your customers are segment 22. If only 5% of the base area contains segment 22 households, then you are penetrating that segment at twice the expected rate (10% vs. 5%). In this case, segment 22 represents a strong segment. By looking at composition and penetration rates, your company can select the most promising target segments for your business development or expansion plans.

Sample Customer Profile

Showing Index for Purchasing ABC Golf Clubs by Cluster

So how can this lifestyle profile be visually displayed on a map? Once you know your expected penetration rates by segment, you can now look at the geographic composition of unknown market areas in terms of the quantity of households in each segment, and thereby estimate your success potential in that market area. Several equations may be used to project these national profiles down to a local level.

Number of Lap of Luxury (Cluster 2) Households in San Jose, CA by Block Group (Star indicates competitive golf shop locations)

A few examples of these equations are demand potential, market index score, and estimated potential customer households. Some software retrieval engines will even calculate these equations for you. For example, to estimate the potential number of customers you might gain by opening a new store at a proposed location, you can look at the number of households in each segment of that site’s trade area. For each segment, the number of households in the trade area is multiplied by the penetration rate for that segment from your overall customer profile. This is aggregated for all segments to estimate the total number of potential new customers who would shop at your new store.

MicroVision Demand Potential for Tampa, Florida
Using ABC Golf Shops MicroVision Profile
Block Groups in Tampa MSA

In GIS terms, this means calculating the estimated demand, households, or index for every sub-geographic area (such as block groups in a ZIP code or ZIP codes in a county). This becomes a variable that may be displayed as a theme on your map. Once the theme is displayed on your map, you will be able to visually notice “hot spots,” by their shade of color within a market, or identify potential new site locations. You can even use this technique at a macro level, such as the United States, to identify metropolitan areas for expansion or growth.

If, however, your company has yet to develop an internal customer database, you may choose to purchase a syndicated profile based on information collected from one of the many national survey companies. This is a good way to begin to use segmentation. If your company has a customer database, you can use syndicated profiles to benchmark your current customer profile against a national average. These syndicated profiles may be applied to a GIS application using the same equations described above.

Investing in segmentation can benefit your company by leveraging its customer database company-wide, from database marketing to spatial analysis and thematic mapping applications. Enhancing the customer database with cluster/segmentation codes is the first step. Your lifestyle profile can then be used to calculate estimated demand for your products and services at a micro-geographic level, adding a new dimension to your mapping capabilities.

Using these techniques, ABC Golf Shop is now able to quantify their success potential in new markets, estimate revenue potential of existing and proposed new stores, and more efficiently reach its best customers and new prospective customers who are most likely to respond, visit a store, and purchase golf products.