When approaching a market segmentation project, the most important thing to understand is the purpose or objective of the segmentation. While objectives can differ markedly, they should all have a common theme: market segments should enable marketing strategy and tactics that are different for each segment. There is no reason to divide customers and prospects into separate groups if they are going to be treated the same. Whether the impact is on message content, targeting of direct marketing or mass advertising, new product development, type of customer service or other aspects of marketing and operations, the best segmentations will result in very different approaches for each segment.
Segmentation methodology should be designed to support the segmentation objective. This column will focus on one of the key distinctions in segmentation methodology: descriptive segmentation vs. predictive segmentation. It is important that this choice be made correctly based upon the segmentation objective. Lower level methodology decisions (e.g., the choice of descriptive segmentation methodology) can usually be left to the experts.
Descriptive segmentation methodologies are useful when a set of variables can be pre-identified based on their importance for making marketing decisions. Some examples will illustrate this point:
- A "behavioral" segmentation is based on variables that represent measurable behavior of individuals. For a retail store, this might be transaction information on frequency, quantity and type of products purchased. Customers who buy mainly products in one category would be likely targets for specialty catalogs or special offers for new products in that category.
- A "motivational" segmentation is based on variables that relate to why customers make purchases (usually collected using surveys on a subset of customers, when motivation cannot be inferred from behavior). Understanding the motivations behind purchases can be used to craft advertising messages and offers that appeal to these motivations.
Based upon the variables selected for the segmentation, descriptive segmentation methodologies divide the data into groups of individuals by putting "similar" individuals together.
It is the selection of the segmentation variables that primarily drives this process. The analytic tool will treat all variables equally, whether or not they are really important for making marketing decisions so choose these variables with care. As an example, suppose a variable for ethnicity is included among the segmentation variables. The resulting segments will likely have very different ethnic profiles. But remember which way the causality runs the analytic tool will separate ethnic groups because it assumes all the designated segmentation variables to be important. Don’t conclude that ethnicity is important because the analytic tool separated ethnic groups.
Predictive segmentation methodologies are useful for understanding what variables distinguish "best customers." The process is fairly simple:
- Predictive segmentation starts with the choice of one or several variables (dependent variables) that are key indicators of good customers. Clearly, these will usually be variables related to profitability quantity purchased, repurchase probability, etc.
- A predictive model is then built that identifies which predictive variables (such as past purchase patterns, life- stage variables, etc.) are drivers of the dependent variables.
- Customers are grouped into segments that have similar values of the predictive variables within each segment, but different values (and therefore different expected values of the dependent variables) across segments.
- "Tree models" such as CHAID and CART are well suited for this purpose because they simultaneously identify which variables are predictive of the dependent variable and group customers into segments based on the predictive variables.
This forms the basis for what I would call a "best customer analysis." Marketing resources should be focused on the most profitable customers. Predictive segmentation identifies not only which customers are most profitable, but what variables are associated with best customers. Therefore, it identifies not only whom to target but can provide valuable insights on how they should be targeted.
An online bookstore, for example, might learn that customers who purchase mainly within one category of literature (e.g., mysteries, romances, biographies) are more likely to purchase in the future. This might suggest that building special Web pages that provide information on specific categories might be well worth the investment.
It’s hard to imagine a business that would not benefit from a better understanding of their best customers. But this doesn’t mean that I would never recommend a descriptive segmentation. In fact, in many instances it may benefit to combine predictive with descriptive segmentations. But there is one obvious benefit to starting with a predictive segmentation: once the "best customer" segments have been identified, subsequent analysis can focus on these customers. For example, a survey based motivational segmentation could be accomplished more efficiently when targeted just at best customers.
If there is no objective more important than distinguishing and understanding your most profitable customers, consider starting with a predictive segmentation/best customer analysis.
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access