I hope you had the chance to read my companion post, Three Statisticians, a Rabbit and Medicare Marketing, but if not, I can wait 😉 Now let’s dive into the nuts and bolts of using data for Medicare marketing:
Three methodologies are crucial to data-driven Medicare marketing, but are often confused with each other. Here is a concise way to think of them:
1. Profiling leverages a wealth of data to make it easier to distinguish why members buy from you. Profiling is a very good baseline analytical project from which to gain member insights and develop marketing strategy by determining those characteristics most prevalent across your member base. Do your members own their own homes or rent? Have they lived in the area a long time or do they move frequently? Are they financially and/or technologically savvy? These examples are just a couple of potential data points to help you create a data-driven portrait of your members. Armed with this information, general marketing messaging along with product and service features can be made more relevant to your members and prospective members that you may want to target. Additionally,by profiling your members and prospects periodically, you have an effective way to identify shifts in your member base that could influence your marketing strategy. Lastly, profiling can provide the building blocks to open the doors to modeling and segmentation strategies that will allow your approach to lead to greater personalization.
2. Segmentation is a process for categorizing members and prospects into smaller groups, where those within the group exhibit the same set of unique attributes. Each “segment” is distinguishable by demographics, social economics, financial behavior, attitudes, etc. By tailoring the offering (creative/communication, product, channel, price) to different segments, you are able to better identify and meet the needs of more members. For example, it would be more effective to market products differently to the five segments below:
While segmentation is frequently used in developing products, features, and service options to appeal to each segment. In marketing, it is also powerfully used to vary messaging and creative treatment to improve open and response rates vs. one-to-many marketing techniques. And it is a great step to move you closer to a “one-to-one” marketing messaging strategy.
3. Modeling is statistical mathematics employed through developing a formula to predict the likelihood that someone will respond to a certain target behavior – such as who is most likely to respond to a Medicare Advantage direct mail campaign during AEP.
Modeling determines the difference between those who respond to your AEP direct mail campaign in the past vs. those who didn’t respond. This is accomplished by inputting hundreds of data elements and then comparing and contrasting them to gradually reduce the number of variables and find the ones that are the most predictive between the two groups. If profiling and segmentation are thought of as two-dimensional, then modeling is three-dimensional. Modeling reduces variables based on the interactions between those variables and the target audience, and is predicated on how the variables interact WITH EACH OTHER.
Modeling is like baking a cake. All of the variables work together – very similar to your cake recipe. All of the data variables are mixed in the bowl. Think of the whisk you use to stir the batter as the statistical tools like regression, neural networks, etc. In contrast to customer profiling or segmentation, modeling does not consider any variable or ingredient by itself, so no “selects” are performed. Each ingredient does its work as they interact with other ingredients. The recipe works beautifully in unison, and one single ingredient may make a small but necessary contribution to the success of the model as a whole. A result that is much greater than the sum of its parts. Have you ever made a cake with baking soda instead of flour or baking powder? It’s disgusting by the way.
The final output of the model is a simple algebraic equation with each final variable having a specific weight so that when the entire target audience is scored, each prospect comes back with a score (or weight) of 1-100 (often bunched in deciles or 10 equal groups). Modeling provides a score for each prospect respective of the degree to which those non-responders ‘look like’ the actual responders. You find the best prospects, so that your budget isn’t spent on those not likely to respond.
So what’s the best – modeling, profiling or segmentation?
Read part 3 of this series tomorrow.