Customer Propensity Modeling

Propensity is the natural tendency to behave in a particular way. Using that definition it should be easy to see customer propensity modeling is the use of predictive models to try and predict customer behavior. Understanding these behaviors can allow your business to positively impact your relationship with your customers. 

With a general understanding of what propensity modeling is, the next step is to define some behaviors that can be modeled. Below are some common examples of types of propensity models used in marketing analytics.

LTV – LTV propensity models look to predict the value of a customer. This can be done at any time in the customer lifecycle. A customer’s LTV can be a useful tool in segmenting your customers. More effort then can be used in starting a relationship with more valuable customers.

Likelihood to Buy – This propensity model tries to predict when a customer is likely to buy. This model may look at data points like website visits, marketing touchpoints, form fills and more to identify people that are in the buying mood. Knowing a customer is in a buying mood can allow you to communicate in relevant ways to close the deal.

Churn – A customer who has churned is one that no longer has a relationship with your company. Knowing when a customer may churn can allow you to create retention programs to try and save the relationship.

Response/Engagement – Another useful propensity model is response modeling. This model predicts the likelihood of a person to respond to a marketing piece such as a mailer, email, or outbound call. This type of model is another effective way to focus limited resources on the people most likely to create a result.

The propensity models listed above are only a few examples of customer behaviors that can be modeled. Depending on the availability of data the possibilities of modeling customer behaviors can be large. Creating the model is only one step in an effective strategy. The next step is to implement the model and a program to provide action. The next section will describe one way to use propensity modeling to increase revenue for your data driven organization.

One of my favorite ways to use propensity modeling is to create a customer quadrant segmentation structure by using LTV and Churn propensity models together. The idea here is to create a quadrant to segment customers into four categories which are HIgh Value Intervention, Low Work High Value, Low Value and High Work Customers. A picture representing the quadrant can be found below.