Analytics in Customer Life Cycle
In general, a typical Customer Life Cycle comprises Acquisition, Activation, Engagement (also includes Up-Sell, Next-Sell, Cross-Sell), Retention and Win Back. Nesting among all of these will be an on-going Customer Loyalty program running in tandem. However, various industry sectors have developed or modified its own Customer Life Cycle model, and that is acceptable given that each sector has its own unique business environment.
Data analytics can be used to drive key decision points in each of the Customer Life Cycle stages. For instance, using the Logistic Regression technique, marketers can utilise the Cross-Sell/Propensity, Look-A-Like/Propensity, Revenue and Attrition Scoring to glean deeper insights from a numbers perspective to craft their marketing strategy.
A Marketing Channel Optimisation Example:
To take this discussion further, let’s say we need to optimise our marketing spend for different segments of customers. One way to go about doing this is to allocate the most expensive method (eg personalised landing page) to the customer segment that has the highest propensity to accept the promotional offer. Do note that the data sets are based on historical customer data, as it the data model will compute the likelihood of success based on past experience of the same customer profile.
The thinking behind this is to invest the most into the groups of customers who are most likely to respond, thereby greatly increasing the cost-efficiency metric. For those segments deemed least likely to respond to that particular promotional offer, if budget allows, it might be worth a cheap shot (literally).
Also, when using Logistic Regression for such methods, we will also need to look at the acceptable Lift Value, which is a ratio of Band Rate over the Average Rate. The greater the Lift Value, the better it is. As a general rule of thumb, we should be looking at least 2-3 times in the value difference. Finally, the difference between the top bin value and the bottom bin value tells us how sharp the scoring system is – the greater the difference, the better the segmentation.
5xTwo 