Regression models (linear, non-linear, single variable or multiple variables) with good R square can help to predict the period two (t+1 period) outcomes. And we can also leverage this outcome to get optimal strategy, such as pricing strategy, promotion solutions and so on.
When the model prototype is ready, the next step is variable selection (pick up the right data dimension). One of the most popular selection is RFM. aka, Recency, Frequency, Monetary Value.
Recency: the last time the customer made a purchase with me, or visited my website, or made a sales call. Or any activity that indicate that this customer might become a valuable client.
Frequency: How many purchase that a customer made over a certain period.
Monetary Value: what is the overall or average value of these economically beneficial activities ?
Importance: R>F>M
(Above picture comes from https://www.zhihu.com/question/49439948/answer/254004098
However, when it comes to customer value, we are not satisfied with the short-outcome. We would like to know how many periods that customer can survive, and where is the customer centricity (who is the right customer?). We need to have visibility in the long run for customer life time value.
We can use Probability Model /BTYD model to predict the outcome:
Assumption: 1) Each customer will act randomly (yes,or no for each decision), however, there is a certain propensity for he/she. 2) The precondition for customer's decision is that the customer is still alive to the product.
Methodology: To Be Continued..
