博客笔记八: [Zylotech] 用户流失模型以及召回的用户remarket

https://www.zylotech.com/blog/the-mechanics-of-predicting-customer-churn-part-1
https://www.zylotech.com/blog/the-mechanics-of-predicting-customer-churn-part-2
https://www.zylotech.com/blog/the-mechanics-of-predicting-customer-churn-part-3
https://www.zylotech.com/blog/got-a-reactivated-customer-dont-celebrate-just-yet

定义：
对于会员制，取消会员就ok了，定义简单。
对于非会员制，查看用户历史购买行为。customers’ transactional behaviors

首先知道他们为什么离开
past purchasing behavior to develop micro-segments, grouping reactivated customers by a best-fit explanation for why they left.

How much did they spend when they were customers? How did they reactivate—was it seasonal, a price discount, or some other factor? These and many others can be used to group customers back together.

Be sure not to ignore demographic information, the amount of time that has lapsed since they last bought from you, and any other information you might have on a group of customers.

不仅要有客户为什么离开的数据，最好也获取那些重新回来的客户的数据，综合来看知道客户为什么离开，也好对症下药。对于那些从来只买打折物品的客户，也不必发全价的促销给他浪费时间。