A simple thought about CLV

Clv,customer lifetime value, user lifecycle values. The user and our business, the process of association can be simply summed up, the introduction, consumption, loss. This whole process is called the user lifecycle.

Product Wang "lucky" to participate in the income category of the project, such as paid products, games, electricity dealers, may be more or less responsible for profit indicators. This time it will be like an operation to think: How to make more money (I am very straightforward oh)?

0, suppose I encountered such a problem: the product has been operating for some time, with a certain number of users, the next time, the project will be able to pay the future? How can you expand your future income? Should I do a promotion? Should we focus on introducing new users or maintaining old users?

1, with CLV analysis of income: Because if users are "dead", I have no income, so, income =CLV. CLV system has a number of variables, time, user status, user status of the corresponding value, the third can be disassembled into the average unit price, the probability of payment. Then, Clv= has acquired the user value +e (future user value)-C (operating and marketing costs), E refers to expectations, how to estimate it. (c not discussed in this article)

2, the Division of user status. The CRM case in university statistics always teaches us to cluster, actually it is a bit of use. For example, to do a game, a certain period of time units, users can be divided into: Do not give money to the cock silk, occasional money to the child paper, often give money to the rich second-generation. University statistics also teach us to use R (recency), F (frequency), M (monetary) Three typical indicators to cluster has reached the cock silk, child paper, rich second generation of grouping purposes. This time I can know that I do the user composition of the game, of course, a more simple than clustering method. Also such as a member service, can be divided into short-term paid users, medium-term paid users, long-term paid users, three of States.

3, after the completion of the user state, the key two estimated variables is the purchase probability of a user state and state transition matrix.

3.1. The purchase probability of a group is estimated by observing a reasonable sample.

3.2, the transfer matrix represents a group of users in the next time unit into a group of probability, the premise is a marketing strategy, national policy (such as not to do a net net action Ah!) ) and other external factors unchanged. A specific chestnut: what are the odds of not paying Dick Silk in the next month, or the chance to become a paid child, or to evolve into a high-handsome one? These are all estimated. Then, a user's state in the next time period is only related to the state of the month, and after estimating each possibility, a Markov transfer matrix is obtained.

4, finally, a simple estimate of a user group of M, can be estimated revenue expectations.

5. The ultimate goal is to maximize revenue expectations. Clv= has acquired user value +e (future user value)-C (operating and marketing costs). To achieve this goal, C is the means!

Changes in the marketing strategy will change the cost and make the revenue expectations change. When the operation purposely to try to calculate how the strategy can bring the most positive impact on key variables, which also has a lot of data analysis methods, and then want to write.