Predicting Customer Lifetime Value with “Buy ‘Til You Die” probabilistic models in Python

https://towardsdatascience.com/predicting-customer-lifetime-value-with-buy-til-you-die-probabilistic-models-in-python-f5cac78758d9

Predicting Customer Lifetime Value with “Buy ‘Til You Die” probabilistic models in Python

 

CLV

1. In non-contractual business settings, where customers can end their relationship with a retailer at any moment and without notice, this can be even trickier.

2. We can only rely on a customer’s past purchases and other less characterizing events (website visits, reviews, etc.).

 

BG/NBD Model

1. In particular, to predict future transactions the model treats the customer purchasing behaviour as a coin tossing game.

2. Each customer has 2 coins: a buy coin that controls the probability of a customer to purchase, and a die coin that controls the probability of a customer to quit and never purchase again.

3. Assumption 1: while active, the number of transactions made by a customer follows a Poisson Process with transaction rate λ (=expected number of transactions in a time interval).

4. Assumption 2: heterogeneity in transaction rates among customers follows a Gamma distribution.

5. Assumption 3: after any transaction, a customer becomes inactive with probability p.
Therefore the point at which the customer “drops out” is distributed across transactions according to a (shifted) Geometric distribution.

6. Assumption 4: heterogeneity in p follows a Beta distribution.

 

 

https://blog.naver.com/mykepzzang/220842759639

[확률과 통계] 40. 연속확률분포(4) - 감마 분포, Gamma Distribution