Welcome to my series of lifelong value forecasts of the customer's lifelong value. In the first part, we dealt with the estimate of *historical* CLV analysis and what you can already do with information that looks like that rear. Next , I presented a tone use case for CLV *prediction* , and went further than a typically limited example in another post about this topic. Now is the time for the practical part, including the data science team, the actual data and the customer.

Once again, there are so many moisture information that does not change to Odyssey and fit one blog post. So today I will focus on historical CLV modeling. I will cover foolish formal, cohort analysis and RFM approach, including the advantages and disadvantages discovered for each. I will do *the same* next time. And we will finish the entire series as a best example of data scientists on how to correct CLVs correctly.

Good sound? Then let's look at the historical CLV analysis and the advantages and “Gotchas” you need to know.

## Methods 1: A foolish simple formula

Perhaps the simplest formula is based on three elements. How many customers usually buy, how often they shop, and the period of loyalty:

For example, if the average customer consumes 25 euros per transaction, makes two transactions every month, and maintains loyalty for 24 months, CLV = € 1200.

We can make this a little more sophisticated in consideration of margins or interests. There are several ways to do this.

## Stupid Simple Simple official V1: Product Star **Margin**

Here, we calculate the average margin per product for all products of the inventory, and then multiply the simple formal results, such as the stupid form of this number, to create an average customer life *margin* .