Welcome back to my series on predicting the lifetime value of customers. It's called "Everything the other tutorials left out." In Part 1, I covered what can already be done with historical CLV analysis, often estimated stages, and information that looks like it's from the rear. Next, I presented use cases for CLV prediction and went further than the typical limited examples on this topic in other posts. Now it's time for the practical part, including everything I've learned while working with the data science team and real data and customers.

Once again, without turning it into an odyssey, there's too much information for one blog post. So today, we'll focus on historical CLV modeling. I'll cover simple formulas, cohort analysis, and the RFM approach, including the pros and cons I found for each. Next, I'll do the same. And I'll finish the entire series with best practices learned by data scientists on how to properly perform CLV.

Sounds good? Then let's take a look at the historical CLV analysis methods and the advantages and "Gotchas" you need to know.

Method 1: Simple Foolish Formula

Perhaps the simplest formula is based on three elements: how much the customer typically spends, how often they shop, and how long they maintain loyalty:

For example, if the average customer spends €25 per transaction, shops twice a month, and maintains loyalty for 24 months, then CLV = €1200.

We can make this a little more sophisticated by considering margin or profit. There are a few ways to do this.

Foolish Simple Formula V1: Margin per Product

Here, we calculate the average margin per product for all products in inventory, then multiply this number by the simple formula result to create the margin for average customer lifetime.

Users who liked