Customer Lifetime Revenue (CLR) Explanation



Once we have established the recency, frequency, and monetary value summary statistics, we can input these into algorithms to estimate the total customer lifetime revenue. This encompasses the sum of their previous spending up until the time the model is run.

Additionally, we forecast the customer's likelihood of re-engagement and repurchasing based on their behavioral patterns. The sum of these two estimates creates the customer lifetime value, representing the anticipated lifetime engagement and spending of a customer with the company.





This measure can be used to evaluate the ROI of customer acquisition costs, measure the overall increase or decrease in the company's customer lifetime revenue, and track changes in customer lifetime revenue over time.

The customer lifetime revenue can further be used to determine the appropriate amount of investment in new customer acquisition. By analyzing each customer's data, we train models to identify patterns in their behavior and predict their likelihood of repurchasing. Our goal is to accurately identify customers who are likely to repurchase and those who are not and assign weights accordingly to assess the accuracy of our models.

Calculating the probability of a customer's re-engagement or repurchase is a crucial aspect of customer lifetime value estimation. A representative customer who made multiple purchases from a store exemplifies this concept. The likelihood of their engagement with the store at the time of purchase is 100%, but as time passes, this probability diminishes. With increasing temporal distance from their last purchase, the likelihood of re-engagement also declines.





An important aspect of customer lifetime value is determining the probability of a customer reengaging or repurchasing from a store. For instance, if a company scores the reengagement probability for all its customers, it can conduct a targeted campaign for customers with a moderate probability of reengagement, say 40% to 60%. This group can then be divided into an A/B component, with the treatment marketing campaign or email campaign only given to the A group while the B group is not engaged.





At the conclusion of the campaign, a causal difference can be measured between the two groups. It is expected that the A group's performance would be what was predicted, with the delta between the two groups representing the impact of the A/B group. The probability of reengagement is a powerful segmentation tool that can be spiced by different segments within the group.

The general idea is to use it as a predictive tool, with a good idea of how the two buckets will perform in the test, and then evaluate the causal impact of the test by assessing the delta on the other side.