Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 13

Customer Lifetime

Value (CLV) Prediction


What is CLV ?
CLV

Customer lifetime value (CLV) is a useful indicator that depicts the total
sum of money a company anticipates a customer will spend on goods and
services over the course of the customer relationship. It assigns a monetary
value to the importance of a particular customer or market segment to the
company and the amount that should be spent maintaining that relationship.

In order to reach their marketing objectives and build a following of


devoted patrons who will promote their brand, businesses can benefit from
insights about their clients and establish smart marketing strategies. In other
words, it assists businesses in putting more of an emphasis on long-term
profitability than on immediate profits, ensuring that both
CLV Elements and Parameters
Calculating Customer Lifetime Value (CLV)

Steps for calculating customer lifetime value are-

1. Calculate average purchase value by dividing the company's total revenue in a time period by the number of purchases in
that time period

2. Calculate the average purchase frequency rate by diving the number of purchases in that time period by the number of
unique customers who made purchases during that time period

3. Calculate customer value by multiplying average purchase value by average purchase frequency rate.

4. Calculate average customer lifespan by finding the average of the number of years a customer continues to buy from the
company

5. Calculate customer lifetime value by multiplying customer value by average customer lifespan. This gives the revenue that
can be expected to be generated during the course of the relationship of a customer with the company.
What Is RFM?
Recency, Frequency, Monetary Value (RFM):

Recency
Recency refers to how recent a customer's last purchase was.
Customers who have made a recent purchase, still have the product and
brand on their minds and are most likely to make a repeat purchase.

Frequency
Frequency is how often the customer makes
purchases, which can help you identify repeat
customers.

Monetary
Monetary value refers to how much a customer spends within a given
period. It's always important to consider because it can tell you a few
things about consumer behavior.
Whats in the code ?
Steps reuired to find CLV

Required libraries are imported


Importing Basic
including the dataset of the invoice.
Libraries and Data

Understanding the EDA & Data Preprocessing


Dataset

We used 8 months of data, calculated


RFM and used it for predicting next 16 Lifetime Value
months. Prediction

Building Machine Predicted CLV using Linear


Regression Model and K-Mean
Learning Model
Clustering.
Result

Biggest cluster we have is cluster 0 which is


78.6% of the total base. If we blindly say,
every customer belongs to cluster 0, then
our accuracy would be 78.6%.

82% vs 78.6% tell us that our machine


learning model is a useful one but needs
some improvement for sure. We should find
out where the model is failing.

The feature set has been ready to


calculate 16 months CLV for each
customer which we used for training
our model.
CONCLUSION
Conclusion and Improvement

CLV is a measure you should pay special attention to because it enables you to detect numerous facets of your organization.
Your business will benefit from your efforts to increase the CLV rate.

Daily customer growth is desirable, but letting go of existing clients who are gradually leaving your business is detrimental to
your own interests. If you want to reduce churn, you might attempt to keep your current clients happy. It should become
second nature to you to regularly calculate your CLV.

As an example, for cluster 0 (Low CLV), if model tells us this customer belongs to cluster 0,87 out of 100 will be correct
(precision). And the model successfully identifies 95% of actual cluster 0 customers (recall).

We really need to improve the model for other clusters. For example, this model can not detect other class CLV customers.
Possible actions to improve those points:

• Adding more features and improve feature engineering


• Apply hyper parameter tuning to current model
• Add more data to the model if possible
Thank you for
listening and watching

You might also like