Cohort Analysis is a subset of behavioral analytics that breaks your customers’ data into related groups that share common characteristics within a time frame. This allows a company to clearly see patterns across the lifecycle of a customer, rather than slicing blindly across all customers without accounting for the natural cycle the customer undergoes. It highlights patterns that may not be visible from a macro perspective.

Hovering over the matrix/table, a narrative report page tooltip tells you the story of your cohort in a natural language whenever you move horizontally or vertically.

The entire Cohort Analysis in this project was done in Power BI using DAX.

I’ve combined different analytic models such as RFM, Customer Lifetime Value (CLV), the Probability of a customer to be “Alive” the following year, and a personalized Recommendation System using Collaborative Filtering for the implicit dataset.

The power of this project is the combination of all these techniques applied to customers, especially in this time of digital transformation where the customer is more than ever at the center of any successful business. It gives you a 360-degree view of your customers and therefore provides a 360-degree solution with concrete actions you can take.

This simple and single-page report quickly shows how the company is performing in terms of sales growth; whether the company is meeting its sales objectives or not. You can choose to compare the actual sales against the prior year or against the target.

The cumulative sales (YTD) and its variance, as well as the monthly trend and its variance let you quickly identify where the company underperforms at the higher level and at the granular/product level, and what was the reason for such a performance.

Currently, there is a lot of documentation on recommender systems where customers explicitly give their feedback or rate products. Sometimes, such reviews or ratings do not truly reflect the real intention of a customer. But, what about implicit feedback where a customer does not explicitly rate your service? That’s what this project addresses.

This solution mimics an online retail company like Amazon in that from the moment you add a product to the cart, the system recommends you both popular and relevant/similar products with a score assigned to each recommendation. We can apply the same technique at the customer level. Based on the purchase history, the system can recommend personalized/relevant products that the customer is likely to purchase, rather than just recommending popular products to all customers.

How are you calculating your Customer Lifetime Value? Are you using the same formulas regardless of your industry (contractual/non-contractual)?

In this project, I applied the results of over 25 years of research from Peter Fader, also known as the Frances and Pei-Yuan Chia Professor of Marketing at The Wharton School of the University of Pennsylvania.

I used Python to accurately calculate the Customer Lifetime Value in the non-contractual setting, considering the probability of a customer being “Alive” or not. Instead of using Formulas, I recommend using this probabilistic/statistical approach as it is more realistic and allows your company to accurately plan and significantly reduce the Customer Acquisition Costs.

This project helps to answer the question “where are my customers right now?” It highlights the importance of running and monitoring customer segmentation over time. Simply put, I run an RFM analysis monthly or quarterly to see if your loyal customers are at risk of churning and how much money you’re likely to lose. The combination of a powerful visual and filters allows you to quickly identify At-Risk customers and immediately take actions to prevent losing them.