Data analytics is about prescribing solutions to improve future events while data analysis is geared toward defining past and current situations. I use tools including Python to create algorithms for machine learning.
Below is a case situation showing how to predict the churn rate of customers for a SaaS company.
How Does a Company Reduce Churn Rate?
Churn is when a client or employee leaves a company. These business stakeholder relationships tend to be seen as more valuable to a company than new business relationships simply because retaining previous clients tends to be easier than converting brand new ones.
As you can see from the scatter plot below, churn rate is highly dependent on years with the company. When a client was with the company for four years or less the probability of that stakeholder leaving was the highest.
Data Pipeline Modeling
I created a data pipeline by first cleaning the csv data and identifying the statistical measures. I partitioned 80% of the data to train the machine learning algorithm and the other 20% to test against the final prediction.
Characteristics of Customer
I discovered certain characteristics could be used to predict churn probability. Three distinct characteristics were used to train the statistical model and thus predict if a customer may churn.
The charts below show the churn data as well as predicted data of a customer’s time with company, age, and time since sign-up. The model was then applied to all current stakeholder in the company and then sorted to reveal the top client most likely to churn.
Customer Lifetime Value
Another useful metric was to predict the the total revenue value generated for each customer. This data was used to generate a sorted list of the biggest potential clients.
The average CLV came out to be $12.5 with an average lifetime of 42.5 months.
Machine Learning with Python
By using the power of Python i created a model and trained it to predict future churn rate probability of a small sample size (however a larger sample is ideal for more accurate predictions).
Would you like to know more about sales analytics? Please get in touch with me here.
