Project 2: Customer Segmentation
Background: The client contacted me, because they needed an effective method to target their customer. A good customer segmentation can help a company to develop more focused strategies. This improves customer retention-rates and leads to increased sales. Marketing campaigns can be run more effectively with the result of a higher ROI (return of investment).
I have used a clustering algorithm (k-means model) to group customers into different segments. Grouping customers into segments (cluster) makes it easier to understand each group’s collective behavior. This will allow the company to handle each cluster differently — i.e. offer exclusive discount coupons to some of them.
The first step of the clustering algorithm is to select the customer characteristics that would help the algorithm categorising them into different clusters. Some characteristics might be:
- Number of orders per month
- Average order value
- Average value per item
- Number of items ordered
- Number of employees
- Customer Demographics
- Customer location
- Number of active membership months
- Number of Customer service contact points
In our case, we have chosen the Customers Age and Yearly Order Value as input into the algorithm.
Plotting the data into a Scatter Plot allows us to get a rough understanding of the distribution of the data points.
Next, we are using the Elbow Technique, to determine the optimal number of clusters for the algorithm. The Elbow Technique confirms that three clusters is the optimal number for our algorithm.
After training a K-Means Clustering Model, the algorithm has grouped the customers as follows:
Every customer got assigned to a specific cluster (see column ‘Cluster’).
The machine learning model can now be used to group new customer instantly.
Result: The company has use this information to optimize their content strategy and create targeted marketing campaigns.
Project 3: Customer Churn Probability
Background: The client asked me to predict the churn probability of each of their customer. Knowing whether a customer is at a high risk of churn is a very valuable information for the company, as it enables them to pro-actively approach high-risk churn customer and thereby, potentially prevent a churn.
Work in progress. – To be continued soon –