Customer segmentation is the process of dividing customers into groups based on common characteristics so companies can market to each group effectively and appropriately.
In business-to-business marketing, a company might segment customers according to a wide range of factors, including:
- Industry
- Number of employees
- Products previously purchased from the company
- Location
In business-to-consumer marketing, companies often segment customers according to demographics that include:
- Age
- Gender
- Marital Status
- Location (urban, surburban, rural)
- Income
The aim of this project is to determine the types of customers (target customers) who can easily convert into loyal customers so that the marketing team can make an informed decision about their approach.
The results obtained from this project can be used by various stakeholders within the bank such as
- Marketing team
- Sales team
- Customer success team
Segmentation allows marketers to better tailor their marketing efforts to various audience subsets. Those efforts can relate to both communications and product development. Specifically, segmentation helps a company:
- Create and communicate targeted marketing messages that will resonate with specific groups of customers, but not with others (who will receive messages tailored to their needs and interests, instead).
- Select the best communication channel for the segment, which might be email, social media posts, radio advertising, or another approach, depending on the segment.
- Focus on the most profitable customers.
- Improve customer service.
Python Version:3.9.12
Packages:Pandas,Numpy,Scikit learn,Matplotlib,Seaborn,Imblearn, Collection, Intertools
Data Source:https://www.kaggle.com/datasets/vjchoudhary7/customer-segmentation-tutorial-in-python
The datasets used in this project were downloaded from https://www.kaggle.com/datasets/vjchoudhary7/customer-segmentation-tutorial-in-python. I then read the two csv files using the pd.read_csv() command.
After downloading the data, I needed to clean it up so that it was usable for our model. In our case, the dataset did not contain any missing values and the data was of the correct format.
The data only consists of 4 numerical variables, and 1 categorical variable. I looked at different distributions for both the categorical and numeric data. Below are highlights from the data visualization section
I will first implement the kmeans algorithm on the data set having 200 rows and 5 columns dataset and see how it works. The data contains the age of the customers, annual income, spending score as well as customer ID. I will be looking at the different relationships between these attributes :
- Age and Spending Score
- Annual Income and Spending Score
- Age and Annual Score
From now on we will be using sklearn implementation of kmeans. Few thing to note here:
- n_init is the number of times of running the kmeans with different centroid’s initialization. The result of the best one will be reported.
- tol is the within-cluster variation metric used to declare convergence.
- The default of init is k-means++ which is supposed to yield a better results than just random initialization of centroids.
Next, we'll calculate the sum of squared error using the elbow method as well as the silhouette score using different number of clusters then plot the results.
Contrary to supervised learning where we have the ground truth to evaluate the model’s performance, clustering analysis doesn’t have a solid evaluation metric that we can use to evaluate the outcome of different clustering algorithms. Moreover, since kmeans requires k as an input and doesn’t learn it from data, there is no right answer in terms of the number of clusters that we should have in any problem.
In order to evaluate the model performance for the different the different clusters, two clustering metrics were used:
- Elbow method
- Silhouette analysis
Elbow Method
Elbow method gives us an idea on what a good k number of clusters would be based on the sum of squared distance (SSE) between data points and their assigned clusters’ centroids. We pick k at the spot where SSE starts to flatten out and forming an elbow. We’ll use the geyser dataset and evaluate SSE for different values of k and see where the curve might form an elbow and flatten out. Below are the plots obtained using this method
Silhouette analysis
Silhouette analysis can be used to determine the degree of separation between clusters. For each sample:
- Compute the average distance from all data points in the same cluster (ai).
- Compute the average distance from all data points in the closest cluster (bi).
- Compute the coefficient:
The coefficient can take values in the interval [-1, 1].
- If it is 0 –> the sample is very close to the neighboring clusters.
- It it is 1 –> the sample is far away from the neighboring clusters.
- It it is -1 –> the sample is assigned to the wrong clusters.
Therefore, we want the coefficients to be as big as possible and close to 1 to have a good clusters.
For clustering problems, we often choose the result that has the highest silhouette score, since it measures the seperation between the clusters. In general, the higher the score, the far apart the clusters are, which is often good since we do not want to have an overlapping within the clusters that might lead to incorrect grouping of data. In our case, we can see from the silho 69CB uette score table that the best relationship is between the Age and Spending Score (1-100) having an Silhouette score of 0.499739 when compared to the other relationships. Therefore, the marketing team of the mall should focus on this group when conducting customer segmentation.
- In this python machine learning project, I built a clustering model using the kmeans algorithm. This project aims to determine the types of customers (target customers) who can easily convert into loyal customers so that the marketing team can make an informed decision about their approach.
- The age vs annual income scatter plot shows that the highest earning customers are males in their early 30's. Overall, the distribution of the data within the scatter plot shows that majority of the customers are between the age of 20 to 50 and earn between 20 - 80 k$
- The age vs spending score scatter plot shows that the highest spending customers are between the age of 20 to 40. Based on this data, the marketing team can focus on these individuals when promoting their products/services in order to increase sales. Overall, the distribution of the data within the scatter plot shows that majority of the customers with a spending score of 60 and higher are less than 40 years while the customers who are 40 years or older have a spending score of 60 or less.
- The annual income vs spending score scatter plot shows that the majority of the customers earn between 40 - 70 k and have a spending score between 40 - 60. In addition, the highest spending customers earns 20 k while the lowest spending customer earns 80 k.
- It is evident from the correlation table that the age and the annual income have a negative correlation with a correlation coefficient of -0.01. Similarly the age and the spending score have a strong negative correlation with a correlation coefficient of -0.33. This shows that as the age increases both the annual income and spending score decrease. Lastly, the annual income ans the spending score has a positive correlation having a correlation coefficient of 0.01. This shows that as the annual income increases, the spending score is more likely to increase.
- The age vs spending score scatter plot shows 4 clusters representing different customer groups. The group of interest in this case green cluster since most of the customers in that have a high spending score compared to the other groups, thus it is advisable for the marketing team of the mall to focus on these customers in order to increase customer base within the mall.
- The above annual income vs spending score scatter plot shows 4 clusters representing different customer groups. The group of interest in this case yellow cluster since most of the customers in that cluster have both a high spending score and annual income compared to the other groups, thus it is advisable for the marketing team of the mall to focus on these customers in order to increase customer base within the mall.
- For clustering problems, we often choose the result that has the highest silhouette score, since it measures the seperation between the clusters. In general, the higher the score, the far apart the clusters are, which is often good since we do not want to have an overlapping within the clusters that might lead to incorrect grouping of data. In our case, we can see from the silhouette score table that the best relationship is between the Age and Spending Score (1-100) having an Silhouette score of 0.499739 when compared to the other relationships. Therefore, the marketing team of the mall should focus on this group when conducting customer segmentation.