Customer segmentation in sales entails grouping customers of a wholesale company based on their annual buying patterns using K-Means and DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithms. The study employs clustering as a stand-alone tool to get insight into the wholesale company client’s buying patterns distribution and identify customer groups with similar characteristics. The result identifies groups of similar customer purchase characteristics and helps guide the company on targeted marketing or promotions.
Exploratory data analysis and preparation techniques were applied to the dataset, which revealed that the company’s major annual sale is from fresh products and groceries. The company’s clients can be roughly clustered into two groups and there are exclusive clients with higher purchasing patterns. Lastly, the attributes were scaled to yield a reliable clustering result.
KMeans
The K-Means algorithm identified five clusters, indicating that the wholesale company clients/customers can be grouped into five based on their buying patterns with a silhouette score of 0.368.
Figure 1: KMeans clusters
Cluster interpretations:
o Cluster 4: regular purchases, minimal spending
o Cluster 3: very high purchase, high spending
o Cluster 2: minimal purchases, very high spending
o Cluster 1: a bit more regular purchases, more spending
o Cluster 0: regular purchases, minimal spending
DBSCAN
DBSCAN algorithm identified only two clusters, indicating that the wholesale company clients can be segmented into two groups with similar characteristics; the regular buyers and the binge buyers with a silhouette score of 0.748699.
Figure 2: DBSCAN clusters
Cluster interpretations:
o Cluster 0: regular purchases, regular spending
o Cluster -1: very high purchase, high spending
The two clustering algorithms, K-Means and DBSCAN, yielded a positive silhouette coefficient in identifying clusters of wholesale company customers with similar buying characteristics. DBSCAN had a much higher score of 0.75, which makes it better than the K-Means algorithm with 0.37. Although the silhouette coefficient tends to be usually higher for density-based clusters, the DBSCAN result for the customer segmentation is best suited for the wholesale company data because it yielded a more realistic result with reliable clusters. With this result, the company would better understand its customers and discover opportunities for better marketing and product development.