Description This project aims to analyze customer behavior and segment customers based on RFM (Recency, Frequency, Monetary) using real sales data from a UK-based retailer. The dataset contains transactions occurring between 01/12/2010 and 09/12/2011. The company primarily sells unique all-occasion gifts, and many customers are wholesalers.
The dataset used in this project is sourced from the UCI Machine Learning Repository.(https://archive.ics.uci.edu/dataset/352/online+retail)
-
Data Cleaning: Convert attributes to the proper data type: Convert 'Monetary', 'Frequency', and 'Recency' attributes to appropriate numerical data types. Convert 'InvoiceDate' to datetime datatype for proper time-based analysis.
-
Scaling: Scale the data as necessary for model building.
-
Model Building: Utilize the K-means clustering algorithm to segment customers based on RFM.
-
Elbow Method: Determine the optimal number of clusters using the elbow method.
-
Result Analysis: Analyze the results of the clustering algorithm to understand customer segments.
-
Business Strategy: *Identify opportunities for increasing sales in specific customer clusters.
-
Develop competitive positioning, pricing strategies, sales and marketing efforts, promotions, and bundling strategies tailored to each cluster.
-
Consider customer segments with growth potential for future expansion.
-
Buying Analysis: The UK alone contributes to 82% of total revenue, indicating the importance of the UK market for the retailer.
-
Geographic Expansion: Consider expanding to other countries with significant sales potential, such as Netherlands, EIRE, Germany, and France, for future geographic expansion.
- Cluster 1 may already be dominated by others. Increase sales in clusters 0 & 2 through suitable competitive positioning, pricing strategy, cohesive sales & marketing efforts, promotions, bundling, etc. Additionally, focus on increasing sales in cluster 1 through suitable marketing efforts, promotions, and bundling.
- Customer segments 0 & 2 have opportunities for growth and future expansion.
- Contributions to improve the analysis techniques, expand insights, or optimize the code are welcome.