Customer behaviour and value analysis and ml clustering

I attempt to understand customer behaviour patterns on an ecommerce sales dataset for better business strategies by using a variety of analysis methods including cohort analysis of purchase activity, RFM value analysis, and unsupervised machine learning.

Cohort analysis

• Creating cohorts based on month of first purchase
• Organising data by cohort and month
• Analysing different metrics: Retention rate, Mean cohort order quantity, Mean cohort order price

Recency, frequency, monetary value analysis

• Wrangled data to calculate metrics at the client level
• Recency (days since last purchase)
• Frequency (how many purchases in the last 12 months)
• Monetary value (How much spent in the last 12 months
• Assigning metric scores based on percentiles to analyse general behaviour
• Added product breadth metric

Clustering

• Removal of outlier clients
• Transforming and scaling data
• Clustering with kmeans (Elbow method, SSE scores, silhouette scores)
• Business actions based on kmeans cluster analysis
• Clustering with density-based scan (Selection of min samples in a cluster, and distance where to pick neighbours)
• Business actions based on dbscan clusters analysis