From Data to Decisions: Enhancing User Conversions Through Predictive Analysis

Our client was interested in more accurately identifying which users would convert (i.e., purchase) after visiting their website. 

The purpose was to identify those who showed a higher likelihood of conversion but had not yet converted. By identifying these users, the sales team could spend their time more efficiently by assisting these customers, and leadership would have better insight into which behaviors were strongly associated with conversions. 

The approach was to manually identify a range of website interactions, such as product views, list creations, and bid placements. Subsequently, they assigned subjective weights to each of these behaviors to estimate a user's inclination to convert. 

Solution 

Methodology: 

• Data Preparation: 
  • We gathered data on website behaviors and sales outcomes. 
  • Tailored metrics to our specific analysis needs.
• Descriptive Statistics: 
  • Managed missing data and outliers. 
  • Removed correlated behaviors to prevent multicollinearity. 
  • Incorporated weights due to an imbalanced dataset.
• Modelling 
  • Regression based Predictive Modelling to identify predictive behaviors. 
  • Selected the optimal model through iterative testing and comparison metrics.

Results: Our model, based on a year's historical data, effectively predicted conversions in recent data with notable accuracy.