Photo by Clemens van Lay on Unsplash

An End-to-End Journey of a Multinational Predictive Conversion Likelihood Model for Marketplaces

Overview

Introduction

Project Big Picture (Image by Author)

Table of Contents

Step 1: Data Exploration and Discovery

Step 2: Feature Engineering

Time Frames for Training Data Generations (Image by Author)

Predictors

Response Variable

Step 3: Modelling and Parameter Optimisation

Model Selection (Source: h2o.ai)
Shifted Training and Validation Datasets Generation (Image by Author)
Feature Importance List (Image by Author)

Step 4: Results

Lift-Gain Chart Bucket Analysis

Bucket Analysis: Comparison of Buckets CRs vs. Overall CR vs Intuitive Way CR (Image by Author)

Feature Importances

Feature Importance Lists of Kijiji Canada’s Buyer and Seller Models (Image by Author)
Feature Importance Lists of eBay Kleinanzeigen’s Buyer and Seller Models (Image by Author)

Partial Dependence Plots

Partial Dependence Plot: Impact of ‘The Most Visited Category’-feature on User Conversion-response variable (Image by Author)
Partial Dependence Plot: Impact of ‘Number of Different Categories Visited’-feature on User Conversion-response variable (Image by Author)

Bucket Transitions

Transitions Between Buckets in 7 Days (Image by Author)
Transitions Between Buckets in 30 Days (Image by Author)

Step 5: Production Jobs

Production Pipeline: Jobs Flow (Image by Author)

Step 6: Behavioural User Segments

Predictive Behavioural User Clusters (Image by Author)

Step 7: A Sample Use-Case

Experiment: How The New Product Performs On Different User Segments (Image by Author)

Conclusion

Author Signature

Currently Amsterdam-based and working at Ebay. Senior Data Scientist with M. Sc degree in Machine Learning and 7 years of professional experience.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store