Project Overview

HOME

PROJECT OVERVIEW

PROJECT FINDINGS

PROJECT DOCUMENTATION

PROJECT MANAGEMENT

ANLY482 HOMEPAGE

Background		Data Source		Methodology

Introduction

For any entity to conduct business online, they must be able to accept payments from customers, and this involves a third-party to facilitate the online transfer of funds. Such entities are called electronic Payment Gateways , and this includes PayPal, Stripe, and our project sponsor.

When a customer makes a purchase on a merchant’s website, our sponsor helps to process the credit card payment. This is done by transferring key information between the payment portal (e.g. merchant’s website) and the merchant’s registered bank account. For each successful transaction that our sponsor processes, they apply a commission, called a Merchant Discount Rate (MDR) , from the transaction.

Business Motivations and Objectives

While our sponsor handles millions of transactions across the globe, the company has not fully been able to derive any conclusive analysis from its transaction data.

By providing our team with their data, our sponsor hopes to gather a deeper understanding of it. The project objectives include developing meaningful insights by performing exploratory data analysis. Using the results drawn from the findings, recommendations will be formulated to aid in future business decision making.

Meetings with Project Coordinator, Project Sponsor and Team

Our team aims to meet our Project Coordinator, Professor Kam Tin Seong on average of at least once every week in order to ensure that we are progressing on the right track. We also met our project sponsor once a month to update them on our progress and validate our analysis, while our internal team meetings are held at least twice a week. All minutes can be found under "Project Documentation".

Project Objectives

We aim to study the complexity of the online transactions carried out by our sponsor’s merchants. In particular, we hope to assess the performance of our sponsor’s merchants, by using number of approved transactions per merchant as a benchmark.

This project shows how we use funnel plots and line of best-fit graphs to compare the performance among entities in a group in an unbiased manner. Though both funnel plots and line of best-fit graphs are prevalent in medical research , they are under-utilized in e-commerce analytics. In our case study, the line of best-fit graphs provide a better model fit. By combining features of funnel plot with line of best-fit, we can identify star and laggard merchants.

Moving on, we want to study the characteristics of transactions carried out by key merchants. We use logistic fit to establish any correlation between independent variables and the number of approved transactions. For independent variables that exhibit correlation, we conduct further analysis by utilizing decision trees to map expected approved-to-rejected transaction ratio for key merchants.

In summary, we show how exploratory and confirmatory techniques can be used as source of business intelligence - setting performance benchmarks for each merchant and improving their approved-to-rejected transaction ratio.

Scope of Project

Task	Description
Gather Requirements	Confirm and gather sponsor requirements
Initial Research and Preparation	Conduct preliminary data exploration and define project objectives and scope
Project Proposal	Prepare project proposal and Wikipage
Data Exploration and Preparation	Ensuring that data is clean and can be analysed using analytical software; We have done data preparation, which include – Interactive Binning and finding the top reason code descriptions. After which, we conducted exploratory data analysis.
Model Building	Through research findings and experience, we will attach suitable models to our data. We have used Interactive Binning, Line of Fit and Time Series Analysis to generate insights.
Project Revision (Mid-Term)	Assisted by RDP through obtaining feedback during our sponsor meeting.
Mid-term Preparation	Prepare mid-term report, presentation and Wikipage.
Model Validation and Refinement	Conduct independent sample t-tests (e.g. Ensure the results are similar when attached to different years of study) and refine analysis of data.
Insights and Recommendations	Create visualisation from analysis results and formulate recommendations for our sponsor.
Project Revision	Assisted by sponsor through obtaining feedback during our sponsor meeting; Align our final deliverables with sponsor requirements.
Final Preparation	Prepare abstract and full paper, final Wikipage update and final presentation with necessary deliverables.