Live Agent Resource Management System

Machine Learning Model based Live Traffic Control System designed to reduce live agent costs for after-sales customer support in an automatic and smart way. The system identifies the users to hide live agent access by predicting that their issues are more likely to be solved without agent assistance.

The model can dynamically control the main entry point to live agent access from chatbot session, which usually contributes to 90% of the overall traffic to live agent

My Role

Defined business problems and prodcut roadmap by working with Regional Operational Teams
Identified user pain points through user interviews with local tperational Teams across 10+ transcontinental markets
Collaborated with Data Scientists and Data Analysts to scope out the user data, build and constantly iterate the model
Designed the A/B testing and suceesfully rolled out to all markets on a global scale

Overview

Established as the largest E-commerce platform in South East Asia while operating globally, Shopee has a cross-continental 90 million monthly active users, generating billions of revenue. Chatbot has become an essential channel for after-sales customer service with a 5 million weekly session to better server customer inquiries with a constantly improving end-to-end shopping experience at Shopee.

What Challenges Customer Service is Facing?

Mission & Business Objective

Shopee’s mission is to use the transformative power of technology and and to change the world for the better by providing a platform to connect buyers and sellers within one community. The business objective of Shopee’s Customer Service, specifically, is to create the best after-sales experience for customer with accessible assistance and satisfying solution to increase customer satisfication.

Customer Segmentation

Customer Pain Points

3RD PARTY VENDOR

Solution

Build a machine learning model to dynamically split the traffic from live agent by directing users with inquires to be solved without agent assistance to chatbot.